Release v1.0: HPRC-trained 35/21-feature calibrated SV+STR models

README.md

@@ -1,109 +1,82 @@
 ---
 license: mit
-tags:
-  - genomics
-  - structural-variants
-  - short-tandem-repeats
-  - random-forest
-  - variant-calling
 library_name: sklearn
+tags:
+- genomics
+- structural-variants
+- short-tandem-repeats
+- variant-calling
+- confidence-calibration
+- random-forest
+pipeline_tag: tabular-classification
 ---
 
-# SVSTR-Score: long-read-guided confidence scoring for short-read SV and STR genotypes
+# SVSTR-Score (v1.0) — long-read-guided confidence scoring for short-read SV/STR calls
 
-Per-class random-forest models that assign a calibrated **confidence score (CS ∈ [0,1])**
-and a four-tier operating point to each **short-read** structural-variant (SV) or
-short-tandem-repeat (STR) call, learned from paired short-/long-read genomes. The primary SV
-model is the 23-feature caller-output random forest; a caller-independent sequence-only model
-is provided as a portable secondary option.
-The long-read genotype is used only to build the training label; **inference needs
-short-read caller output only**.
+Two per-class **RandomForest + isotonic calibrator** models that assign a
+**calibrated confidence score** `CS ∈ [0,1]` and a four-tier operating point to
+each short-read **structural-variant (SV)** or **short-tandem-repeat (STR)** call.
+The long-read genotype is used **only to build the training label**, so inference
+needs short-read caller output only.
 
-## Files
-| File | Description |
-|---|---|
-| `sv_model_v13_parents.joblib` | SV random forest (23 caller-output features; **primary SV model / external-validation headline**) |
-| `str_model_v13_parents.joblib` | STR random forest (25 features) |
-| `seqonly/` | **caller-independent sequence-only SV model (svspr, 11 features) — portable secondary option** for callers whose fields differ from training. `seqonly/model/svspr_v14_seq.pkl` + the `svspr` inference package (VCF + reference → CS/tier; tiers HIGH≥0.9/MOD≥0.7/WARN≥0.5) + example. 11 features = SV length/log-length, SVTYPE one-hot, ±100 bp flank GC/AT/inner-GC, motif-2/3 counts; no caller fields |
-| `str_locus_lookup.parquet` | per-locus historical concordance (`locus_conc_rate`) keyed by (chrom, pos); 163,726 loci |
-| `sv_config.json` / `str_config.json` | feature order, tier thresholds, lookup metadata |
-| `feature_manifest.json` | definition of every SV/STR feature |
-| `tier_thresholds.json` | tier cut-offs + override rule |
-| `score_svstr.py` | inference entry point |
-| `requirements.txt` | pinned dependencies (**scikit-learn==1.5.1**) |
-| `example/` | real example calls from the study cohort (input features + scored output) |
+`CS = isotonic_calibrator( RandomForest.predict_proba(X)[:, 1] )` = P(the
+short-read call is concordant with the long-read truth).
 
-> Models were trained with **scikit-learn 1.5.1**; load with the same version to
-> avoid `InconsistentVersionWarning` and guarantee identical results.
+| | SV | STR |
+|---|---|---|
+| weights | `sv_model.joblib` + `sv_calibrator.joblib` | `str_model.joblib` + `str_calibrator.joblib` |
+| features | 35 (`feature_builder.py`) | 21 |
+| short-read caller | Manta | ExpansionHunter |
+| long-read truth (label only) | sawfish | TRGT |
+| training cohort | 208 HPRC paired genomes | 208 HPRC |
 
-## Quick start
-```bash
-pip install -r requirements.txt
-# SV
-python score_svstr.py --variant sv  --model-dir . --features example/sv_features.tsv  --out sv_scored.tsv
-# STR (locus_conc_rate / locus_in_lookup are filled here from str_locus_lookup.parquet)
-python score_svstr.py --variant str --model-dir . --features example/str_features.tsv --out str_scored.tsv
-```
-Input is a table extracted from the caller VCF (e.g. with `bcftools query`) holding
-the features in `*_config.json`. For STR, supply the 23 caller-output features plus
-`chrom,pos`; the script joins the catalogue lookup to add `locus_conc_rate` and
-`locus_in_lookup`.
+## Tiers
+`HIGH CS≥0.70` · `MODERATE 0.50–0.70` · `WARNING 0.30–0.50` · `LOW <0.30`.
+Tiers are buckets of the calibrated CS (no heuristic overrides). **HIGH** is the
+candidate-triage tier.
 
-`example/` holds real calls drawn from the study cohort (`*_features.tsv`) and
-their scored output (`*_scored.tsv`), spanning the tier range. They double as a
-reproducibility check: the SV set includes precise PASS calls scored HIGH and
-override cases (`is_imprecise==1` or `filter_pass==0`) demoted to LOW; the STR
-set includes high-concordance loci scored HIGH and a feature-clean but
-historically discordant locus scored LOW, illustrating that `locus_conc_rate`
-dominates the STR score.
+## Intended use
+Triage / down-weight *emitted* short-read SV & STR calls by their probability of
+matching long-read truth, **without long-read sequencing every sample**. Not a
+variant caller; does not recover missed variants; does not replace long-read
+sequencing for complete discovery.
 
-## Confidence tiers
-| Tier | Cut |
-|---|---|
-| HIGH | CS ≥ 0.70 |
-| MODERATE | 0.50 ≤ CS < 0.70 |
-| WARNING | 0.30 ≤ CS < 0.50 |
-| LOW | CS < 0.30 |
+## Performance
+**Internal — 5-fold GroupKFold by sample (out-of-fold), 208 HPRC genomes**
 
-**Override.** Deterministic quality-demotion rules in each `*_config.json` (`rules`)
-are applied on top of the CS tier by `score_svstr.py`; any rule triggered →
-`override_target` regardless of CS. **SV → LOW** (QUAL<20, GQ<15, total_alt_support≤2,
-vaf_estimate<0.15, is_imprecise==1); **STR → WARNING** (locus_coverage<20, is_low_depth==1,
-support_type_a2≤1, total_support_a2<5, ci_width_a2>3, allele_balance<0.3). **Only the
-MODERATE→HIGH range is a monotone precision ladder; LOW and WARNING are not rank-ordered
-by precision.** Use the **HIGH tier as a candidate-triage filter**.
+| | AUROC | AUPRC | per-sample AUROC median |
+|---|---|---|---|
+| SV | 0.950 | 0.951 | 0.981 |
+| STR | 0.834 | 0.886 | 0.835 |
 
-## Intended use & performance
-- **Sample-LOSO cross-validation** (143 unrelated parents): SV F1 = **0.9308**, STR F1 = **0.9960**.
-- **Within-cohort generalization** — re-scored on 3,608 unseen short-read genomes without retraining;
-  the score distribution and per-1-Mb genomic reliability map reproduce (training↔unseen Spearman ρ = **0.90**).
-- **External multi-sample validation — 194 HPRC genomes** (sequence-aware Truvari labels; long-read truth Sawfish/SV, TRGT/STR):
-  - SV (this 23-feature model, the primary external object): per-genome median AUROC **0.927**, above raw
-    QUAL (0.741), GQ (0.476) and read-depth (0.671) and the applicable dedicated tools (duphold, Paragraph),
-    most clearly for deletions and insertions (AUPRC 0.948, expected calibration error 0.053). Robust across
-    the tested callers (median **0.915** on an independent DRAGEN reanalysis). The companion
-    **caller-independent sequence-only model** transfers at a lower ceiling (median AUROC **0.851**).
-  - STR (this model): per-genome median AUROC **0.912**; in-catalogue 0.909 vs out-of-catalogue 0.688 (catalogue-bound reliability atlas).
-- **GIAB HG002** (single-sample supplementary check): HIGH-tier SV precision **97.7 %** (consensus v0.6).
+**External — 295 ASD genomes, applied unchanged (no retraining)**
 
-## Limitations (read before use)
-- **External validation is multi-sample (194 HPRC genomes; see above).** Outputs remain
-  candidate triage, not validated clinical calls. The 23-feature caller-output model is robust
-  across the tested Manta-family callers (Manta v1.6, DRAGEN), but it consumes caller-specific
-  fields and is untested on structurally different callers (e.g. DELLY, GATK-SV); for those use
-  the caller-independent sequence-only model (lower ceiling).
-- **STR scoring is a per-locus reliability atlas.** ~94 % of STR model importance is
-  `locus_conc_rate`; removing it drops cross-validated AUROC 0.998 → 0.62. The score
-  does **not** extend to STR loci outside the 163,726-locus catalogue (out-of-catalogue
-  loci get a conservative score by construction), and only the HIGH tier transfers
-  across callers — lower-tier ranking does not.
-- Trained on a single ancestry and pipeline (Illumina DRAGEN + Manta/ExpansionHunter;
-  PacBio HiFi + Sawfish/TRGT, GRCh38). Behaviour under other pipelines is untested.
-- `.joblib` is a pickle: load only from trusted sources and with the pinned versions.
+| | AUROC | observed concordance LOW / WARN / MOD / HIGH |
+|---|---|---|
+| SV | 0.891 | 12% / 35% / 50% / **86%** |
+| STR | 0.831 | 16% / 40% / 60% / **85%** |
 
-## Citation
-Kim W*, Yeom K*, et al. SVSTR-Score: long-read-guided confidence scoring for
-short-read SV and STR genotypes. (manuscript in preparation.)
+Calibrated tiers stay monotone and meaningful on the external cohort; isotonic
+calibration improves Brier (SV 0.077→0.074, STR 0.167→0.159).
 
-## Licence
-MIT.
+## Usage
+```bash
+pip install -U huggingface_hub scikit-learn==1.7.1
+hf download khyeom/SVSTR-Score sv_model.joblib sv_calibrator.joblib sv_config.json score_svstr.py feature_builder.py --local-dir svstr/
+# features from a single-sample VCF, then score:
+python svstr/feature_builder.py --vcf sample.manta.vcf.gz --caller manta \
+  --fasta GRCh38.fa --giab-dir giab_prepared --repeatmasker rmsk_class.bed.gz -o feats.tsv
+python svstr/score_svstr.py --variant sv --model-dir svstr --features feats.tsv --out scored.tsv
+```
+Code, training pipeline and reproduction scripts: https://github.com/khyeom0608/SVSTR-Score
+
+## Limitations
+- Relies on **caller support / breakpoint-confidence fields** (PR/SR, CIPOS/CIEND,
+  VAF, GQ, depth). On **merged or heavily filtered call sets that drop these**,
+  scores deflate and tiers are unreliable (rank-discrimination degrades only
+  moderately, but calibration breaks).
+- Strongest for **DEL/INS**; DUP/BND are mostly down-weighted/triaged.
+- STR scoring is **bound to the caller's genotyped locus set**.
+
+## Citation
+Kim W\*, Yeom K\*, et al. *SVSTR-Score* (manuscript in preparation). Licence: MIT.

example/str_features.tsv

@@ -1,7 +1,51 @@
-chrom	pos	repcn_a1	repcn_a2	ru_length	ref_count	delta_from_ref_a1	delta_from_ref_a2	ci_width_a1	ci_width_a2	adsp_a1	adsp_a2	adfl_a1	adfl_a2	adir_a1	adir_a2	total_support_a1	total_support_a2	spanning_frac_a1	spanning_frac_a2	locus_coverage	allele_balance	support_type_a1	support_type_a2	is_low_depth
-chr12	10728964	8.0	8.0	2	19	-11.0	-11.0	7.0	146.0	3.0	3.0	2.0	2.0	0.0	0.0	5.0	5.0	0.6	0.6	23.5197	1.0	2.0	2.0	1
-chr13	70022282	11.0	18.0	2	11	0.0	7.0	0.0	23.0	34.0	1.0	5.0	9.0	0.0	0.0	39.0	10.0	0.8717948717948718	0.1	40.7512	0.2564102564102564	2.0	2.0	0
-chr4	98868919	6.0	6.0	3	6	0.0	0.0	0.0	0.0	27.0	27.0	9.0	9.0	0.0	0.0	36.0	36.0	0.75	0.75	37.1579	1.0	2.0	2.0	0
-chr6	88530656	6.0	5.0	4	6	0.0	-1.0	0.0	0.0	13.0	8.0	8.0	4.0	0.0	0.0	21.0	12.0	0.6190476190476191	0.6666666666666666	37.5663	0.5714285714285714	2.0	2.0	0
-chr8	37146469	14.0	18.0	2	14	0.0	4.0	0.0	0.0	15.0	15.0	14.0	18.0	0.0	0.0	29.0	33.0	0.5172413793103449	0.4545454545454545	43.7729	0.8787878787878788	2.0	2.0	0
-chr8	50668495	9.0	15.0	3	14	-5.0	1.0	0.0	0.0	8.0	6.0	16.0	18.0	0.0	0.0	24.0	24.0	0.3333333333333333	0.25	37.7296	1.0	2.0	2.0	0
+sample	caller	variant_ID	is_pass	motif_len	ref_copynum	gt_repcn_max	gt_repcn_min	expansion_over_ref	repci_width_max	spanning_reads	flanking_reads	inrepeat_reads	locus_depth	gt_hom	ref_tract_bp	spanning_frac	allele_vs_readlen	motif_is_homopolymer	gc_flank	entropy_flank	in_segdup	in_difficult	flank_lowmap
+HG00097	expansionhunter	chr1:165954:165962	1	4.0	2.0	2.0	1.0	0.0	0.0	25.0	2.0	0.0	34.5405	0.0	8.0	0.9259259259259259	0.05333333333333334	0	0.297029702970297	1.8767240669197331	1	1	1
+HG00097	expansionhunter	chr1:370632:370648	1	4.0	4.0	4.0	3.0	0.0	1.0	30.0	14.0	0.0	40.3784	0.0	16.0	0.6818181818181818	0.10666666666666667	0	0.37623762376237624	1.7475795513453003	1	1	1
+HG00097	expansionhunter	chr1:832736:832781	1	5.0	9.0	10.0	10.0	1.0	0.0	44.0	104.0	0.0	42.8919	1.0	45.0	0.2972972972972973	0.3333333333333333	0	0.36633663366336633	1.7496871374624745	1	1	0
+HG00097	expansionhunter	chr1:932613:932621	1	4.0	2.0	1.0	1.0	-1.0	0.0	88.0	4.0	0.0	49.3784	1.0	8.0	0.9565217391304348	0.02666666666666667	0	0.6138613861386139	1.93240710844821	0	1	0
+HG00097	expansionhunter	chr1:1010481:1010497	1	4.0	4.0	3.0	3.0	-1.0	0.0	62.0	20.0	0.0	35.1081	1.0	16.0	0.7560975609756098	0.08	0	0.13861386138613863	1.5081706750248076	0	1	0
+HG00097	expansionhunter	chr1:1052514:1052528	1	2.0	7.0	8.0	7.0	1.0	0.0	40.0	4.0	0.0	46.3784	0.0	14.0	0.9090909090909091	0.10666666666666667	0	0.46534653465346537	1.8443047122008371	0	1	0
+HG00097	expansionhunter	chr1:1063661:1063681	1	2.0	10.0	10.0	9.0	0.0	0.0	39.0	26.0	0.0	51.0811	0.0	20.0	0.6	0.13333333333333333	0	0.5643564356435643	1.7595456304802637	0	1	0
+HG00097	expansionhunter	chr1:1265603:1265649	1	2.0	23.0	24.0	23.0	1.0	0.0	19.0	44.0	0.0	38.3514	0.0	46.0	0.30158730158730157	0.32	0	0.48514851485148514	1.4319496839589465	0	1	0
+HG00097	expansionhunter	chr1:1431026:1431058	1	8.0	4.0	5.0	4.0	1.0	0.0	25.0	65.0	0.0	53.2703	0.0	32.0	0.2777777777777778	0.26666666666666666	0	0.7128712871287128	1.8074915272734042	0	1	0
+HG00097	expansionhunter	chr1:1585949:1585976	1	3.0	9.0	16.0	15.0	7.0	0.0	23.0	57.0	0.0	44.4324	0.0	27.0	0.2875	0.32	0	0.33663366336633666	1.870380810485376	0	1	0
+HG00097	expansionhunter	chr1:1653825:1653863	1	2.0	19.0	21.0	21.0	2.0	1.0	22.0	28.0	0.0	34.8649	1.0	38.0	0.44	0.28	0	0.49504950495049505	1.861174864442864	1	1	0
+HG00097	expansionhunter	chr1:1752908:1752935	1	3.0	9.0	10.0	9.0	1.0	0.0	39.0	31.0	0.0	49.4595	0.0	27.0	0.5571428571428572	0.2	0	0.6435643564356436	1.8970057796754802	0	1	0
+HG00097	expansionhunter	chr1:1762759:1762791	1	4.0	8.0	9.0	8.0	1.0	0.0	38.0	32.0	0.0	40.5405	0.0	32.0	0.5428571428571428	0.24	0	0.38613861386138615	1.9107593839500945	0	1	0
+HG00097	expansionhunter	chr1:1769969:1770011	1	6.0	7.0	5.0	5.0	-2.0	0.0	36.0	38.0	0.0	44.7568	1.0	42.0	0.4864864864864865	0.2	0	0.32673267326732675	1.680234216638318	0	1	0
+HG00097	expansionhunter	chr1:1776461:1776473	1	4.0	3.0	2.0	2.0	-1.0	0.0	74.0	12.0	0.0	38.7568	1.0	12.0	0.8604651162790697	0.05333333333333334	0	0.45544554455445546	1.8897699414887892	0	0	0
+HG00097	expansionhunter	chr1:1812407:1812429	1	2.0	11.0	12.0	12.0	1.0	0.0	64.0	20.0	0.0	39.6486	1.0	22.0	0.7619047619047619	0.16	0	0.40594059405940597	1.8347498835870788	0	1	0
+HG00097	expansionhunter	chr1:1845824:1845872	1	2.0	24.0	23.0	20.0	-1.0	0.0	37.0	34.0	0.0	43.3784	0.0	48.0	0.5211267605633803	0.30666666666666664	0	0.40594059405940597	1.8265902247349062	0	1	0
+HG00097	expansionhunter	chr1:1891654:1891658	1	2.0	2.0	4.0	4.0	2.0	0.0	60.0	4.0	0.0	40.7027	1.0	4.0	0.9375	0.05333333333333334	0	0.7128712871287128	1.854785288155074	0	1	0
+HG00097	expansionhunter	chr1:1904424:1904448	1	4.0	6.0	8.0	8.0	2.0	0.0	44.0	32.0	0.0	39.2432	1.0	24.0	0.5789473684210527	0.21333333333333335	0	0.3069306930693069	1.7889467525708116	0	1	0
+HG00097	expansionhunter	chr1:1948412:1948428	1	2.0	8.0	10.0	10.0	2.0	0.0	78.0	20.0	0.0	37.7838	1.0	16.0	0.7959183673469388	0.13333333333333333	0	0.33663366336633666	1.908726104503474	0	1	0
+HG00097	expansionhunter	chr1:2003928:2003940	1	6.0	2.0	3.0	2.0	1.0	0.0	51.0	19.0	0.0	45.8919	0.0	12.0	0.7285714285714285	0.12	0	0.7920792079207921	1.7281983769914673	0	1	0
+HG00097	expansionhunter	chr1:2012279:2012309	1	5.0	6.0	8.0	6.0	2.0	0.0	23.0	59.0	0.0	40.8649	0.0	30.0	0.2804878048780488	0.26666666666666666	0	0.2871287128712871	1.8125908839274099	0	1	0
+HG00097	expansionhunter	chr1:2018334:2018389	1	5.0	11.0	11.0	10.0	0.0	0.0	29.0	43.0	0.0	34.7838	0.0	55.0	0.4027777777777778	0.36666666666666664	0	0.36633663366336633	1.8582276266039703	0	1	0
+HG00097	expansionhunter	chr1:2112675:2112691	1	4.0	4.0	5.0	5.0	1.0	0.0	68.0	26.0	0.0	48.4054	1.0	16.0	0.723404255319149	0.13333333333333333	0	0.37623762376237624	1.7586128527134226	0	1	0
+HG00097	expansionhunter	chr1:2173095:2173103	1	4.0	2.0	2.0	1.0	0.0	0.0	44.0	8.0	0.0	41.6757	0.0	8.0	0.8461538461538461	0.05333333333333334	0	0.504950495049505	1.9625707732852478	0	0	0
+HG00097	expansionhunter	chr1:2207909:2207955	1	2.0	23.0	25.0	20.0	2.0	0.0	25.0	65.0	0.0	39.8919	0.0	46.0	0.2777777777777778	0.3333333333333333	0	0.5841584158415841	1.7558294024999062	0	1	0
+HG00097	expansionhunter	chr1:2219851:2219871	1	5.0	4.0	4.0	3.0	0.0	0.0	21.0	9.0	0.0	41.7568	0.0	20.0	0.7	0.13333333333333333	0	0.4752475247524752	1.8595044600250858	0	1	0
+HG00097	expansionhunter	chr1:2345829:2345855	1	2.0	13.0	14.0	10.0	1.0	0.0	24.0	25.0	0.0	47.1892	0.0	26.0	0.4897959183673469	0.18666666666666668	0	0.5247524752475248	1.9180855002676722	0	1	0
+HG00097	expansionhunter	chr1:2371211:2371241	1	3.0	10.0	13.0	10.0	3.0	0.0	27.0	40.0	0.0	50.1892	0.0	30.0	0.40298507462686567	0.26	0	0.5148514851485149	1.8305650137637772	0	1	0
+HG00097	expansionhunter	chr1:2431330:2431346	1	2.0	8.0	9.0	8.0	1.0	0.0	41.0	32.0	0.0	48.6486	0.0	16.0	0.5616438356164384	0.12	0	0.5841584158415841	1.9166005274309357	0	1	0
+HG00097	expansionhunter	chr1:2435454:2435489	1	7.0	5.0	5.0	4.0	0.0	0.0	39.0	30.0	0.0	39.0	0.0	35.0	0.5652173913043478	0.23333333333333334	0	0.5742574257425742	1.9335830591930787	0	1	0
+HG00097	expansionhunter	chr1:2449499:2449535	1	4.0	9.0	9.0	8.0	0.0	0.0	17.0	48.0	0.0	40.9459	0.0	36.0	0.26153846153846155	0.24	0	0.5247524752475248	1.8838136852474037	0	1	0
+HG00097	expansionhunter	chr1:2508612:2508630	1	2.0	9.0	11.0	9.0	2.0	0.0	35.0	18.0	0.0	49.0541	0.0	18.0	0.660377358490566	0.14666666666666667	0	0.3564356435643564	1.933754201968858	0	1	0
+HG00097	expansionhunter	chr1:2566825:2566869	1	4.0	11.0	12.0	11.0	1.0	0.0	19.0	124.0	0.0	47.8378	0.0	44.0	0.13286713286713286	0.32	0	0.4158415841584158	1.9433008996133987	0	1	0
+HG00097	expansionhunter	chr1:2580534:2580555	1	3.0	7.0	8.0	8.0	1.0	0.0	42.0	32.0	0.0	33.8108	1.0	21.0	0.5675675675675675	0.16	0	0.37623762376237624	1.9016529508428732	0	1	0
+HG00097	expansionhunter	chr1:2600684:2600708	1	4.0	6.0	6.0	3.0	0.0	0.0	35.0	17.0	0.0	37.9459	0.0	24.0	0.6730769230769231	0.16	0	0.44554455445544555	1.8855340778294845	0	1	0
+HG00097	expansionhunter	chr1:2782508:2782518	1	2.0	5.0	6.0	6.0	1.0	0.0	80.0	8.0	0.0	40.2162	1.0	10.0	0.9090909090909091	0.08	0	0.31683168316831684	1.539579197931996	0	1	0
+HG00097	expansionhunter	chr1:2784990:2785010	1	10.0	2.0	4.0	3.0	2.0	0.0	33.0	28.0	0.0	40.1351	0.0	20.0	0.5409836065573771	0.26666666666666666	0	0.693069306930693	1.849024777034605	0	1	0
+HG00097	expansionhunter	chr1:2824891:2824941	1	2.0	25.0	23.0	22.0	-2.0	0.0	20.0	28.0	0.0	41.1081	0.0	50.0	0.4166666666666667	0.30666666666666664	0	0.5148514851485149	1.736792200782953	0	1	0
+HG00097	expansionhunter	chr1:2847577:2847593	1	8.0	2.0	2.0	1.0	0.0	0.0	45.0	4.0	0.0	41.3514	0.0	16.0	0.9183673469387755	0.10666666666666667	0	0.6039603960396039	1.893728291589638	0	1	0
+HG00097	expansionhunter	chr1:2899043:2899075	1	4.0	8.0	7.0	7.0	-1.0	0.0	54.0	32.0	0.0	41.5946	1.0	32.0	0.627906976744186	0.18666666666666668	0	0.44554455445544555	1.7976208220041983	0	1	0
+HG00097	expansionhunter	chr1:2952909:2952927	1	2.0	9.0	12.0	12.0	3.0	0.0	60.0	22.0	0.0	36.7297	1.0	18.0	0.7317073170731707	0.16	0	0.6138613861386139	1.4586949056098804	0	1	0
+HG00097	expansionhunter	chr1:3089915:3089939	1	4.0	6.0	8.0	8.0	2.0	0.0	72.0	34.0	0.0	49.6216	1.0	24.0	0.6792452830188679	0.21333333333333335	0	0.44554455445544555	1.9438900377229549	0	1	0
+HG00097	expansionhunter	chr1:3109083:3109116	1	3.0	11.0	11.0	9.0	0.0	0.0	24.0	49.0	0.0	41.3514	0.0	33.0	0.3287671232876712	0.22	0	0.32673267326732675	1.8491066059426058	0	1	0
+HG00097	expansionhunter	chr1:3152348:3152380	1	4.0	8.0	6.0	6.0	-2.0	0.0	58.0	10.0	0.0	44.9189	1.0	32.0	0.8529411764705882	0.16	0	0.43564356435643564	1.6103551420752904	0	1	0
+HG00097	expansionhunter	chr1:3175983:3176011	1	4.0	7.0	8.0	8.0	1.0	0.0	50.0	42.0	0.0	45.2432	1.0	28.0	0.5434782608695652	0.21333333333333335	0	0.5643564356435643	1.8171755888564571	0	1	0
+HG00097	expansionhunter	chr1:3225530:3225574	1	2.0	22.0	21.0	17.0	-1.0	0.0	20.0	40.0	0.0	42.973	0.0	44.0	0.3333333333333333	0.28	0	0.5742574257425742	1.7700475389123633	0	1	0
+HG00097	expansionhunter	chr1:3240056:3240106	1	5.0	10.0	5.0	4.0	-5.0	0.0	31.0	16.0	0.0	39.6486	0.0	50.0	0.6595744680851063	0.16666666666666666	0	0.5544554455445545	1.0599256723036075	0	1	0
+HG00097	expansionhunter	chr1:3273438:3273450	1	2.0	6.0	4.0	4.0	-2.0	0.0	64.0	12.0	0.0	36.7297	1.0	12.0	0.8421052631578947	0.05333333333333334	0	0.5247524752475248	1.856780293719988	0	1	0
+HG00097	expansionhunter	chr1:3273817:3273851	1	2.0	17.0	17.0	15.0	0.0	0.0	25.0	34.0	0.0	36.973	0.0	34.0	0.423728813559322	0.22666666666666666	0	0.4752475247524752	1.78618987974027	0	1	0

example/str_scored.tsv

@@ -1,7 +1,51 @@
-chrom	pos	repcn_a1	repcn_a2	ru_length	ref_count	delta_from_ref_a1	delta_from_ref_a2	ci_width_a1	ci_width_a2	adsp_a1	adsp_a2	adfl_a1	adfl_a2	adir_a1	adir_a2	total_support_a1	total_support_a2	spanning_frac_a1	spanning_frac_a2	locus_coverage	allele_balance	support_type_a1	support_type_a2	is_low_depth	locus_conc_rate	locus_in_lookup	CS	tier
-chr12	10728964	8.0	8.0	2	19	-11.0	-11.0	7.0	146.0	3.0	3.0	2.0	2.0	0.0	0.0	5.0	5.0	0.6	0.6	23.5197	1.0	2.0	2.0	1	0.0	1	0.005	WARNING
-chr13	70022282	11.0	18.0	2	11	0.0	7.0	0.0	23.0	34.0	1.0	5.0	9.0	0.0	0.0	39.0	10.0	0.8717948717948718	0.1	40.7512	0.2564102564102564	2.0	2.0	0	0.0	1	0.005	WARNING
-chr4	98868919	6.0	6.0	3	6	0.0	0.0	0.0	0.0	27.0	27.0	9.0	9.0	0.0	0.0	36.0	36.0	0.75	0.75	37.1579	1.0	2.0	2.0	0	1.0	1	1.0	HIGH
-chr6	88530656	6.0	5.0	4	6	0.0	-1.0	0.0	0.0	13.0	8.0	8.0	4.0	0.0	0.0	21.0	12.0	0.6190476190476191	0.6666666666666666	37.5663	0.5714285714285714	2.0	2.0	0	0.0	1	0.0	LOW
-chr8	37146469	14.0	18.0	2	14	0.0	4.0	0.0	0.0	15.0	15.0	14.0	18.0	0.0	0.0	29.0	33.0	0.5172413793103449	0.4545454545454545	43.7729	0.8787878787878788	2.0	2.0	0	0.7062937062937062	1	0.97	HIGH
-chr8	50668495	9.0	15.0	3	14	-5.0	1.0	0.0	0.0	8.0	6.0	16.0	18.0	0.0	0.0	24.0	24.0	0.3333333333333333	0.25	37.7296	1.0	2.0	2.0	0	0.986013986013986	1	1.0	HIGH
+sample	caller	variant_ID	is_pass	motif_len	ref_copynum	gt_repcn_max	gt_repcn_min	expansion_over_ref	repci_width_max	spanning_reads	flanking_reads	inrepeat_reads	locus_depth	gt_hom	ref_tract_bp	spanning_frac	allele_vs_readlen	motif_is_homopolymer	gc_flank	entropy_flank	in_segdup	in_difficult	flank_lowmap	CS_raw	CS	tier
+HG00097	expansionhunter	chr1:165954:165962	1	4.0	2.0	2.0	1.0	0.0	0.0	25.0	2.0	0.0	34.5405	0.0	8.0	0.925925925925926	0.0533333333333333	0	0.297029702970297	1.8767240669197327	1	1	1	0.7378	0.8546	HIGH
+HG00097	expansionhunter	chr1:370632:370648	1	4.0	4.0	4.0	3.0	0.0	1.0	30.0	14.0	0.0	40.3784	0.0	16.0	0.6818181818181818	0.1066666666666666	0	0.3762376237623762	1.7475795513453003	1	1	1	0.3337	0.3967	WARNING
+HG00097	expansionhunter	chr1:832736:832781	1	5.0	9.0	10.0	10.0	1.0	0.0	44.0	104.0	0.0	42.8919	1.0	45.0	0.2972972972972973	0.3333333333333333	0	0.3663366336633663	1.7496871374624745	1	1	0	0.9082	0.9876	HIGH
+HG00097	expansionhunter	chr1:932613:932621	1	4.0	2.0	1.0	1.0	-1.0	0.0	88.0	4.0	0.0	49.3784	1.0	8.0	0.9565217391304348	0.0266666666666666	0	0.6138613861386139	1.93240710844821	0	1	0	0.4569	0.5696	MODERATE
+HG00097	expansionhunter	chr1:1010481:1010497	1	4.0	4.0	3.0	3.0	-1.0	0.0	62.0	20.0	0.0	35.1081	1.0	16.0	0.7560975609756098	0.08	0	0.1386138613861386	1.5081706750248076	0	1	0	0.1697	0.1673	LOW
+HG00097	expansionhunter	chr1:1052514:1052528	1	2.0	7.0	8.0	7.0	1.0	0.0	40.0	4.0	0.0	46.3784	0.0	14.0	0.9090909090909092	0.1066666666666666	0	0.4653465346534653	1.8443047122008367	0	1	0	0.1755	0.1752	LOW
+HG00097	expansionhunter	chr1:1063661:1063681	1	2.0	10.0	10.0	9.0	0.0	0.0	39.0	26.0	0.0	51.0811	0.0	20.0	0.6	0.1333333333333333	0	0.5643564356435643	1.7595456304802637	0	1	0	0.0987	0.0615	LOW
+HG00097	expansionhunter	chr1:1265603:1265649	1	2.0	23.0	24.0	23.0	1.0	0.0	19.0	44.0	0.0	38.3514	0.0	46.0	0.3015873015873015	0.32	0	0.4851485148514851	1.4319496839589465	0	1	0	0.1248	0.0944	LOW
+HG00097	expansionhunter	chr1:1431026:1431058	1	8.0	4.0	5.0	4.0	1.0	0.0	25.0	65.0	0.0	53.2703	0.0	32.0	0.2777777777777778	0.2666666666666666	0	0.7128712871287128	1.807491527273404	0	1	0	0.6394	0.7505	HIGH
+HG00097	expansionhunter	chr1:1585949:1585976	1	3.0	9.0	16.0	15.0	7.0	0.0	23.0	57.0	0.0	44.4324	0.0	27.0	0.2875	0.32	0	0.3366336633663366	1.870380810485376	0	1	0	0.8984	0.9819	HIGH
+HG00097	expansionhunter	chr1:1653825:1653863	1	2.0	19.0	21.0	21.0	2.0	1.0	22.0	28.0	0.0	34.8649	1.0	38.0	0.44	0.28	0	0.495049504950495	1.861174864442864	1	1	0	0.6052	0.7199	HIGH
+HG00097	expansionhunter	chr1:1752908:1752935	1	3.0	9.0	10.0	9.0	1.0	0.0	39.0	31.0	0.0	49.4595	0.0	27.0	0.5571428571428572	0.2	0	0.6435643564356436	1.89700577967548	0	1	0	0.7839	0.8853	HIGH
+HG00097	expansionhunter	chr1:1762759:1762791	1	4.0	8.0	9.0	8.0	1.0	0.0	38.0	32.0	0.0	40.5405	0.0	32.0	0.5428571428571428	0.24	0	0.3861386138613861	1.9107593839500945	0	1	0	0.9107	0.988	HIGH
+HG00097	expansionhunter	chr1:1769969:1770011	1	6.0	7.0	5.0	5.0	-2.0	0.0	36.0	38.0	0.0	44.7568	1.0	42.0	0.4864864864864865	0.2	0	0.3267326732673267	1.680234216638318	0	1	0	0.8992	0.9829	HIGH
+HG00097	expansionhunter	chr1:1776461:1776473	1	4.0	3.0	2.0	2.0	-1.0	0.0	74.0	12.0	0.0	38.7568	1.0	12.0	0.8604651162790697	0.0533333333333333	0	0.4554455445544554	1.8897699414887887	0	0	0	0.8386	0.9299	HIGH
+HG00097	expansionhunter	chr1:1812407:1812429	1	2.0	11.0	12.0	12.0	1.0	0.0	64.0	20.0	0.0	39.6486	1.0	22.0	0.7619047619047619	0.16	0	0.4059405940594059	1.8347498835870788	0	1	0	0.2376	0.2636	LOW
+HG00097	expansionhunter	chr1:1845824:1845872	1	2.0	24.0	23.0	20.0	-1.0	0.0	37.0	34.0	0.0	43.3784	0.0	48.0	0.5211267605633803	0.3066666666666666	0	0.4059405940594059	1.8265902247349064	0	1	0	0.7847	0.8853	HIGH
+HG00097	expansionhunter	chr1:1891654:1891658	1	2.0	2.0	4.0	4.0	2.0	0.0	60.0	4.0	0.0	40.7027	1.0	4.0	0.9375	0.0533333333333333	0	0.7128712871287128	1.854785288155074	0	1	0	0.1098	0.0718	LOW
+HG00097	expansionhunter	chr1:1904424:1904448	1	4.0	6.0	8.0	8.0	2.0	0.0	44.0	32.0	0.0	39.2432	1.0	24.0	0.5789473684210527	0.2133333333333333	0	0.3069306930693069	1.7889467525708116	0	1	0	0.6801	0.8002	HIGH
+HG00097	expansionhunter	chr1:1948412:1948428	1	2.0	8.0	10.0	10.0	2.0	0.0	78.0	20.0	0.0	37.7838	1.0	16.0	0.7959183673469388	0.1333333333333333	0	0.3366336633663366	1.908726104503474	0	1	0	0.3487	0.4133	WARNING
+HG00097	expansionhunter	chr1:2003928:2003940	1	6.0	2.0	3.0	2.0	1.0	0.0	51.0	19.0	0.0	45.8919	0.0	12.0	0.7285714285714285	0.12	0	0.7920792079207921	1.7281983769914673	0	1	0	0.7141	0.8336	HIGH
+HG00097	expansionhunter	chr1:2012279:2012309	1	5.0	6.0	8.0	6.0	2.0	0.0	23.0	59.0	0.0	40.8649	0.0	30.0	0.2804878048780488	0.2666666666666666	0	0.2871287128712871	1.8125908839274096	0	1	0	0.9037	0.985	HIGH
+HG00097	expansionhunter	chr1:2018334:2018389	1	5.0	11.0	11.0	10.0	0.0	0.0	29.0	43.0	0.0	34.7838	0.0	55.0	0.4027777777777778	0.3666666666666666	0	0.3663366336633663	1.8582276266039703	0	1	0	0.8468	0.9378	HIGH
+HG00097	expansionhunter	chr1:2112675:2112691	1	4.0	4.0	5.0	5.0	1.0	0.0	68.0	26.0	0.0	48.4054	1.0	16.0	0.723404255319149	0.1333333333333333	0	0.3762376237623762	1.7586128527134226	0	1	0	0.371	0.4441	WARNING
+HG00097	expansionhunter	chr1:2173095:2173103	1	4.0	2.0	2.0	1.0	0.0	0.0	44.0	8.0	0.0	41.6757	0.0	8.0	0.8461538461538461	0.0533333333333333	0	0.504950495049505	1.962570773285248	0	0	0	0.6207	0.7324	HIGH
+HG00097	expansionhunter	chr1:2207909:2207955	1	2.0	23.0	25.0	20.0	2.0	0.0	25.0	65.0	0.0	39.8919	0.0	46.0	0.2777777777777778	0.3333333333333333	0	0.5841584158415841	1.7558294024999062	0	1	0	0.7201	0.8388	HIGH
+HG00097	expansionhunter	chr1:2219851:2219871	1	5.0	4.0	4.0	3.0	0.0	0.0	21.0	9.0	0.0	41.7568	0.0	20.0	0.7	0.1333333333333333	0	0.4752475247524752	1.8595044600250856	0	1	0	0.7221	0.8402	HIGH
+HG00097	expansionhunter	chr1:2345829:2345855	1	2.0	13.0	14.0	10.0	1.0	0.0	24.0	25.0	0.0	47.1892	0.0	26.0	0.4897959183673469	0.1866666666666666	0	0.5247524752475248	1.918085500267672	0	1	0	0.3912	0.4822	WARNING
+HG00097	expansionhunter	chr1:2371211:2371241	1	3.0	10.0	13.0	10.0	3.0	0.0	27.0	40.0	0.0	50.1892	0.0	30.0	0.4029850746268656	0.26	0	0.5148514851485149	1.8305650137637768	0	1	0	0.796	0.8924	HIGH
+HG00097	expansionhunter	chr1:2431330:2431346	1	2.0	8.0	9.0	8.0	1.0	0.0	41.0	32.0	0.0	48.6486	0.0	16.0	0.5616438356164384	0.12	0	0.5841584158415841	1.916600527430936	0	1	0	0.2435	0.2759	LOW
+HG00097	expansionhunter	chr1:2435454:2435489	1	7.0	5.0	5.0	4.0	0.0	0.0	39.0	30.0	0.0	39.0	0.0	35.0	0.5652173913043478	0.2333333333333333	0	0.5742574257425742	1.9335830591930787	0	1	0	0.7708	0.8778	HIGH
+HG00097	expansionhunter	chr1:2449499:2449535	1	4.0	9.0	9.0	8.0	0.0	0.0	17.0	48.0	0.0	40.9459	0.0	36.0	0.2615384615384615	0.24	0	0.5247524752475248	1.883813685247404	0	1	0	0.8208	0.9087	HIGH
+HG00097	expansionhunter	chr1:2508612:2508630	1	2.0	9.0	11.0	9.0	2.0	0.0	35.0	18.0	0.0	49.0541	0.0	18.0	0.660377358490566	0.1466666666666666	0	0.3564356435643564	1.933754201968858	0	1	0	0.6421	0.7584	HIGH
+HG00097	expansionhunter	chr1:2566825:2566869	1	4.0	11.0	12.0	11.0	1.0	0.0	19.0	124.0	0.0	47.8378	0.0	44.0	0.1328671328671328	0.32	0	0.4158415841584158	1.9433008996133987	0	1	0	0.7715	0.8778	HIGH
+HG00097	expansionhunter	chr1:2580534:2580555	1	3.0	7.0	8.0	8.0	1.0	0.0	42.0	32.0	0.0	33.8108	1.0	21.0	0.5675675675675675	0.16	0	0.3762376237623762	1.9016529508428728	0	1	0	0.8111	0.9036	HIGH
+HG00097	expansionhunter	chr1:2600684:2600708	1	4.0	6.0	6.0	3.0	0.0	0.0	35.0	17.0	0.0	37.9459	0.0	24.0	0.6730769230769231	0.16	0	0.4455445544554455	1.8855340778294845	0	1	0	0.8571	0.9491	HIGH
+HG00097	expansionhunter	chr1:2782508:2782518	1	2.0	5.0	6.0	6.0	1.0	0.0	80.0	8.0	0.0	40.2162	1.0	10.0	0.9090909090909092	0.08	0	0.3168316831683168	1.539579197931996	0	1	0	0.0884	0.052	LOW
+HG00097	expansionhunter	chr1:2784990:2785010	1	10.0	2.0	4.0	3.0	2.0	0.0	33.0	28.0	0.0	40.1351	0.0	20.0	0.5409836065573771	0.2666666666666666	0	0.693069306930693	1.849024777034605	0	1	0	0.4897	0.5908	MODERATE
+HG00097	expansionhunter	chr1:2824891:2824941	1	2.0	25.0	23.0	22.0	-2.0	0.0	20.0	28.0	0.0	41.1081	0.0	50.0	0.4166666666666667	0.3066666666666666	0	0.5148514851485149	1.736792200782953	0	1	0	0.6974	0.8162	HIGH
+HG00097	expansionhunter	chr1:2847577:2847593	1	8.0	2.0	2.0	1.0	0.0	0.0	45.0	4.0	0.0	41.3514	0.0	16.0	0.9183673469387756	0.1066666666666666	0	0.6039603960396039	1.893728291589638	0	1	0	0.523	0.6287	MODERATE
+HG00097	expansionhunter	chr1:2899043:2899075	1	4.0	8.0	7.0	7.0	-1.0	0.0	54.0	32.0	0.0	41.5946	1.0	32.0	0.627906976744186	0.1866666666666666	0	0.4455445544554455	1.7976208220041985	0	1	0	0.8667	0.9568	HIGH
+HG00097	expansionhunter	chr1:2952909:2952927	1	2.0	9.0	12.0	12.0	3.0	0.0	60.0	22.0	0.0	36.7297	1.0	18.0	0.7317073170731707	0.16	0	0.6138613861386139	1.4586949056098804	0	1	0	0.0848	0.0452	LOW
+HG00097	expansionhunter	chr1:3089915:3089939	1	4.0	6.0	8.0	8.0	2.0	0.0	72.0	34.0	0.0	49.6216	1.0	24.0	0.6792452830188679	0.2133333333333333	0	0.4455445544554455	1.9438900377229549	0	1	0	0.9023	0.985	HIGH
+HG00097	expansionhunter	chr1:3109083:3109116	1	3.0	11.0	11.0	9.0	0.0	0.0	24.0	49.0	0.0	41.3514	0.0	33.0	0.3287671232876712	0.22	0	0.3267326732673267	1.8491066059426056	0	1	0	0.8115	0.9036	HIGH
+HG00097	expansionhunter	chr1:3152348:3152380	1	4.0	8.0	6.0	6.0	-2.0	0.0	58.0	10.0	0.0	44.9189	1.0	32.0	0.8529411764705882	0.16	0	0.4356435643564356	1.6103551420752904	0	1	0	0.4377	0.5542	MODERATE
+HG00097	expansionhunter	chr1:3175983:3176011	1	4.0	7.0	8.0	8.0	1.0	0.0	50.0	42.0	0.0	45.2432	1.0	28.0	0.5434782608695652	0.2133333333333333	0	0.5643564356435643	1.8171755888564567	0	1	0	0.6944	0.8135	HIGH
+HG00097	expansionhunter	chr1:3225530:3225574	1	2.0	22.0	21.0	17.0	-1.0	0.0	20.0	40.0	0.0	42.973	0.0	44.0	0.3333333333333333	0.28	0	0.5742574257425742	1.770047538912363	0	1	0	0.6244	0.7366	HIGH
+HG00097	expansionhunter	chr1:3240056:3240106	1	5.0	10.0	5.0	4.0	-5.0	0.0	31.0	16.0	0.0	39.6486	0.0	50.0	0.6595744680851063	0.1666666666666666	0	0.5544554455445545	1.0599256723036077	0	1	0	0.2903	0.3499	WARNING
+HG00097	expansionhunter	chr1:3273438:3273450	1	2.0	6.0	4.0	4.0	-2.0	0.0	64.0	12.0	0.0	36.7297	1.0	12.0	0.8421052631578947	0.0533333333333333	0	0.5247524752475248	1.856780293719988	0	1	0	0.2091	0.2239	LOW
+HG00097	expansionhunter	chr1:3273817:3273851	1	2.0	17.0	17.0	15.0	0.0	0.0	25.0	34.0	0.0	36.973	0.0	34.0	0.423728813559322	0.2266666666666666	0	0.4752475247524752	1.78618987974027	0	1	0	0.2897	0.3499	WARNING

example/sv_features.tsv

@@ -1,8 +1,51 @@
-chrom	pos	qual	gq	pr_ref	pr_alt	sr_ref	sr_alt	vf_ref	vf_alt	total_alt_support	sr_pr_ratio	vaf_estimate	strand_bias_fs	cipos_width	ciend_width	homlen	svlen_abs	svtype_DEL	svtype_INS	svtype_DUP	svtype_BND	is_imprecise	filter_pass	n_filter_flags
-chr10	42254638	182	6	2.0	0.0	3.0	5.0	5.0	5.0	5.0	5.0	0.5	9.7			0.0	50.0	1	0	0	0	0	1	0
-chr12	12391920	999	281	11.0	20.0	3.0	14.0	12.0	30.0	34.0	0.6666666666666666	0.7083333333333334	7.186			0.0		0	0	0	1	0	1	0
-chr16	8562503	693	693	54.0	22.0	36.0	9.0	69.0	26.0	31.0	0.391304347826087	0.256198347107438	0.0			0.0	1781.0	0	0	1	0	0	1	0
-chr17	23232202	310	38	0.0	0.0	0.0	21.0	0.0	21.0	21.0	21.0	1.0	0.0	5.0	5.0	5.0		0	1	0	0	0	0	1
-chr3	163746379	999	118	0.0	10.0	0.0	33.0	0.0	43.0	43.0	3.0	1.0	0.0	8.0		8.0	306.0	1	0	0	0	0	1	0
-chr6	32648154	82	82	7.0	5.0			7.0	5.0	5.0		0.4166666666666667		415.0	238.0	0.0	4156.0	1	0	0	0	1	1	0
-chr8	91664843	999	148	0.0	28.0	0.0	36.0	0.0	56.0	64.0	1.2413793103448276	1.0	0.0	15.0		15.0	316.0	0	1	0	0	0	1	0
+sample	caller	variant_ID	is_pass	svtype_DEL	svtype_DUP	svtype_INS	svtype_INV	svtype_BND	svlen_log	cipos_width	ciend_width	is_imprecise	pe_support	sr_support	total_support	vaf	gt_hom	gq	qual_norm	local_depth	gc_min	gc_max	entropy_min	microhom_max	in_segdup_either	in_segdup_both	in_difficult_either	in_difficult_both	in_lowmap_either	in_tandem_either	in_Alu_either	in_L1_either	in_SVA_either	in_LTR_either	frac_span_repeat	n_neighbors	nn_log_dist
+HG00097	manta	chr1:861885:BND:None	1	0	0	0	0	1	-99999.0	31.0	-99999.0	0	5.0	7.0	12.0	0.15384615384615385	0	182.0	182.0	43.0	0.297029702970297	0.297029702970297	1.6350564503142975	-99999.0	1	1	1	1	0	1	0	0	0	0	-99999.0	2	4.424979586745809
+HG00097	manta	chr1:888490:DEL:888577	1	1	0	0	0	0	1.9444826721501687	55.0	-99999.0	0	0.0	5.0	5.0	0.2631578947368421	0	93.0	116.0	5.0	0.49504950495049505	0.5247524752475248	1.119875913690787	50.0	1	1	1	1	0	1	0	0	0	0	1.0	2	4.20382130251655
+HG00097	manta	chr1:904478:DEL:904576	1	1	0	0	0	0	1.99563519459755	14.0	-99999.0	0	0.0	10.0	10.0	0.4166666666666667	0	31.0	294.0	3.0	0.7227722772277227	0.7722772277227723	1.6490755403817534	14.0	0	0	1	1	0	1	0	0	0	0	1.0	3	4.20382130251655
+HG00097	manta	chr1:998763:INS:None	1	0	0	1	0	0	1.7708520116421442	-99999.0	-99999.0	0	0.0	15.0	15.0	0.6521739130434783	0	43.0	538.0	-99999.0	0.7920792079207921	0.7920792079207921	1.6906469647232745	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	3	4.472888033769636
+HG00097	manta	chr1:1028471:DEL:1029081	1	1	0	0	0	0	2.786041210242554	19.0	-99999.0	0	13.0	12.0	25.0	0.352112676056338	0	477.0	654.0	36.0	0.6534653465346535	0.7227722772277227	1.7262819301695698	19.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.472888033769636
+HG00097	manta	chr1:1068824:INS:None	1	0	0	1	0	0	1.8976270912904414	10.0	-99999.0	0	0.0	29.0	29.0	0.9666666666666667	1	75.0	999.0	-99999.0	0.8514851485148515	0.8514851485148515	1.5447689726449716	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	4.60588658966101
+HG00097	manta	chr1:1427385:DEL:1427442	0	1	0	0	0	0	1.7634279935629373	-99999.0	-99999.0	0	0.0	2.0	2.0	0.16666666666666666	0	13.0	13.0	3.0	0.693069306930693	0.7029702970297029	1.3844773445130296	0.0	0	0	1	1	0	1	0	0	0	0	1.0	0	5.22457708545644
+HG00097	manta	chr1:1595101:DEL:1595183	1	1	0	0	0	0	1.919078092376074	2.0	-99999.0	0	0.0	10.0	10.0	0.19230769230769232	0	266.0	266.0	12.0	0.5148514851485149	0.5247524752475248	1.4455984547909617	2.0	0	0	1	1	1	1	0	0	0	0	1.0	1	4.856571815297643
+HG00097	manta	chr1:1666974:DEL:1667141	1	1	0	0	0	0	2.225309281725863	18.0	-99999.0	0	11.0	21.0	32.0	1.0	1	59.0	888.0	11.0	0.5247524752475248	0.5445544554455446	1.9680479893568856	18.0	1	1	1	1	1	0	1	0	0	0	1.0	3	4.763375575548453
+HG00097	manta	chr1:1724966:DEL:1726924	1	1	0	0	0	0	3.2920344359947364	27.0	27.0	0	8.0	10.0	18.0	0.2727272727272727	0	420.0	420.0	31.0	0.5643564356435643	0.6435643564356436	1.9231444533753872	27.0	1	1	1	1	1	1	1	0	0	0	1.0	2	4.391640703492388
+HG00097	manta	chr1:1749605:INS:None	1	0	0	1	0	0	1.7075701760979363	23.0	-99999.0	0	0.0	10.0	10.0	0.37037037037037035	0	262.0	358.0	-99999.0	0.4752475247524752	0.4752475247524752	1.6885390899981634	-99999.0	1	1	1	1	1	1	0	0	0	0	-99999.0	2	4.391640703492388
+HG00097	manta	chr1:1924223:INS:None	1	0	0	1	0	0	2.0170333392987803	10.0	-99999.0	0	0.0	16.0	16.0	1.0	1	41.0	630.0	-99999.0	0.7425742574257426	0.7425742574257426	1.6063691498853343	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	3.71281800020785
+HG00097	manta	chr1:1929384:INS:None	1	0	0	1	0	0	-99999.0	5.0	5.0	0	0.0	19.0	19.0	1.0	1	39.0	580.0	-99999.0	0.693069306930693	0.693069306930693	1.4041502486751618	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	3.71281800020785
+HG00097	manta	chr1:1934989:DEL:1935584	1	1	0	0	0	0	2.7752462597402365	10.0	-99999.0	0	16.0	18.0	34.0	1.0	1	86.0	934.0	16.0	0.44554455445544555	0.504950495049505	1.8451028377340413	10.0	0	0	1	1	0	1	0	0	0	0	1.0	5	3.7486530934242674
+HG00097	manta	chr1:1948934:INS:None	1	0	0	1	0	0	2.0492180226701815	2.0	-99999.0	0	3.0	22.0	25.0	1.0	1	64.0	864.0	3.0	0.5643564356435643	0.5643564356435643	1.0561905395876316	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	4.1444496608689
+HG00097	manta	chr1:1993704:INS:None	1	0	0	1	0	0	2.1702617153949575	15.0	-99999.0	0	8.0	45.0	53.0	1.0	1	128.0	999.0	8.0	0.4752475247524752	0.4752475247524752	1.9812115261970087	-99999.0	0	0	0	0	0	0	1	0	0	0	-99999.0	5	4.406829613621544
+HG00097	manta	chr1:2019220:INS:None	1	0	0	1	0	0	-99999.0	23.0	23.0	0	2.0	12.0	14.0	1.0	1	20.0	261.0	2.0	0.8613861386138614	0.8613861386138614	1.3524948796891727	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	4.406829613621544
+HG00097	manta	chr1:2421838:BND:None	1	0	0	0	0	1	-99999.0	3.0	-99999.0	0	4.0	25.0	29.0	0.37662337662337664	0	582.0	999.0	23.0	0.49504950495049505	0.49504950495049505	1.983371708749389	-99999.0	0	0	1	1	0	1	1	0	0	0	-99999.0	0	5.255942731372637
+HG00097	manta	chr1:2602115:DUP:2602189	1	0	1	0	0	0	1.8750612633917	-99999.0	-99999.0	0	0.0	11.0	11.0	0.12941176470588237	0	255.0	255.0	30.0	0.5643564356435643	0.6435643564356436	1.913521655146875	1.0	0	0	0	0	0	0	0	0	0	0	0.0	1	1.6989700043360187
+HG00097	manta	chr1:2602164:INS:None	1	0	0	1	0	0	1.919078092376074	-99999.0	-99999.0	0	0.0	22.0	22.0	0.6470588235294118	0	137.0	697.0	-99999.0	0.594059405940594	0.594059405940594	1.9680479893568856	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	1	1.6989700043360187
+HG00097	manta	chr1:3026038:INS:None	0	0	0	1	0	0	-99999.0	5.0	5.0	0	2.0	26.0	28.0	0.875	1	6.0	831.0	6.0	0.48514851485148514	0.48514851485148514	1.556688426030284	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	0	5.192483819902663
+HG00097	manta	chr1:3181807:DEL:3181925	1	1	0	0	0	0	2.075546961392531	14.0	-99999.0	0	0.0	27.0	27.0	1.0	1	65.0	999.0	-99999.0	0.44554455445544555	0.46534653465346537	1.9395446972377677	14.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.525860772931853
+HG00097	manta	chr1:3215369:INS:None	1	0	0	1	0	0	1.792391689498254	54.0	-99999.0	0	0.0	27.0	27.0	1.0	1	68.0	999.0	-99999.0	0.6831683168316832	0.6831683168316832	1.8503664672483442	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	4.525860772931853
+HG00097	manta	chr1:3260742:INS:None	1	0	0	1	0	0	-99999.0	5.0	5.0	0	8.0	29.0	37.0	1.0	1	95.0	999.0	8.0	0.4752475247524752	0.4752475247524752	1.6383659495376368	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	3	4.6568070667104475
+HG00097	manta	chr1:3316611:DEL:3316667	1	1	0	0	0	0	1.7558748556724915	23.0	-99999.0	0	0.0	10.0	10.0	0.37037037037037035	0	177.0	306.0	5.0	0.49504950495049505	0.49504950495049505	1.8200734972498984	23.0	0	0	1	1	0	1	0	0	0	0	1.0	1	4.747178671360165
+HG00097	manta	chr1:3832372:DEL:3832444	1	1	0	0	0	0	1.863322860120456	45.0	-99999.0	0	0.0	5.0	5.0	0.625	0	25.0	127.0	1.0	0.5247524752475248	0.5841584158415841	1.9137004533259432	45.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.560086048497414
+HG00097	manta	chr1:3868686:INS:None	1	0	0	1	0	0	2.4712917110589387	1.0	-99999.0	0	4.0	38.0	42.0	1.0	1	29.0	708.0	4.0	0.6336633663366337	0.6336633663366337	1.5250608688455582	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	4.560086048497414
+HG00097	manta	chr1:3929498:INS:None	1	0	0	1	0	0	2.4668676203541096	4.0	-99999.0	0	2.0	27.0	29.0	1.0	1	62.0	707.0	2.0	0.44554455445544555	0.44554455445544555	1.4880199108409586	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	3	4.7839964283643
+HG00097	manta	chr1:3999775:DEL:3999895	1	1	0	0	0	0	2.08278537031645	-99999.0	-99999.0	0	0.0	25.0	25.0	1.0	1	73.0	999.0	-99999.0	0.6633663366336634	0.693069306930693	1.7648687041978608	1.0	0	0	1	1	1	1	0	0	0	0	1.0	1	4.846819393669543
+HG00097	manta	chr1:4129813:DEL:4129873	1	1	0	0	0	0	1.7853298350107671	45.0	-99999.0	0	0.0	9.0	9.0	0.34615384615384615	0	162.0	312.0	6.0	0.6039603960396039	0.6831683168316832	1.8738460050301151	45.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.16660770308391
+HG00097	manta	chr1:4144488:DUP:4144620	0	0	1	0	0	0	2.123851640967086	4.0	4.0	0	0.0	22.0	22.0	0.43137254901960786	0	332.0	518.0	11.0	0.42574257425742573	0.48514851485148514	1.4784771363149791	4.0	0	0	1	1	0	1	0	0	0	0	1.0	2	2.1492191126553797
+HG00097	manta	chr1:4144628:INS:None	0	0	0	1	0	0	1.9084850188786497	2.0	-99999.0	0	0.0	28.0	28.0	1.0	1	71.0	982.0	-99999.0	0.48514851485148514	0.48514851485148514	1.4687802326926382	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	2.1492191126553797
+HG00097	manta	chr1:4333580:BND:None	1	0	0	0	0	1	-99999.0	-99999.0	-99999.0	0	12.0	0.0	12.0	0.5	0	142.0	142.0	24.0	0.5643564356435643	0.5643564356435643	1.9623130244403804	-99999.0	0	0	1	1	0	1	0	0	0	1	-99999.0	1	3.465680211598278
+HG00097	manta	chr1:4336501:INS:None	1	0	0	1	0	0	-99999.0	27.0	27.0	0	3.0	37.0	40.0	1.0	1	68.0	929.0	3.0	0.40594059405940597	0.40594059405940597	1.747964514410695	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	3.465680211598278
+HG00097	manta	chr1:4939764:BND:None	0	0	0	0	0	1	-99999.0	599.0	-99999.0	1	14.0	0.0	14.0	0.4117647058823529	0	115.0	115.0	34.0	0.45544554455445546	0.45544554455445546	1.9728402340561	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	0	5.133363239048624
+HG00097	manta	chr1:5075708:INS:None	1	0	0	1	0	0	1.863322860120456	13.0	-99999.0	0	0.0	12.0	12.0	1.0	1	26.0	288.0	-99999.0	0.07920792079207921	0.07920792079207921	1.3986907959255488	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	4.927334427020689
+HG00097	manta	chr1:5160300:INS:None	0	0	0	1	0	0	2.765668554759014	-99999.0	-99999.0	0	0.0	46.0	46.0	0.92	1	52.0	999.0	4.0	0.37623762376237624	0.37623762376237624	1.7050759800383022	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	4.927334427020689
+HG00097	manta	chr1:5387025:BND:None	1	0	0	0	0	1	-99999.0	466.0	-99999.0	1	15.0	0.0	15.0	0.6	0	116.0	228.0	25.0	0.5148514851485149	0.5148514851485149	1.3511081047001698	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	2.161368002234975
+HG00097	manta	chr1:5387169:DUP:5387401	1	0	1	0	0	0	2.367355921026019	-99999.0	-99999.0	0	1.0	23.0	24.0	0.42857142857142855	0	338.0	481.0	13.0	0.45544554455445546	0.4752475247524752	1.5481463016741694	0.0	0	0	1	1	0	1	0	0	0	0	1.0	2	2.161368002234975
+HG00097	manta	chr1:5414152:DEL:5414227	1	1	0	0	0	0	1.8808135922807914	18.0	-99999.0	0	0.0	7.0	7.0	0.3333333333333333	0	102.0	157.0	6.0	0.5346534653465347	0.5346534653465347	1.9698531038077964	18.0	0	0	1	1	0	1	0	0	0	0	1.0	3	4.431106328181145
+HG00097	manta	chr1:5499440:DEL:5499546	1	1	0	0	0	0	2.0293837776852097	8.0	-99999.0	0	0.0	9.0	9.0	0.21951219512195122	0	276.0	276.0	7.0	0.26732673267326734	0.26732673267326734	1.6246863594780123	8.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.930893022406026
+HG00097	manta	chr1:5593288:INS:None	1	0	0	1	0	0	1.806179973983887	2.0	-99999.0	0	0.0	22.0	22.0	0.6470588235294118	0	142.0	889.0	-99999.0	0.5544554455445545	0.5544554455445545	1.9892256298878759	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	2	4.9076585372801205
+HG00097	manta	chr1:5674133:BND:None	0	0	0	0	0	1	-99999.0	7.0	-99999.0	0	11.0	47.0	58.0	0.3020833333333333	0	999.0	999.0	78.0	0.38613861386138615	0.38613861386138615	1.9050918955475322	-99999.0	0	0	0	0	0	0	0	1	0	0	-99999.0	1	4.9076585372801205
+HG00097	manta	chr1:5874228:INS:None	1	0	0	1	0	0	2.4533183400470375	13.0	-99999.0	0	18.0	23.0	41.0	0.5394736842105263	0	327.0	571.0	37.0	0.44554455445544555	0.44554455445544555	1.9714887292053216	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	0	5.116717299877443
+HG00097	manta	chr1:6005060:DEL:6005340	1	1	0	0	0	0	2.44870631990508	-99999.0	-99999.0	0	14.0	15.0	29.0	0.9666666666666667	1	23.0	314.0	15.0	0.6237623762376238	0.6831683168316832	1.5376840111363732	2.0	0	0	1	1	0	1	0	0	0	0	1.0	1	4.317666442356501
+HG00097	manta	chr1:6025840:DEL:6025908	1	1	0	0	0	0	1.8388490907372552	74.0	-99999.0	0	0.0	6.0	6.0	0.2222222222222222	0	182.0	184.0	9.0	0.6435643564356436	0.7029702970297029	1.7307956407165326	50.0	0	0	1	1	1	1	0	0	0	0	1.0	1	4.317666442356501
+HG00097	manta	chr1:6742557:INS:None	1	0	0	1	0	0	-99999.0	4.0	4.0	0	5.0	31.0	36.0	1.0	1	101.0	999.0	5.0	0.07920792079207921	0.07920792079207921	1.320097363865938	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	0	5.297445341827969
+HG00097	manta	chr1:6940912:INS:None	1	0	0	1	0	0	-99999.0	4.0	4.0	0	0.0	34.0	34.0	0.9714285714285714	1	95.0	999.0	-99999.0	0.6336633663366337	0.6336633663366337	1.902230293166466	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	1	4.635272580112365
+HG00097	manta	chr1:6984090:INS:None	1	0	0	1	0	0	2.311753861055754	2.0	-99999.0	0	6.0	35.0	41.0	1.0	1	72.0	999.0	6.0	0.48514851485148514	0.48514851485148514	1.5136654464703396	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	4.635272580112365
+HG00097	manta	chr1:7510011:DEL:7511458	1	1	0	0	0	0	3.1607685618611283	-99999.0	-99999.0	0	11.0	15.0	26.0	0.36619718309859156	0	527.0	711.0	35.0	0.46534653465346537	0.48514851485148514	1.8175669943687356	0.0	0	0	1	0	0	1	0	0	0	0	0.015193370165745856	1	4.5124575861973435

example/sv_scored.tsv

@@ -1,8 +1,51 @@
-chrom	pos	qual	gq	pr_ref	pr_alt	sr_ref	sr_alt	vf_ref	vf_alt	total_alt_support	sr_pr_ratio	vaf_estimate	strand_bias_fs	cipos_width	ciend_width	homlen	svlen_abs	svtype_DEL	svtype_INS	svtype_DUP	svtype_BND	is_imprecise	filter_pass	n_filter_flags	CS	tier
-chr10	42254638	182	6	2.0	0.0	3.0	5.0	5.0	5.0	5.0	5.0	0.5	9.7			0.0	50.0	1	0	0	0	0	1	0	0.37	LOW
-chr12	12391920	999	281	11.0	20.0	3.0	14.0	12.0	30.0	34.0	0.6666666666666666	0.7083333333333334	7.186			0.0		0	0	0	1	0	1	0	0.1762	LOW
-chr16	8562503	693	693	54.0	22.0	36.0	9.0	69.0	26.0	31.0	0.391304347826087	0.256198347107438	0.0			0.0	1781.0	0	0	1	0	0	1	0	0.075	LOW
-chr17	23232202	310	38	0.0	0.0	0.0	21.0	0.0	21.0	21.0	21.0	1.0	0.0	5.0	5.0	5.0		0	1	0	0	0	0	1	0.3	WARNING
-chr3	163746379	999	118	0.0	10.0	0.0	33.0	0.0	43.0	43.0	3.0	1.0	0.0	8.0		8.0	306.0	1	0	0	0	0	1	0	1.0	HIGH
-chr6	32648154	82	82	7.0	5.0			7.0	5.0	5.0		0.4166666666666667		415.0	238.0	0.0	4156.0	1	0	0	0	1	1	0	0.495	LOW
-chr8	91664843	999	148	0.0	28.0	0.0	36.0	0.0	56.0	64.0	1.2413793103448276	1.0	0.0	15.0		15.0	316.0	0	1	0	0	0	1	0	1.0	HIGH
+sample	caller	variant_ID	is_pass	svtype_DEL	svtype_DUP	svtype_INS	svtype_INV	svtype_BND	svlen_log	cipos_width	ciend_width	is_imprecise	pe_support	sr_support	total_support	vaf	gt_hom	gq	qual_norm	local_depth	gc_min	gc_max	entropy_min	microhom_max	in_segdup_either	in_segdup_both	in_difficult_either	in_difficult_both	in_lowmap_either	in_tandem_either	in_Alu_either	in_L1_either	in_SVA_either	in_LTR_either	frac_span_repeat	n_neighbors	nn_log_dist	CS_raw	CS	tier
+HG00097	manta	chr1:861885:BND:None	1	0	0	0	0	1	-99999.0	31.0	-99999.0	0	5.0	7.0	12.0	0.1538461538461538	0	182.0	182.0	43.0	0.297029702970297	0.297029702970297	1.6350564503142977	-99999.0	1	1	1	1	0	1	0	0	0	0	-99999.0	2	4.424979586745809	0.0077	0.0016	LOW
+HG00097	manta	chr1:888490:DEL:888577	1	1	0	0	0	0	1.944482672150169	55.0	-99999.0	0	0.0	5.0	5.0	0.2631578947368421	0	93.0	116.0	5.0	0.495049504950495	0.5247524752475248	1.119875913690787	50.0	1	1	1	1	0	1	0	0	0	0	1.0	2	4.20382130251655	0.7047	0.8104	HIGH
+HG00097	manta	chr1:904478:DEL:904576	1	1	0	0	0	0	1.99563519459755	14.0	-99999.0	0	0.0	10.0	10.0	0.4166666666666667	0	31.0	294.0	3.0	0.7227722772277227	0.7722772277227723	1.6490755403817534	14.0	0	0	1	1	0	1	0	0	0	0	1.0	3	4.20382130251655	0.616	0.6822	MODERATE
+HG00097	manta	chr1:998763:INS:None	1	0	0	1	0	0	1.7708520116421442	-99999.0	-99999.0	0	0.0	15.0	15.0	0.6521739130434783	0	43.0	538.0	-99999.0	0.7920792079207921	0.7920792079207921	1.6906469647232745	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	3	4.472888033769636	0.6379	0.7226	HIGH
+HG00097	manta	chr1:1028471:DEL:1029081	1	1	0	0	0	0	2.786041210242554	19.0	-99999.0	0	13.0	12.0	25.0	0.352112676056338	0	477.0	654.0	36.0	0.6534653465346535	0.7227722772277227	1.7262819301695698	19.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.472888033769636	0.809	0.8988	HIGH
+HG00097	manta	chr1:1068824:INS:None	1	0	0	1	0	0	1.8976270912904412	10.0	-99999.0	0	0.0	29.0	29.0	0.9666666666666668	1	75.0	999.0	-99999.0	0.8514851485148515	0.8514851485148515	1.5447689726449716	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	4.60588658966101	0.7489	0.8545	HIGH
+HG00097	manta	chr1:1427385:DEL:1427442	0	1	0	0	0	0	1.7634279935629371	-99999.0	-99999.0	0	0.0	2.0	2.0	0.1666666666666666	0	13.0	13.0	3.0	0.693069306930693	0.7029702970297029	1.3844773445130296	0.0	0	0	1	1	0	1	0	0	0	0	1.0	0	5.22457708545644	0.1506	0.1012	LOW
+HG00097	manta	chr1:1595101:DEL:1595183	1	1	0	0	0	0	1.919078092376074	2.0	-99999.0	0	0.0	10.0	10.0	0.1923076923076923	0	266.0	266.0	12.0	0.5148514851485149	0.5247524752475248	1.4455984547909615	2.0	0	0	1	1	1	1	0	0	0	0	1.0	1	4.856571815297643	0.4295	0.4349	WARNING
+HG00097	manta	chr1:1666974:DEL:1667141	1	1	0	0	0	0	2.225309281725863	18.0	-99999.0	0	11.0	21.0	32.0	1.0	1	59.0	888.0	11.0	0.5247524752475248	0.5445544554455446	1.9680479893568856	18.0	1	1	1	1	1	0	1	0	0	0	1.0	3	4.763375575548453	0.7289	0.8283	HIGH
+HG00097	manta	chr1:1724966:DEL:1726924	1	1	0	0	0	0	3.2920344359947364	27.0	27.0	0	8.0	10.0	18.0	0.2727272727272727	0	420.0	420.0	31.0	0.5643564356435643	0.6435643564356436	1.9231444533753872	27.0	1	1	1	1	1	1	1	0	0	0	1.0	2	4.391640703492388	0.8005	0.8964	HIGH
+HG00097	manta	chr1:1749605:INS:None	1	0	0	1	0	0	1.7075701760979365	23.0	-99999.0	0	0.0	10.0	10.0	0.3703703703703703	0	262.0	358.0	-99999.0	0.4752475247524752	0.4752475247524752	1.6885390899981634	-99999.0	1	1	1	1	1	1	0	0	0	0	-99999.0	2	4.391640703492388	0.7091	0.8117	HIGH
+HG00097	manta	chr1:1924223:INS:None	1	0	0	1	0	0	2.0170333392987803	10.0	-99999.0	0	0.0	16.0	16.0	1.0	1	41.0	630.0	-99999.0	0.7425742574257426	0.7425742574257426	1.6063691498853343	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	3.71281800020785	0.526	0.5557	MODERATE
+HG00097	manta	chr1:1929384:INS:None	1	0	0	1	0	0	-99999.0	5.0	5.0	0	0.0	19.0	19.0	1.0	1	39.0	580.0	-99999.0	0.693069306930693	0.693069306930693	1.4041502486751618	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	3.71281800020785	0.2185	0.1761	LOW
+HG00097	manta	chr1:1934989:DEL:1935584	1	1	0	0	0	0	2.7752462597402365	10.0	-99999.0	0	16.0	18.0	34.0	1.0	1	86.0	934.0	16.0	0.4455445544554455	0.504950495049505	1.8451028377340413	10.0	0	0	1	1	0	1	0	0	0	0	1.0	5	3.748653093424267	0.686	0.7801	HIGH
+HG00097	manta	chr1:1948934:INS:None	1	0	0	1	0	0	2.049218022670181	2.0	-99999.0	0	3.0	22.0	25.0	1.0	1	64.0	864.0	3.0	0.5643564356435643	0.5643564356435643	1.0561905395876316	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	4.1444496608689	0.7533	0.8545	HIGH
+HG00097	manta	chr1:1993704:INS:None	1	0	0	1	0	0	2.170261715394957	15.0	-99999.0	0	8.0	45.0	53.0	1.0	1	128.0	999.0	8.0	0.4752475247524752	0.4752475247524752	1.981211526197009	-99999.0	0	0	0	0	0	0	1	0	0	0	-99999.0	5	4.406829613621544	0.8105	0.8988	HIGH
+HG00097	manta	chr1:2019220:INS:None	1	0	0	1	0	0	-99999.0	23.0	23.0	0	2.0	12.0	14.0	1.0	1	20.0	261.0	2.0	0.8613861386138614	0.8613861386138614	1.3524948796891727	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	5	4.406829613621544	0.5454	0.5772	MODERATE
+HG00097	manta	chr1:2421838:BND:None	1	0	0	0	0	1	-99999.0	3.0	-99999.0	0	4.0	25.0	29.0	0.3766233766233766	0	582.0	999.0	23.0	0.495049504950495	0.495049504950495	1.983371708749389	-99999.0	0	0	1	1	0	1	1	0	0	0	-99999.0	0	5.255942731372637	0.002	0.0003	LOW
+HG00097	manta	chr1:2602115:DUP:2602189	1	0	1	0	0	0	1.8750612633917	-99999.0	-99999.0	0	0.0	11.0	11.0	0.1294117647058823	0	255.0	255.0	30.0	0.5643564356435643	0.6435643564356436	1.913521655146875	1.0	0	0	0	0	0	0	0	0	0	0	0.0	1	1.6989700043360187	0.0	0.0	LOW
+HG00097	manta	chr1:2602164:INS:None	1	0	0	1	0	0	1.919078092376074	-99999.0	-99999.0	0	0.0	22.0	22.0	0.6470588235294118	0	137.0	697.0	-99999.0	0.594059405940594	0.594059405940594	1.9680479893568856	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	1	1.6989700043360187	0.7853	0.8865	HIGH
+HG00097	manta	chr1:3026038:INS:None	0	0	0	1	0	0	-99999.0	5.0	5.0	0	2.0	26.0	28.0	0.875	1	6.0	831.0	6.0	0.4851485148514851	0.4851485148514851	1.556688426030284	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	0	5.192483819902663	0.8821	0.935	HIGH
+HG00097	manta	chr1:3181807:DEL:3181925	1	1	0	0	0	0	2.075546961392531	14.0	-99999.0	0	0.0	27.0	27.0	1.0	1	65.0	999.0	-99999.0	0.4455445544554455	0.4653465346534653	1.939544697237768	14.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.525860772931853	0.6932	0.798	HIGH
+HG00097	manta	chr1:3215369:INS:None	1	0	0	1	0	0	1.792391689498254	54.0	-99999.0	0	0.0	27.0	27.0	1.0	1	68.0	999.0	-99999.0	0.6831683168316832	0.6831683168316832	1.850366467248344	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	4.525860772931853	0.8417	0.9154	HIGH
+HG00097	manta	chr1:3260742:INS:None	1	0	0	1	0	0	-99999.0	5.0	5.0	0	8.0	29.0	37.0	1.0	1	95.0	999.0	8.0	0.4752475247524752	0.4752475247524752	1.6383659495376368	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	3	4.656807066710448	0.8274	0.9091	HIGH
+HG00097	manta	chr1:3316611:DEL:3316667	1	1	0	0	0	0	1.7558748556724917	23.0	-99999.0	0	0.0	10.0	10.0	0.3703703703703703	0	177.0	306.0	5.0	0.495049504950495	0.495049504950495	1.8200734972498984	23.0	0	0	1	1	0	1	0	0	0	0	1.0	1	4.747178671360165	0.8483	0.917	HIGH
+HG00097	manta	chr1:3832372:DEL:3832444	1	1	0	0	0	0	1.863322860120456	45.0	-99999.0	0	0.0	5.0	5.0	0.625	0	25.0	127.0	1.0	0.5247524752475248	0.5841584158415841	1.9137004533259432	45.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.560086048497414	0.6316	0.7133	HIGH
+HG00097	manta	chr1:3868686:INS:None	1	0	0	1	0	0	2.4712917110589387	1.0	-99999.0	0	4.0	38.0	42.0	1.0	1	29.0	708.0	4.0	0.6336633663366337	0.6336633663366337	1.5250608688455582	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	4.560086048497414	0.6765	0.7715	HIGH
+HG00097	manta	chr1:3929498:INS:None	1	0	0	1	0	0	2.4668676203541096	4.0	-99999.0	0	2.0	27.0	29.0	1.0	1	62.0	707.0	2.0	0.4455445544554455	0.4455445544554455	1.4880199108409586	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	3	4.7839964283643	0.4086	0.3836	WARNING
+HG00097	manta	chr1:3999775:DEL:3999895	1	1	0	0	0	0	2.08278537031645	-99999.0	-99999.0	0	0.0	25.0	25.0	1.0	1	73.0	999.0	-99999.0	0.6633663366336634	0.693069306930693	1.7648687041978608	1.0	0	0	1	1	1	1	0	0	0	0	1.0	1	4.846819393669543	0.7755	0.874	HIGH
+HG00097	manta	chr1:4129813:DEL:4129873	1	1	0	0	0	0	1.7853298350107671	45.0	-99999.0	0	0.0	9.0	9.0	0.3461538461538461	0	162.0	312.0	6.0	0.6039603960396039	0.6831683168316832	1.8738460050301151	45.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.16660770308391	0.8085	0.8988	HIGH
+HG00097	manta	chr1:4144488:DUP:4144620	0	0	1	0	0	0	2.123851640967086	4.0	4.0	0	0.0	22.0	22.0	0.4313725490196078	0	332.0	518.0	11.0	0.4257425742574257	0.4851485148514851	1.4784771363149791	4.0	0	0	1	1	0	1	0	0	0	0	1.0	2	2.14921911265538	0.0007	0.0	LOW
+HG00097	manta	chr1:4144628:INS:None	0	0	0	1	0	0	1.9084850188786495	2.0	-99999.0	0	0.0	28.0	28.0	1.0	1	71.0	982.0	-99999.0	0.4851485148514851	0.4851485148514851	1.4687802326926382	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	2.14921911265538	0.1717	0.1265	LOW
+HG00097	manta	chr1:4333580:BND:None	1	0	0	0	0	1	-99999.0	-99999.0	-99999.0	0	12.0	0.0	12.0	0.5	0	142.0	142.0	24.0	0.5643564356435643	0.5643564356435643	1.9623130244403804	-99999.0	0	0	1	1	0	1	0	0	0	1	-99999.0	1	3.465680211598278	0.0055	0.0012	LOW
+HG00097	manta	chr1:4336501:INS:None	1	0	0	1	0	0	-99999.0	27.0	27.0	0	3.0	37.0	40.0	1.0	1	68.0	929.0	3.0	0.4059405940594059	0.4059405940594059	1.747964514410695	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	3.465680211598278	0.7674	0.8684	HIGH
+HG00097	manta	chr1:4939764:BND:None	0	0	0	0	0	1	-99999.0	599.0	-99999.0	1	14.0	0.0	14.0	0.4117647058823529	0	115.0	115.0	34.0	0.4554455445544554	0.4554455445544554	1.9728402340561	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	0	5.133363239048624	0.0303	0.0134	LOW
+HG00097	manta	chr1:5075708:INS:None	1	0	0	1	0	0	1.863322860120456	13.0	-99999.0	0	0.0	12.0	12.0	1.0	1	26.0	288.0	-99999.0	0.0792079207920792	0.0792079207920792	1.3986907959255488	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	4.927334427020689	0.749	0.8545	HIGH
+HG00097	manta	chr1:5160300:INS:None	0	0	0	1	0	0	2.765668554759014	-99999.0	-99999.0	0	0.0	46.0	46.0	0.92	1	52.0	999.0	4.0	0.3762376237623762	0.3762376237623762	1.7050759800383022	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	4.927334427020689	0.7965	0.8914	HIGH
+HG00097	manta	chr1:5387025:BND:None	1	0	0	0	0	1	-99999.0	466.0	-99999.0	1	15.0	0.0	15.0	0.6	0	116.0	228.0	25.0	0.5148514851485149	0.5148514851485149	1.3511081047001698	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	2	2.161368002234975	0.0015	0.0	LOW
+HG00097	manta	chr1:5387169:DUP:5387401	1	0	1	0	0	0	2.367355921026019	-99999.0	-99999.0	0	1.0	23.0	24.0	0.4285714285714285	0	338.0	481.0	13.0	0.4554455445544554	0.4752475247524752	1.5481463016741694	0.0	0	0	1	1	0	1	0	0	0	0	1.0	2	2.161368002234975	0.0005	0.0	LOW
+HG00097	manta	chr1:5414152:DEL:5414227	1	1	0	0	0	0	1.8808135922807916	18.0	-99999.0	0	0.0	7.0	7.0	0.3333333333333333	0	102.0	157.0	6.0	0.5346534653465347	0.5346534653465347	1.9698531038077964	18.0	0	0	1	1	0	1	0	0	0	0	1.0	3	4.431106328181145	0.6073	0.67	MODERATE
+HG00097	manta	chr1:5499440:DEL:5499546	1	1	0	0	0	0	2.0293837776852097	8.0	-99999.0	0	0.0	9.0	9.0	0.2195121951219512	0	276.0	276.0	7.0	0.2673267326732673	0.2673267326732673	1.6246863594780123	8.0	0	0	1	1	0	1	0	0	0	0	1.0	2	4.930893022406026	0.5474	0.5784	MODERATE
+HG00097	manta	chr1:5593288:INS:None	1	0	0	1	0	0	1.806179973983887	2.0	-99999.0	0	0.0	22.0	22.0	0.6470588235294118	0	142.0	889.0	-99999.0	0.5544554455445545	0.5544554455445545	1.989225629887876	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	2	4.9076585372801205	0.9166	0.9509	HIGH
+HG00097	manta	chr1:5674133:BND:None	0	0	0	0	0	1	-99999.0	7.0	-99999.0	0	11.0	47.0	58.0	0.3020833333333333	0	999.0	999.0	78.0	0.3861386138613861	0.3861386138613861	1.905091895547532	-99999.0	0	0	0	0	0	0	0	1	0	0	-99999.0	1	4.9076585372801205	0.0368	0.0188	LOW
+HG00097	manta	chr1:5874228:INS:None	1	0	0	1	0	0	2.453318340047037	13.0	-99999.0	0	18.0	23.0	41.0	0.5394736842105263	0	327.0	571.0	37.0	0.4455445544554455	0.4455445544554455	1.9714887292053216	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	0	5.116717299877443	0.9601	0.9697	HIGH
+HG00097	manta	chr1:6005060:DEL:6005340	1	1	0	0	0	0	2.44870631990508	-99999.0	-99999.0	0	14.0	15.0	29.0	0.9666666666666668	1	23.0	314.0	15.0	0.6237623762376238	0.6831683168316832	1.5376840111363732	2.0	0	0	1	1	0	1	0	0	0	0	1.0	1	4.317666442356501	0.7321	0.8286	HIGH
+HG00097	manta	chr1:6025840:DEL:6025908	1	1	0	0	0	0	1.8388490907372552	74.0	-99999.0	0	0.0	6.0	6.0	0.2222222222222222	0	182.0	184.0	9.0	0.6435643564356436	0.7029702970297029	1.7307956407165326	50.0	0	0	1	1	1	1	0	0	0	0	1.0	1	4.317666442356501	0.876	0.9325	HIGH
+HG00097	manta	chr1:6742557:INS:None	1	0	0	1	0	0	-99999.0	4.0	4.0	0	5.0	31.0	36.0	1.0	1	101.0	999.0	5.0	0.0792079207920792	0.0792079207920792	1.320097363865938	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	0	5.297445341827969	0.6745	0.7695	HIGH
+HG00097	manta	chr1:6940912:INS:None	1	0	0	1	0	0	-99999.0	4.0	4.0	0	0.0	34.0	34.0	0.9714285714285714	1	95.0	999.0	-99999.0	0.6336633663366337	0.6336633663366337	1.902230293166466	-99999.0	0	0	0	0	0	0	0	0	0	0	-99999.0	1	4.635272580112365	0.7578	0.8633	HIGH
+HG00097	manta	chr1:6984090:INS:None	1	0	0	1	0	0	2.311753861055754	2.0	-99999.0	0	6.0	35.0	41.0	1.0	1	72.0	999.0	6.0	0.4851485148514851	0.4851485148514851	1.5136654464703396	-99999.0	0	0	1	1	0	1	0	0	0	0	-99999.0	1	4.635272580112365	0.8751	0.9325	HIGH
+HG00097	manta	chr1:7510011:DEL:7511458	1	1	0	0	0	0	3.1607685618611283	-99999.0	-99999.0	0	11.0	15.0	26.0	0.3661971830985915	0	527.0	711.0	35.0	0.4653465346534653	0.4851485148514851	1.817566994368736	0.0	0	0	1	0	0	1	0	0	0	0	0.0151933701657458	1	4.512457586197344	0.7788	0.8742	HIGH

feature_builder.py

@@ -0,0 +1,708 @@
+#!/usr/bin/env python3
+"""
+SVSTR_Score feature builder (VCF + reference only, single sample).
+
+Computes the RandomForest input features defined in:
+    sv_features.tsv   (callers: manta, delly, lumpy)
+    str_features.tsv  (callers: expansionhunter, gangstr)
+
+Design constraints (head model):
+- Inputs are ONLY a short-read VCF + reference FASTA + static annotation BEDs.
+  No BAM, no cohort, no long-read (long-read is used elsewhere for labeling only).
+- Features are caller-common *concepts*; each caller is parsed by its own parser.
+- `caller` is recorded for bookkeeping but is NOT emitted as a model feature.
+
+Annotation BEDs must be sorted, bgzipped and tabix-indexed (see
+scripts/prepare_annotations or the resources/ prep step).
+
+ExpansionHunter input is its flat (optionally gzipped) TSV, not a VCF — pass it to --vcf.
+
+VALIDATION: validated on HG00097 (Manta/Delly/GangSTR VCFs + ExpansionHunter TSV).
+Four parsing bugs were found & fixed against real data:
+  1. GangSTR REPCN/REPCI come back from pysam as tuples (Number=2), not strings.
+  2. pysam returns absent Flags as False (not KeyError) -> is_imprecise used `in rec.info`.
+  3. INFO/END is consumed into rec.stop; rec.info['END'] is empty.
+  4. missing sentinel must be out-of-range (-99999); -1 collided with real negative
+     expansion_over_ref (contractions). LUMPY (smoove/SVTyper) not yet run.
+
+Usage:
+    python feature_builder.py \
+        --vcf sample.manta.vcf.gz --caller manta \
+        --fasta GRCh38.fa \
+        --giab-dir ../resources/giab_prepared \
+        --repeatmasker ../resources/repeatmasker/rmsk_class.bed.gz \
+        -o sample.manta.features.tsv
+"""
+
+import os
+import sys
+import math
+import bisect
+import argparse
+from collections import defaultdict
+
+import numpy as np
+import pandas as pd
+import pysam
+
+MISSING = -99999.0  # out-of-range sentinel for missing fields (paired with *_missing indicators).
+# Must be outside every feature's real range: expansion_over_ref can legitimately be negative,
+# so a small sentinel like -1 would collide with real contractions.
+SV_CALLERS = {"manta", "delly", "lumpy"}
+STR_CALLERS = {"expansionhunter", "gangstr"}
+PRIMARY_CONTIGS = ({f"chr{c}" for c in list(range(1, 23)) + ["X", "Y", "M"]}
+                   | {str(c) for c in list(range(1, 23)) + ["X", "Y", "MT", "M"]})
+
+# Features that can legitimately be MISSING. Their `<feat>_missing` indicator is
+# emitted ALWAYS (even if all-zero for a given caller) so every caller's output
+# has an identical, fixed column schema — one trained model consumes any caller's
+# converted VCF directly, no per-caller alignment needed.
+SV_MISSING_INDICATORS = [
+    "svlen_log", "cipos_width", "ciend_width", "vaf", "qual_norm", "gq",
+    "local_depth", "gt_hom", "gc_min", "gc_max", "entropy_min", "microhom_max",
+    "frac_span_repeat", "nn_log_dist",
+]
+STR_MISSING_INDICATORS = [
+    "motif_len", "ref_copynum", "locus_depth", "gt_hom", "gt_repcn_max", "gt_repcn_min",
+    "expansion_over_ref", "repci_width_max", "spanning_frac", "ref_tract_bp",
+    "allele_vs_readlen", "motif_is_homopolymer", "gc_flank", "entropy_flank",
+]
+
+
+# ---------------------------------------------------------------------------
+# Reference-sequence features (reused from A2Denovo conventions)
+# ---------------------------------------------------------------------------
+def gc_content(seq):
+    if not seq:
+        return MISSING
+    seq = seq.upper()
+    n = sum(1 for b in seq if b in "ACGT")
+    if n == 0:
+        return MISSING
+    return sum(1 for b in seq if b in "GC") / n
+
+
+def shannon_entropy(seq):
+    if not seq:
+        return MISSING
+    seq = seq.upper()
+    counts = defaultdict(int)
+    for b in seq:
+        if b in "ACGT":
+            counts[b] += 1
+    total = sum(counts.values())
+    if total == 0:
+        return MISSING
+    h = 0.0
+    for c in counts.values():
+        p = c / total
+        h -= p * math.log2(p)
+    return h
+
+
+def fetch(fasta, chrom, start, end):
+    """0-based half-open fetch with clamping; returns '' on failure."""
+    try:
+        start = max(0, start)
+        return fasta.fetch(chrom, start, end)
+    except Exception:
+        return ""
+
+
+def gc_entropy_at(fasta, chrom, pos1, win):
+    """GC and entropy in pos +/- win (pos is 1-based)."""
+    seq = fetch(fasta, chrom, pos1 - 1 - win, pos1 + win)
+    return gc_content(seq), shannon_entropy(seq)
+
+
+def microhomology(fasta, chrom, pos1, end1, max_k=50):
+    """
+    Approximate microhomology between the two breakpoints of an intra-chromosomal
+    SV: longest k (<=max_k) where the sequence adjacent to bp1 matches bp2.
+    Returns MISSING for inter-chromosomal / undefined cases.
+    """
+    if end1 is None or end1 <= pos1:
+        return MISSING
+    left = fetch(fasta, chrom, pos1 - max_k, pos1 + max_k).upper()
+    right = fetch(fasta, chrom, end1 - max_k, end1 + max_k).upper()
+    if len(left) < 2 * max_k or len(right) < 2 * max_k:
+        return MISSING
+    k = 0
+    while k < max_k and left[max_k + k] == right[max_k + k]:  # rightward match
+        k += 1
+    j = 0
+    while j < max_k and left[max_k - 1 - j] == right[max_k - 1 - j]:  # leftward
+        j += 1
+    return float(max(k, j))
+
+
+# ---------------------------------------------------------------------------
+# Tabix annotation (binary overlap + RepeatMasker element class)
+# ---------------------------------------------------------------------------
+class Annotator:
+    """Binary overlap against tabixed BEDs, with chr-naming fallback."""
+
+    RMSK_ELEMENTS = {  # label prefix in rmsk_class.bed (repClass/repFamily) -> flag
+        "SINE/Alu": "Alu",
+        "LINE/L1": "L1",
+        "Retroposon/SVA": "SVA",
+        "LTR": "LTR",
+    }
+
+    def __init__(self, giab_dir=None, repeatmasker=None):
+        self.tbx = {}
+        if giab_dir:
+            for name in ("segdups", "lowmap", "tandem", "difficult"):
+                p = os.path.join(giab_dir, f"{name}.bed.gz")
+                if os.path.exists(p):
+                    self.tbx[name] = pysam.TabixFile(p)
+                else:
+                    sys.stderr.write(f"[warn] missing GIAB bed: {p}\n")
+        self.rmsk = pysam.TabixFile(repeatmasker) if repeatmasker and os.path.exists(repeatmasker) else None
+
+    def _contigs(self, tbx, chrom):
+        if chrom in tbx.contigs:
+            return chrom
+        alt = chrom[3:] if chrom.startswith("chr") else "chr" + chrom
+        return alt if alt in tbx.contigs else None
+
+    def overlaps(self, name, chrom, pos1):
+        """1 if 1-based pos overlaps any interval in bed `name`, else 0."""
+        tbx = self.tbx.get(name)
+        if tbx is None:
+            return MISSING
+        c = self._contigs(tbx, chrom)
+        if c is None:
+            return 0
+        try:
+            for _ in tbx.fetch(c, pos1 - 1, pos1):
+                return 1
+        except Exception:
+            return 0
+        return 0
+
+    def frac_overlap(self, name, chrom, start1, end1):
+        """Fraction of [start1,end1] (1-based inclusive) covered by bed `name`."""
+        tbx = self.tbx.get(name)
+        if tbx is None or end1 is None or end1 < start1:
+            return MISSING
+        c = self._contigs(tbx, chrom)
+        if c is None:
+            return 0.0
+        span = end1 - start1 + 1
+        covered = 0
+        try:
+            for row in tbx.fetch(c, start1 - 1, end1):
+                f = row.split("\t")
+                s, e = int(f[1]), int(f[2])
+                covered += max(0, min(end1, e) - max(start1 - 1, s))
+        except Exception:
+            return 0.0
+        return min(1.0, covered / span) if span > 0 else 0.0
+
+    def rmsk_elements(self, chrom, pos1):
+        """Return dict {Alu,L1,SVA,LTR -> 0/1} for the position."""
+        flags = {"Alu": 0, "L1": 0, "SVA": 0, "LTR": 0}
+        if self.rmsk is None:
+            return {k: MISSING for k in flags}
+        c = self._contigs(self.rmsk, chrom)
+        if c is None:
+            return flags
+        try:
+            for row in self.rmsk.fetch(c, pos1 - 1, pos1):
+                label = row.split("\t")[3]
+                for prefix, flag in self.RMSK_ELEMENTS.items():
+                    if label.startswith(prefix):
+                        flags[flag] = 1
+        except Exception:
+            pass
+        return flags
+
+
+def agg_either_both(a, b):
+    """Order-invariant aggregation for the two breakpoints."""
+    if a == MISSING or b == MISSING:
+        v = a if b == MISSING else b
+        return v, v
+    return (1 if (a or b) else 0), (1 if (a and b) else 0)
+
+
+# ---------------------------------------------------------------------------
+# Small helpers for VCF field access
+# ---------------------------------------------------------------------------
+def info(rec, key, default=None):
+    try:
+        return rec.info[key]
+    except Exception:
+        return default
+
+
+def fmt(rec, key, default=None):
+    try:
+        return rec.samples[0][key]
+    except Exception:
+        return default
+
+
+def is_pass(rec):
+    fk = list(rec.filter.keys())
+    return 1 if (not fk or fk == ["PASS"] or fk == ["."]) else 0
+
+
+def gt_is_hom_alt(rec):
+    gt = fmt(rec, "GT")
+    if not gt or any(a is None for a in gt):
+        return MISSING
+    alleles = [a for a in gt]
+    return 1 if all(a == alleles[0] and a > 0 for a in alleles) else 0
+
+
+def first(x, default=MISSING):
+    """Coerce a possibly-tuple INFO/FORMAT value to a scalar number."""
+    if x is None:
+        return default
+    if isinstance(x, (tuple, list)):
+        x = x[0] if x else default
+    try:
+        return float(x)
+    except Exception:
+        return default
+
+
+def width(ci):
+    if not ci or not isinstance(ci, (tuple, list)) or len(ci) < 2:
+        return MISSING
+    try:
+        return abs(float(ci[1]) - float(ci[0]))
+    except Exception:
+        return MISSING
+
+
+def norm_svtype(rec):
+    st = info(rec, "SVTYPE")
+    if st is None:
+        alt = str(rec.alts[0]) if rec.alts else ""
+        st = alt.strip("<>").split(":")[0] if alt.startswith("<") else "BND"
+    st = str(st).upper().split(":")[0]
+    if st in ("TRA", "CTX"):
+        st = "BND"
+    if st not in ("DEL", "DUP", "INS", "INV", "BND"):
+        st = "BND"
+    return st
+
+
+# ---------------------------------------------------------------------------
+# Per-caller SV parsers  ->  normalized concept dict
+# ---------------------------------------------------------------------------
+def parse_sv_common(rec):
+    st = norm_svtype(rec)
+    chrom = rec.chrom
+    pos = rec.pos
+    # pysam consumes INFO/END into rec.stop; meaningful only for spanned SVs.
+    # BND/INS are annotated at their primary breakend only (bp2 = bp1 via end=None).
+    end = rec.stop if st in ("DEL", "DUP", "INV") else None
+    return {
+        "chrom": chrom, "pos": pos, "end": end, "chrom2": chrom,
+        "svtype": st,
+        "is_pass": is_pass(rec),
+        "cipos_width": width(info(rec, "CIPOS") or info(rec, "CIPOS95")),
+        "ciend_width": width(info(rec, "CIEND") or info(rec, "CIEND95")),
+        "is_imprecise": 1 if ("IMPRECISE" in rec.info) else 0,
+        "gt_hom": gt_is_hom_alt(rec),
+        "svlen_raw": info(rec, "SVLEN"),
+    }
+
+
+def parse_manta(rec):
+    d = parse_sv_common(rec)
+    pr = fmt(rec, "PR") or (None, None)
+    sr = fmt(rec, "SR") or (None, None)
+    pr_ref, pr_alt = (first(pr[0], 0), first(pr[1], 0)) if len(pr) == 2 else (0, 0)
+    sr_ref, sr_alt = (first(sr[0], 0), first(sr[1], 0)) if len(sr) == 2 else (0, 0)
+    tot = pr_ref + pr_alt + sr_ref + sr_alt
+    d.update({
+        "pe_support": pr_alt, "sr_support": sr_alt, "total_support": pr_alt + sr_alt,
+        "vaf": (pr_alt + sr_alt) / tot if tot > 0 else MISSING,
+        "gq": first(fmt(rec, "GQ")), "qual_norm": first(rec.qual),
+        "local_depth": (pr_ref + pr_alt) or first(info(rec, "BND_DEPTH")),
+    })
+    return d
+
+
+def parse_delly(rec):
+    d = parse_sv_common(rec)
+    dr, dv = first(fmt(rec, "DR"), 0), first(fmt(rec, "DV"), 0)
+    rr, rv = first(fmt(rec, "RR"), 0), first(fmt(rec, "RV"), 0)
+    tot = dr + dv + rr + rv
+    if d["svlen_raw"] is None and d["end"] is not None:  # v0.7 has no SVLEN
+        d["svlen_raw"] = d["end"] - d["pos"]
+    d.update({
+        "pe_support": dv, "sr_support": rv, "total_support": dv + rv,
+        "vaf": (dv + rv) / tot if tot > 0 else MISSING,
+        "gq": first(fmt(rec, "GQ")), "qual_norm": first(rec.qual),
+        "local_depth": dr + dv,
+    })
+    return d
+
+
+def parse_lumpy(rec):
+    d = parse_sv_common(rec)
+    ao, ro = first(fmt(rec, "AO"), 0), first(fmt(rec, "RO"), 0)
+    ab = fmt(rec, "AB")
+    # smoove/LUMPY put SU/PE/SR in INFO (site-level), not FORMAT; fall back to FORMAT for other dialects
+    pe = info(rec, "PE"); pe = first(pe) if pe is not None else first(fmt(rec, "PE"), 0)
+    sr = info(rec, "SR"); sr = first(sr) if sr is not None else first(fmt(rec, "SR"), 0)
+    su = info(rec, "SU"); su = first(su) if su is not None else first(fmt(rec, "SU"), 0)
+    d.update({
+        "pe_support": pe, "sr_support": sr, "total_support": su,
+        "vaf": first(ab) if ab is not None else ((ao / (ao + ro)) if (ao + ro) > 0 else MISSING),
+        "gq": first(fmt(rec, "GQ")), "qual_norm": first(fmt(rec, "SQ")),
+        "local_depth": first(fmt(rec, "DP")),
+    })
+    return d
+
+
+SV_PARSERS = {"manta": parse_manta, "delly": parse_delly, "lumpy": parse_lumpy}
+
+
+# ---------------------------------------------------------------------------
+# Per-caller STR parsers
+# ---------------------------------------------------------------------------
+def _split_pair(val, sep):
+    if val is None:
+        return []
+    if isinstance(val, (tuple, list)):  # pysam returns Number=2 fields (e.g. GangSTR REPCN) as tuples
+        out = []
+        for x in val:
+            try:
+                out.append(float(x))
+            except Exception:
+                pass
+        return out
+    s = str(val)
+    for d in sep:
+        s = s.replace(d, "|")
+    out = []
+    for tok in s.split("|"):
+        try:
+            out.append(float(tok))
+        except Exception:
+            pass
+    return out
+
+
+def parse_eh(rec):
+    ru = info(rec, "RU") or ""
+    repcn = _split_pair(fmt(rec, "REPCN"), "/")
+    ref_cn = first(info(rec, "REF"))
+    adsp = sum(_split_pair(fmt(rec, "ADSP"), "/"))
+    adfl = sum(_split_pair(fmt(rec, "ADFL"), "/"))
+    adir = sum(_split_pair(fmt(rec, "ADIR"), "/"))
+    return {
+        "chrom": rec.chrom, "pos": rec.pos, "end": rec.stop,
+        "is_pass": is_pass(rec), "motif_len": float(len(ru)) if ru else first(info(rec, "RL")),
+        "ref_copynum": ref_cn,
+        "repcn": repcn, "repci_raw": fmt(rec, "REPCI"),
+        "spanning_reads": adsp, "flanking_reads": adfl, "inrepeat_reads": adir,
+        "locus_depth": first(fmt(rec, "LC")), "gt_hom": gt_is_hom_alt(rec),
+        "qual_post": first(rec.qual), "ref_tract_bp": first(info(rec, "RL")),
+        "ru": ru,
+    }
+
+
+def parse_gangstr(rec):
+    ru = info(rec, "RU") or ""
+    period = first(info(rec, "PERIOD"))
+    repcn = _split_pair(fmt(rec, "REPCN"), ",")
+    ref_cn = first(info(rec, "REF"))
+    rc = _split_pair(fmt(rec, "RC"), ",")  # enclosing,spanning,FRR,bounding
+    enclosing, spanning, frr, bounding = (rc + [0, 0, 0, 0])[:4]
+    return {
+        "chrom": rec.chrom, "pos": rec.pos, "end": rec.stop,
+        "is_pass": is_pass(rec), "motif_len": period if period != MISSING else float(len(ru)),
+        "ref_copynum": ref_cn,
+        "repcn": repcn, "repci_raw": fmt(rec, "REPCI"),
+        "spanning_reads": enclosing + spanning, "flanking_reads": bounding, "inrepeat_reads": frr,
+        "locus_depth": first(fmt(rec, "DP")), "gt_hom": gt_is_hom_alt(rec),
+        "qual_post": first(fmt(rec, "Q")),
+        "ref_tract_bp": (ref_cn * period) if (ref_cn != MISSING and period != MISSING) else MISSING,
+        "ru": ru,
+    }
+
+
+def _num(x, default=MISSING):
+    try:
+        if x is None or x == "":
+            return default
+        v = float(x)
+        return default if v != v else v  # NaN guard
+    except Exception:
+        return default
+
+
+def parse_eh_tsv(row):
+    """One row of an ExpansionHunter flat TSV:
+    chrom,pos,end,filter,repid,ru,rl,ref,repcn,repci,adsp,adfl,adir,lc,so"""
+    ru = str(row.get("ru") or "")
+    repcn = _split_pair(row.get("repcn"), "/")
+    ref_cn = _num(row.get("ref"))
+    rl = _num(row.get("rl"))
+    adsp = sum(_split_pair(row.get("adsp"), "/"))
+    adfl = sum(_split_pair(row.get("adfl"), "/"))
+    adir = sum(_split_pair(row.get("adir"), "/"))
+    gt_hom = MISSING
+    if len(repcn) >= 2:  # hom-ALT = both alleles equal and differ from reference
+        gt_hom = 1 if (repcn[0] == repcn[1] and repcn[0] != ref_cn) else 0
+    return {
+        "chrom": str(row["chrom"]), "pos": int(float(row["pos"])), "end": _num(row.get("end")),
+        "is_pass": 1 if str(row.get("filter", "")).upper() == "PASS" else 0,
+        "motif_len": float(len(ru)) if ru else rl,
+        "ref_copynum": ref_cn,
+        "repcn": repcn, "repci_raw": row.get("repci"),
+        "spanning_reads": adsp, "flanking_reads": adfl, "inrepeat_reads": adir,
+        "locus_depth": _num(row.get("lc")), "gt_hom": gt_hom,
+        "qual_post": MISSING,  # EH TSV carries no site quality
+        "ref_tract_bp": rl, "ru": ru,
+    }
+
+
+STR_PARSERS = {"expansionhunter": parse_eh, "gangstr": parse_gangstr}
+
+
+def repci_width_max(repci_raw):
+    """Max allele CI width. EH: '2-2/10-10' (str); GangSTR: ('1-2','2-2') (pysam tuple)."""
+    if repci_raw is None:
+        return MISSING
+    if isinstance(repci_raw, (tuple, list)):
+        alleles = [str(x) for x in repci_raw]
+    else:
+        alleles = str(repci_raw).replace("/", ",").split(",")
+    best = MISSING
+    for allele in alleles:
+        if "-" in allele:
+            try:
+                parts = allele.split("-")
+                w = abs(float(parts[1]) - float(parts[0]))
+                best = w if best == MISSING else max(best, w)
+            except Exception:
+                pass
+    return best
+
+
+# ---------------------------------------------------------------------------
+# Feature assembly
+# ---------------------------------------------------------------------------
+def sv_features(d, ann, fasta, win):
+    chrom, pos, end = d["chrom"], d["pos"], d["end"]
+    chrom2, end2 = d["chrom2"], (end if end is not None else pos)
+    st = d["svtype"]
+    svlen = first(d["svlen_raw"])
+    f = {
+        "variant_ID": f"{chrom}:{pos}:{st}:{end}",
+        "is_pass": d["is_pass"],
+        "svtype_DEL": int(st == "DEL"), "svtype_DUP": int(st == "DUP"),
+        "svtype_INS": int(st == "INS"), "svtype_INV": int(st == "INV"),
+        "svtype_BND": int(st == "BND"),
+        "svlen_log": math.log10(abs(svlen) + 1) if svlen != MISSING else MISSING,
+        "cipos_width": d["cipos_width"], "ciend_width": d["ciend_width"],
+        "is_imprecise": d["is_imprecise"],
+        "pe_support": d["pe_support"], "sr_support": d["sr_support"],
+        "total_support": d["total_support"], "vaf": d["vaf"],
+        "gt_hom": d["gt_hom"], "gq": d["gq"], "qual_norm": d["qual_norm"],
+        "local_depth": d["local_depth"],
+    }
+    # reference sequence context at both breakpoints
+    gc1, e1 = gc_entropy_at(fasta, chrom, pos, win)
+    gc2, e2 = gc_entropy_at(fasta, chrom2, end2, win)
+    f["gc_min"], f["gc_max"] = (min(gc1, gc2), max(gc1, gc2)) if MISSING not in (gc1, gc2) else (MISSING, MISSING)
+    f["entropy_min"] = min(e1, e2) if MISSING not in (e1, e2) else MISSING
+    f["microhom_max"] = microhomology(fasta, chrom, pos, end if chrom2 == chrom else None)
+    # GIAB binary overlap, both breakpoints
+    for name, key in (("segdups", "segdup"), ("difficult", "difficult")):
+        ei, bo = agg_either_both(ann.overlaps(name, chrom, pos), ann.overlaps(name, chrom2, end2))
+        f[f"in_{key}_either"], f[f"in_{key}_both"] = ei, bo
+    for name, key in (("lowmap", "lowmap"), ("tandem", "tandem")):
+        ei, _ = agg_either_both(ann.overlaps(name, chrom, pos), ann.overlaps(name, chrom2, end2))
+        f[f"in_{key}_either"] = ei
+    # RepeatMasker element class, either breakpoint
+    r1 = ann.rmsk_elements(chrom, pos)
+    r2 = ann.rmsk_elements(chrom2, end2)
+    for elt in ("Alu", "L1", "SVA", "LTR"):
+        ei, _ = agg_either_both(r1[elt], r2[elt])
+        f[f"in_{elt}_either"] = ei
+    # fraction of the SV interval covered by repeats (intra-chrom interval SVs only)
+    if st in ("DEL", "DUP", "INV") and end is not None and chrom2 == chrom:
+        f["frac_span_repeat"] = max(ann.frac_overlap("tandem", chrom, pos, end),
+                                    ann.frac_overlap("segdups", chrom, pos, end))
+    else:
+        f["frac_span_repeat"] = MISSING
+    # neighbor density (SV only) — precomputed onto d by compute_clustering()
+    f["n_neighbors"] = d.get("n_neighbors", 0)
+    f["nn_log_dist"] = d.get("nn_log_dist", MISSING)
+    return f
+
+
+def str_features(d, ann, fasta, win, read_len):
+    chrom, pos = d["chrom"], d["pos"]
+    repcn = d["repcn"] or []
+    cn_max = max(repcn) if repcn else MISSING
+    cn_min = min(repcn) if repcn else MISSING
+    ref_cn = d["ref_copynum"]
+    motif = d["motif_len"]
+    f = {
+        "variant_ID": f"{chrom}:{pos}:{info_end(d)}",
+        "is_pass": d["is_pass"], "motif_len": motif, "ref_copynum": ref_cn,
+        "gt_repcn_max": cn_max, "gt_repcn_min": cn_min,
+        "expansion_over_ref": (cn_max - ref_cn) if MISSING not in (cn_max, ref_cn) else MISSING,
+        "repci_width_max": repci_width_max(d["repci_raw"]),
+        "spanning_reads": d["spanning_reads"], "flanking_reads": d["flanking_reads"],
+        "inrepeat_reads": d["inrepeat_reads"],
+        "locus_depth": d["locus_depth"], "gt_hom": d["gt_hom"],
+        # qual_post dropped: EH never emits it -> structurally-missing -> caller-identity proxy
+        "ref_tract_bp": d["ref_tract_bp"],
+    }
+    tot = d["spanning_reads"] + d["flanking_reads"] + d["inrepeat_reads"]
+    f["spanning_frac"] = d["spanning_reads"] / tot if tot > 0 else MISSING
+    f["allele_vs_readlen"] = (cn_max * motif / read_len) if MISSING not in (cn_max, motif) else MISSING
+    f["motif_is_homopolymer"] = int(motif == 1) if motif != MISSING else MISSING
+    gc, ent = gc_entropy_at(fasta, chrom, pos, win)
+    f["gc_flank"], f["entropy_flank"] = gc, ent
+    f["in_segdup"] = ann.overlaps("segdups", chrom, pos)
+    f["in_difficult"] = ann.overlaps("difficult", chrom, pos)
+    f["flank_lowmap"] = ann.overlaps("lowmap", chrom, pos)
+    return f
+
+
+def info_end(d):
+    return int(d["end"]) if d.get("end") is not None else d["pos"]
+
+
+# ---------------------------------------------------------------------------
+# Clustering (SV) — within-callset neighbor density
+# ---------------------------------------------------------------------------
+def compute_clustering(parsed, radius):
+    """Set on each parsed SV dict:
+      nn_log_dist  = log10(distance to nearest other call + 1), UNCAPPED (isolation).
+      n_neighbors  = number of other calls within +/-radius.
+    SV calls are sparse (median nearest neighbor ~5-90 kb), so radius must be SV-scale
+    (default 100 kb), not the 1 kb used for dense small variants. Vectorized per chrom."""
+    by_chrom = defaultdict(list)
+    for j, d in enumerate(parsed):
+        by_chrom[d["chrom"]].append((d["pos"], j))
+    for items in by_chrom.values():
+        items.sort()
+        pos = np.array([p for p, _ in items])
+        n = len(pos)
+        for k, (_, j) in enumerate(items):
+            if n < 2:
+                parsed[j]["nn_log_dist"], parsed[j]["n_neighbors"] = MISSING, 0
+                continue
+            p = pos[k]
+            nearest = min((p - pos[k - 1]) if k > 0 else float("inf"),
+                          (pos[k + 1] - p) if k < n - 1 else float("inf"))
+            parsed[j]["nn_log_dist"] = math.log10(nearest + 1)
+            lo = int(np.searchsorted(pos, p - radius, "left"))
+            hi = int(np.searchsorted(pos, p + radius, "right"))
+            parsed[j]["n_neighbors"] = hi - lo - 1  # exclude self
+
+
+# ---------------------------------------------------------------------------
+# Main
+# ---------------------------------------------------------------------------
+def main():
+    ap = argparse.ArgumentParser(description="SVSTR_Score feature builder")
+    ap.add_argument("--vcf", required=True)
+    ap.add_argument("--caller", required=True,
+                    choices=sorted(SV_CALLERS | STR_CALLERS))
+    ap.add_argument("--fasta", required=True)
+    ap.add_argument("--giab-dir", default=None, help="dir with segdups/lowmap/tandem/difficult .bed.gz (tabixed)")
+    ap.add_argument("--repeatmasker", default=None, help="tabixed rmsk_class.bed.gz")
+    ap.add_argument("--win", type=int, default=50, help="GC/entropy window (+/- bp)")
+    ap.add_argument("--neighbor-radius", type=int, default=100000,
+                    help="SV clustering radius for n_neighbors (+/- bp). Default 100kb — SV calls are "
+                         "sparse (median nearest ~5-90kb); 1kb is for dense small variants.")
+    ap.add_argument("--read-len", type=int, default=150, help="short-read length (STR spanning feasibility)")
+    ap.add_argument("--primary-only", dest="primary_only", action="store_true", default=True,
+                    help="keep only primary-assembly contigs chr1-22,X,Y,M (default on)")
+    ap.add_argument("--all-contigs", dest="primary_only", action="store_false",
+                    help="include ALT/decoy/HLA contigs (off by default)")
+    ap.add_argument("--str-drop-homref", action="store_true",
+                    help="(STR) drop hom-ref 0/0 genotype loci (catalog non-variants)")
+    ap.add_argument("--sample", default=None,
+                    help="sample id (default: auto from VCF's single sample, or EH-TSV filename prefix). "
+                         "Emitted as a `sample` column — the label join key with the truth set.")
+    ap.add_argument("--missing-indicators", action="store_true",
+                    help="also emit <feat>_missing 0/1 columns. OFF by default: redundant for tree "
+                         "models (the -99999 sentinel is already split-separable). Turn on for linear/NN models.")
+    ap.add_argument("-o", "--output", required=True)
+    args = ap.parse_args()
+
+    variant_class = "SV" if args.caller in SV_CALLERS else "STR"
+    fasta = pysam.FastaFile(args.fasta)
+    ann = Annotator(args.giab_dir, args.repeatmasker)
+
+    eh_tsv = (args.caller == "expansionhunter")  # EH ships a flat (gzipped) TSV, not a VCF
+    if eh_tsv:
+        with open(args.vcf, "rb") as fh:
+            comp = "gzip" if fh.read(2) == b"\x1f\x8b" else None
+        records = pd.read_csv(args.vcf, sep="\t", dtype=str, compression=comp).to_dict("records")
+        sample = args.sample or os.path.basename(args.vcf).split(".")[0]
+        get_chrom = lambda r: str(r["chrom"])
+        def is_homref(r):
+            cn, ref = _split_pair(r.get("repcn"), "/"), _num(r.get("ref"))
+            return bool(cn) and all(x == ref for x in cn)
+    else:
+        vf = pysam.VariantFile(args.vcf)
+        hdr = list(vf.header.samples)
+        sample = args.sample or (hdr[0] if len(hdr) == 1 else None)
+        if sample is None:
+            sys.exit(f"[error] --sample required: VCF has {len(hdr)} samples {hdr}")
+        records = list(vf)
+        get_chrom = lambda r: r.chrom
+        is_homref = lambda r: not (set(fmt(r, "GT") or ()) - {0})
+    sys.stderr.write(f"[info] sample={sample}\n")
+
+    n_raw = len(records)
+    if args.primary_only:
+        records = [r for r in records if get_chrom(r) in PRIMARY_CONTIGS]
+        sys.stderr.write(f"[info] primary-only: dropped {n_raw - len(records):,} non-primary-contig records\n")
+    if variant_class == "STR" and args.str_drop_homref:
+        before = len(records)
+        records = [r for r in records if not is_homref(r)]
+        sys.stderr.write(f"[info] str-drop-homref: dropped {before - len(records):,} hom-ref loci\n")
+    sys.stderr.write(f"[info] {len(records):,} records to process | caller={args.caller} class={variant_class}\n")
+
+    rows = []
+    if variant_class == "SV":
+        parser = SV_PARSERS[args.caller]
+        parsed = [parser(r) for r in records]
+        compute_clustering(parsed, args.neighbor_radius)
+        for d in parsed:
+            f = sv_features(d, ann, fasta, args.win)
+            f["caller"] = args.caller
+            rows.append(f)
+    else:
+        parser = parse_eh_tsv if eh_tsv else STR_PARSERS[args.caller]
+        for r in records:
+            d = parser(r)
+            f = str_features(d, ann, fasta, args.win, args.read_len)
+            f["caller"] = args.caller
+            rows.append(f)
+
+    out = pd.DataFrame(rows)
+    out["sample"] = sample
+    # Missingness is carried by the -99999 sentinel in each feature (trees split on it
+    # directly). Optional explicit indicators (fixed list -> stable schema) for linear/NN.
+    if args.missing_indicators:
+        indicators = SV_MISSING_INDICATORS if variant_class == "SV" else STR_MISSING_INDICATORS
+        for col in indicators:
+            out[f"{col}_missing"] = (out[col] == MISSING).astype(int) if col in out.columns else 0
+    # meta (label join key) first: sample, caller, variant_ID — NOT model features
+    meta = [c for c in ("sample", "caller", "variant_ID") if c in out.columns]
+    out = out[meta + [c for c in out.columns if c not in meta]]
+    out.to_csv(args.output, sep="\t", index=False)
+    sys.stderr.write(f"[info] wrote {len(out):,} rows x {out.shape[1]} cols -> {args.output}\n")
+
+
+if __name__ == "__main__":
+    main()

feature_manifest.json

@@ -1,54 +1,64 @@
 {
+  "release_version": "1.0",
   "sv_features": [
-    "qual",
-    "gq",
-    "pr_ref",
-    "pr_alt",
-    "sr_ref",
-    "sr_alt",
-    "vf_ref",
-    "vf_alt",
-    "total_alt_support",
-    "sr_pr_ratio",
-    "vaf_estimate",
-    "strand_bias_fs",
-    "cipos_width",
-    "ciend_width",
-    "homlen",
-    "svlen_abs",
+    "is_pass",
     "svtype_DEL",
-    "svtype_INS",
     "svtype_DUP",
+    "svtype_INS",
+    "svtype_INV",
     "svtype_BND",
+    "svlen_log",
+    "cipos_width",
+    "ciend_width",
     "is_imprecise",
-    "filter_pass",
-    "n_filter_flags"
+    "pe_support",
+    "sr_support",
+    "total_support",
+    "vaf",
+    "gt_hom",
+    "gq",
+    "qual_norm",
+    "local_depth",
+    "gc_min",
+    "gc_max",
+    "entropy_min",
+    "microhom_max",
+    "in_segdup_either",
+    "in_segdup_both",
+    "in_difficult_either",
+    "in_difficult_both",
+    "in_lowmap_either",
+    "in_tandem_either",
+    "in_Alu_either",
+    "in_L1_either",
+    "in_SVA_either",
+    "in_LTR_either",
+    "frac_span_repeat",
+    "n_neighbors",
+    "nn_log_dist"
   ],
   "str_features": [
-    "repcn_a1",
-    "repcn_a2",
-    "ru_length",
-    "ref_count",
-    "delta_from_ref_a1",
-    "delta_from_ref_a2",
-    "ci_width_a1",
-    "ci_width_a2",
-    "adsp_a1",
-    "adsp_a2",
-    "adfl_a1",
-    "adfl_a2",
-    "adir_a1",
-    "adir_a2",
-    "total_support_a1",
-    "total_support_a2",
-    "spanning_frac_a1",
-    "spanning_frac_a2",
-    "locus_coverage",
-    "allele_balance",
-    "support_type_a1",
-    "support_type_a2",
-    "is_low_depth",
-    "locus_conc_rate",
-    "locus_in_lookup"
-  ]
+    "is_pass",
+    "motif_len",
+    "ref_copynum",
+    "gt_repcn_max",
+    "gt_repcn_min",
+    "expansion_over_ref",
+    "repci_width_max",
+    "spanning_reads",
+    "flanking_reads",
+    "inrepeat_reads",
+    "locus_depth",
+    "gt_hom",
+    "ref_tract_bp",
+    "spanning_frac",
+    "allele_vs_readlen",
+    "motif_is_homopolymer",
+    "gc_flank",
+    "entropy_flank",
+    "in_segdup",
+    "in_difficult",
+    "flank_lowmap"
+  ],
+  "note": "Produced by feature_builder.py from a short-read VCF + reference FASTA + static BEDs; missing fields use the -99999 sentinel."
 }

requirements.txt

@@ -1,4 +1,4 @@
-scikit-learn==1.5.1
+scikit-learn==1.7.1
 pandas>=2.0
 numpy>=1.24
 joblib>=1.3

score_svstr.py

@@ -1,110 +1,91 @@
 #!/usr/bin/env python3
 """
 score_svstr.py — apply the SVSTR-Score confidence model to short-read SV or STR
-calls and emit a per-call confidence score (CS) and tier.
+calls and emit a per-call calibrated confidence score (CS) and tier.
 
-This is the inference entry point for the released models
-(sv_model_v13_parents.joblib / str_model_v13_parents.joblib). It loads the
-trained random forest + its feature/config sidecar and (for STR) the per-locus
-catalogue lookup, then scores a tabular feature matrix extracted from the
-caller VCF.
+Inference entry point for the released models (sv_model.joblib / str_model.joblib,
+each paired with an isotonic calibrator). It loads the trained random forest + its
+isotonic calibrator + the feature/config sidecar, then scores a tabular feature
+matrix produced by `feature_builder.py` from the caller VCF (short-read VCF +
+reference FASTA + static annotation BEDs):
 
-TIERS (as published):
-  HIGH      CS >= 0.70
+    CS = isotonic_calibrator( RF.predict_proba(X)[:, 1] )    # P(concordant)
+
+TIERS:
+  HIGH      CS >= 0.70        (candidate-triage filter)
   MODERATE  0.50 <= CS < 0.70
   WARNING   0.30 <= CS < 0.50
   LOW       CS < 0.30
-Override: the deployed quality-demotion rules in each `*_config.json` ("rules")
-are applied on top of the CS tier. A call flagged by ANY rule is reassigned to
-the config's `override_target` regardless of CS — SV -> LOW (5 rules:
-QUAL<20, GQ<15, total_alt_support<=2, vaf_estimate<0.15, is_imprecise==1);
-STR -> WARNING (6 rules: locus_coverage<20, is_low_depth==1, support_type_a2<=1,
-total_support_a2<5, ci_width_a2>3, allele_balance<0.3). Only the MODERATE-to-HIGH
-range is a monotone precision ladder; LOW/WARNING are not.
-
-INPUT (--features): tab/comma table extracted from the caller VCF.
-  SV  : the 23 features in sv_config.json  (+ chrom,pos optional for provenance)
-  STR : the 23 caller-output features in str_config.json EXCLUDING locus_conc_rate
-        and locus_in_lookup, PLUS chrom,pos (used to join the catalogue lookup).
-        locus_conc_rate / locus_in_lookup are filled here from the released lookup.
+
+The score is isotonic-calibrated, so the tier is a pure bucket of the calibrated
+CS — there are no heuristic override rules, and STR needs no per-locus catalogue
+lookup (its features are self-contained). Missing features (fields a merged or
+filtered callset may not carry) are filled with the -99999 sentinel that the
+trees were trained to split on.
 
 USAGE
-  python score_svstr.py --variant sv  --model-dir . --features sv_feats.tsv  --out sv_scored.tsv
-  python score_svstr.py --variant str --model-dir . --features str_feats.tsv --out str_scored.tsv
+  python score_svstr.py --variant sv  --model-dir . --features sv_features.tsv  --out sv_scored.tsv
+  python score_svstr.py --variant str --model-dir . --features str_features.tsv --out str_scored.tsv
 
-Requires the package versions in requirements.txt (scikit-learn==1.5.1).
-Licence: MIT.
+Requires the versions in requirements.txt (scikit-learn==1.7.1). Licence: MIT.
 """
 import argparse, json, os, sys
 import numpy as np
 import pandas as pd
 import joblib
 
-# out-of-catalogue STR loci: published convention (Methods) — conservative score
-STR_MISS_LOCUS_CONC_RATE = 0.0
-STR_MISS_LOCUS_IN_LOOKUP = 0
-
-_OP = {"<": lambda a, v: a < v, "<=": lambda a, v: a <= v,
-       ">": lambda a, v: a > v, ">=": lambda a, v: a >= v, "==": lambda a, v: a == v}
+MISSING = -99999.0
+TIER_EDGES = (0.30, 0.50, 0.70)
+TIER_NAMES = ("LOW", "WARNING", "MODERATE", "HIGH")
 
 
-def assign_tier(cs, df, cfg):
-    """CS tier (HIGH/MODERATE/WARNING/LOW), then apply the config's deterministic
-    quality-demotion rules: any rule triggered -> cfg['override_target'] (SV: LOW,
-    STR: WARNING), regardless of CS. NaN feature values never trigger a rule."""
-    tier = np.where(cs >= 0.70, "HIGH",
-           np.where(cs >= 0.50, "MODERATE",
-           np.where(cs >= 0.30, "WARNING", "LOW")))
-    rules = cfg.get("rules", {})
-    target = cfg.get("override_target", "LOW")
-    trig = np.zeros(len(cs), dtype=bool)
-    for r in rules.values():
-        f = r["feature"]
-        if f in df.columns:
-            v = pd.to_numeric(df[f], errors="coerce").values
-            trig = trig | np.nan_to_num(_OP[r["op"]](v, r["value"]), nan=False).astype(bool)
-    return np.where(trig, target, tier)
+def to_tier(cs):
+    return np.asarray(TIER_NAMES)[np.digitize(np.asarray(cs, float), TIER_EDGES, right=False)]
 
 
 def main():
     ap = argparse.ArgumentParser(description=__doc__,
                                  formatter_class=argparse.RawDescriptionHelpFormatter)
     ap.add_argument("--variant", choices=["sv", "str"], required=True)
-    ap.add_argument("--model-dir", default=".", help="dir with *_config.json, *.joblib, lookup")
-    ap.add_argument("--features", required=True, help="feature table (tsv/csv)")
+    ap.add_argument("--model-dir", default=".", help="dir with *_config.json + *.joblib")
+    ap.add_argument("--features", required=True,
+                    help="feature table from feature_builder.py (tsv/csv[.gz])")
     ap.add_argument("--out", required=True)
+    ap.add_argument("--raw", action="store_true", help="also emit the uncalibrated RF score (CS_raw)")
     a = ap.parse_args()
 
     cfg = json.load(open(os.path.join(a.model_dir, f"{a.variant}_config.json")))
     model = joblib.load(os.path.join(a.model_dir, cfg["model_file"]))
+    cal = joblib.load(os.path.join(a.model_dir, cfg["calibrator_file"]))
     feats = cfg["features"]
 
     sep = "\t" if a.features.endswith((".tsv", ".txt", ".gz")) else ","
     df = pd.read_csv(a.features, sep=sep)
 
-    if a.variant == "str":
-        lk = pd.read_parquet(os.path.join(a.model_dir, cfg["lookup_file"]))
-        keys = cfg.get("lookup_keys", ["chrom", "pos"])
-        rate_col = [c for c in lk.columns if c not in keys][0] if "locus_conc_rate" not in lk.columns else "locus_conc_rate"
-        lk = lk.rename(columns={rate_col: "locus_conc_rate"})[keys + ["locus_conc_rate"]]
-        df = df.merge(lk, on=keys, how="left")
-        df["locus_in_lookup"] = df["locus_conc_rate"].notna().astype(int)
-        df["locus_conc_rate"] = df["locus_conc_rate"].fillna(STR_MISS_LOCUS_CONC_RATE)
-        df.loc[df["locus_in_lookup"] == 0, "locus_in_lookup"] = STR_MISS_LOCUS_IN_LOOKUP
-
-    missing = [f for f in feats if f not in df.columns]
-    if missing:
-        sys.exit(f"ERROR: input is missing required features: {missing}")
-
-    X = df[feats]
-    cs = model.predict_proba(X)[:, 1]
-    df["CS"] = np.round(cs, 4)
-    df["tier"] = assign_tier(cs, df, cfg)
-
-    df.to_csv(a.out, sep="\t", index=False)
-    n = len(df)
-    hi = int((df["tier"] == "HIGH").sum())
-    print(f"{a.variant.upper()}: scored {n:,} calls; HIGH {hi:,} ({hi/n:.1%})", file=sys.stderr)
+    absent = [f for f in feats if f not in df.columns]
+    if absent:
+        print(f"[warn] {len(absent)} model features absent -> -99999 sentinel: {absent}",
+              file=sys.stderr)
+        for f in absent:
+            df[f] = MISSING
+
+    X = df[feats].astype("float32")
+    p = model.predict_proba(X)
+    raw = p[:, list(model.classes_).index(1)] if p.shape[1] > 1 else p[:, 0]
+    cs = np.clip(cal.predict(raw), 0.0, 1.0)
+
+    out = df.copy()
+    if a.raw:
+        out["CS_raw"] = np.round(raw, 4)
+    out["CS"] = np.round(cs, 4)
+    out["tier"] = to_tier(cs)
+    out.to_csv(a.out, sep="\t", index=False)
+
+    n = len(out)
+    vc = pd.Series(out["tier"]).value_counts()
+    hi = int(vc.get("HIGH", 0))
+    print(f"{a.variant.upper()}: scored {n:,} calls; HIGH {hi:,} ({hi/n:.1%}) | "
+          + " ".join(f"{t}:{int(vc.get(t,0)):,}" for t in TIER_NAMES), file=sys.stderr)
     print(f"wrote {a.out}", file=sys.stderr)
 
 

str_locus_lookup.parquet → str_calibrator.joblib

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f9e0f5d6becccb4f130486de1eb80341b44e8e1d66b935a41e5c81f9ae79f8e6
-size 2356330
+oid sha256:3af5671f62a57815797d4c3b71b1c5b5b0102d36d9275651b3ba088a3827e453
+size 16219

str_config.json

@@ -1,74 +1,53 @@
 {
-  "model_file": "str_model_v13_parents.joblib",
+  "release_version": "1.0",
+  "model_file": "str_model.joblib",
+  "calibrator_file": "str_calibrator.joblib",
+  "calibration": "isotonic regression on out-of-fold scores",
   "features": [
-    "repcn_a1",
-    "repcn_a2",
-    "ru_length",
-    "ref_count",
-    "delta_from_ref_a1",
-    "delta_from_ref_a2",
-    "ci_width_a1",
-    "ci_width_a2",
-    "adsp_a1",
-    "adsp_a2",
-    "adfl_a1",
-    "adfl_a2",
-    "adir_a1",
-    "adir_a2",
-    "total_support_a1",
-    "total_support_a2",
-    "spanning_frac_a1",
-    "spanning_frac_a2",
-    "locus_coverage",
-    "allele_balance",
-    "support_type_a1",
-    "support_type_a2",
-    "is_low_depth",
-    "locus_conc_rate",
-    "locus_in_lookup"
+    "is_pass",
+    "motif_len",
+    "ref_copynum",
+    "gt_repcn_max",
+    "gt_repcn_min",
+    "expansion_over_ref",
+    "repci_width_max",
+    "spanning_reads",
+    "flanking_reads",
+    "inrepeat_reads",
+    "locus_depth",
+    "gt_hom",
+    "ref_tract_bp",
+    "spanning_frac",
+    "allele_vs_readlen",
+    "motif_is_homopolymer",
+    "gc_flank",
+    "entropy_flank",
+    "in_segdup",
+    "in_difficult",
+    "flank_lowmap"
   ],
-  "tier_thresholds": {
-    "low": 0.3,
-    "high": 0.7
+  "n_features": 21,
+  "missing_sentinel": -99999.0,
+  "tiers": {
+    "HIGH": "CS>=0.70",
+    "MODERATE": "0.50<=CS<0.70",
+    "WARNING": "0.30<=CS<0.50",
+    "LOW": "CS<0.30"
   },
-  "rules": {
-    "low_coverage": {
-      "feature": "locus_coverage",
-      "op": "<",
-      "value": 20
-    },
-    "low_depth_flag": {
-      "feature": "is_low_depth",
-      "op": "==",
-      "value": 1
-    },
-    "no_spanning_a2": {
-      "feature": "support_type_a2",
-      "op": "<=",
-      "value": 1
-    },
-    "low_support_a2": {
-      "feature": "total_support_a2",
-      "op": "<",
-      "value": 5
-    },
-    "wide_ci_a2": {
-      "feature": "ci_width_a2",
-      "op": ">",
-      "value": 3
-    },
-    "low_allele_balance": {
-      "feature": "allele_balance",
-      "op": "<",
-      "value": 0.3
-    }
-  },
-  "lookup_file": "str_locus_lookup.parquet",
-  "lookup_keys": [
-    "chrom",
-    "pos"
+  "tier_edges": [
+    0.3,
+    0.5,
+    0.7
   ],
-  "sklearn_version_trained": "1.5.1",
+  "score": "CS = isotonic-calibrated P(call concordant with long-read truth)",
   "variant_class": "STR",
-  "override_target": "WARNING"
+  "sklearn_version_trained": "1.7.1",
+  "training": {
+    "cohort": "HPRC",
+    "n_samples": 208,
+    "n_train_rows": 22651133,
+    "cv": "5-fold GroupKFold by sample",
+    "oof_auroc": 0.8342,
+    "oof_auprc": 0.886
+  }
 }

str_model_v13_parents.joblib → str_model.joblib

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a8a3f4a19de1cfd46ea3241b6c2ad31b4f71bddcd2e2a23d0e7f1c35d54bf044
-size 1067649121
+oid sha256:589539ef52b9c0d6518ac7e6a7beda82abfe767d7a5e27fb1dd9a55099f373d5
+size 512205159

str_model_meta.json

@@ -0,0 +1,448 @@
+{
+  "variant": "str",
+  "created_unix": 1782043477,
+  "feature_cols": [
+    "is_pass",
+    "motif_len",
+    "ref_copynum",
+    "gt_repcn_max",
+    "gt_repcn_min",
+    "expansion_over_ref",
+    "repci_width_max",
+    "spanning_reads",
+    "flanking_reads",
+    "inrepeat_reads",
+    "locus_depth",
+    "gt_hom",
+    "ref_tract_bp",
+    "spanning_frac",
+    "allele_vs_readlen",
+    "motif_is_homopolymer",
+    "gc_flank",
+    "entropy_flank",
+    "in_segdup",
+    "in_difficult",
+    "flank_lowmap"
+  ],
+  "n_features": 21,
+  "tier_edges": [
+    0.3,
+    0.5,
+    0.7
+  ],
+  "tier_names": [
+    "LOW",
+    "Warning",
+    "Moderate",
+    "High"
+  ],
+  "missing_sentinel": -99999.0,
+  "rf_params": {
+    "bootstrap": true,
+    "ccp_alpha": 0.0,
+    "class_weight": "balanced_subsample",
+    "criterion": "gini",
+    "max_depth": null,
+    "max_features": "sqrt",
+    "max_leaf_nodes": null,
+    "max_samples": 2000000,
+    "min_impurity_decrease": 0.0,
+    "min_samples_leaf": 50,
+    "min_samples_split": 2,
+    "min_weight_fraction_leaf": 0.0,
+    "monotonic_cst": null,
+    "n_estimators": 300,
+    "n_jobs": -1,
+    "oob_score": false,
+    "random_state": 42,
+    "verbose": 0,
+    "warm_start": false
+  },
+  "n_train_rows": 22651133,
+  "n_samples": 208,
+  "qc": {
+    "label_rows_raw": 36254400,
+    "label_dist_raw": {
+      "concordant": 21350382,
+      "discordant": 13838163,
+      "unlabeled": 1065855
+    },
+    "label_rows_usable": 35188545,
+    "ambiguous_keys_dropped": 0,
+    "ambiguous_feat_rows": 0,
+    "ambiguous_label_rows": 0,
+    "dup_keys_feature": 0,
+    "dup_keys_label": 0,
+    "merged_rows": 22651133,
+    "match_rate_vs_labels": 0.6437075758602693,
+    "match_rate_vs_features": 0.9832673629385175,
+    "class_balance": {
+      "concordant": 13960015,
+      "discordant": 8691118
+    },
+    "concordant_rate": 0.6163053742168217
+  },
+  "cv_folds": 5,
+  "cv_fold_metrics": [
+    {
+      "n": 4469639,
+      "pos_rate": 0.6172603648751052,
+      "auroc": 0.8345731588413778,
+      "auprc": 0.8868311937682424,
+      "brier": 0.16715199887480572,
+      "logloss": 0.505031384190826,
+      "fold": 0,
+      "seconds": 404.5
+    },
+    {
+      "n": 4469658,
+      "pos_rate": 0.6172628867801518,
+      "auroc": 0.8348793797657998,
+      "auprc": 0.8871277104956028,
+      "brier": 0.16710046702995693,
+      "logloss": 0.5048207582711781,
+      "fold": 1,
+      "seconds": 457.3
+    },
+    {
+      "n": 4569998,
+      "pos_rate": 0.6173429397562099,
+      "auroc": 0.8345632397054213,
+      "auprc": 0.8867279699640327,
+      "brier": 0.16717765756008388,
+      "logloss": 0.5050623605875793,
+      "fold": 2,
+      "seconds": 480.6
+    },
+    {
+      "n": 4570859,
+      "pos_rate": 0.6168989242503433,
+      "auroc": 0.8350534258010407,
+      "auprc": 0.8870572426757822,
+      "brier": 0.1669604630273807,
+      "logloss": 0.5044600822147348,
+      "fold": 3,
+      "seconds": 546.9
+    },
+    {
+      "n": 4570979,
+      "pos_rate": 0.6128043904817765,
+      "auroc": 0.8317845587452297,
+      "auprc": 0.8823297885222531,
+      "brier": 0.16790578845588436,
+      "logloss": 0.5066430427730261,
+      "fold": 4,
+      "seconds": 558.6
+    }
+  ],
+  "cv_report": {
+    "overall": {
+      "n": 22651133,
+      "pos_rate": 0.6163053742168217,
+      "auroc": 0.8341539493365068,
+      "auprc": 0.885996637709877,
+      "brier": 0.16726047042063633,
+      "logloss": 0.5052060179718258
+    },
+    "calibration": [
+      {
+        "bin": "[0.0,0.1)",
+        "n": 759079,
+        "mean_pred": 0.06623314081824806,
+        "obs_rate": 0.027333123429840636
+      },
+      {
+        "bin": "[0.1,0.2)",
+        "n": 1807689,
+        "mean_pred": 0.15353118408631086,
+        "obs_rate": 0.1398288090484591
+      },
+      {
+        "bin": "[0.2,0.3)",
+        "n": 2278662,
+        "mean_pred": 0.250703986073481,
+        "obs_rate": 0.2854271497922904
+      },
+      {
+        "bin": "[0.3,0.4)",
+        "n": 2401825,
+        "mean_pred": 0.35114505321433914,
+        "obs_rate": 0.4219845325950059
+      },
+      {
+        "bin": "[0.4,0.5)",
+        "n": 2503890,
+        "mean_pred": 0.4496778698066448,
+        "obs_rate": 0.5559477453083003
+      },
+      {
+        "bin": "[0.5,0.6)",
+        "n": 2743182,
+        "mean_pred": 0.5514420283736253,
+        "obs_rate": 0.6633803371413198
+      },
+      {
+        "bin": "[0.6,0.7)",
+        "n": 3201411,
+        "mean_pred": 0.6513120336728542,
+        "obs_rate": 0.7673941271520589
+      },
+      {
+        "bin": "[0.7,0.8)",
+        "n": 2972899,
+        "mean_pred": 0.7478180823491758,
+        "obs_rate": 0.8596629081579966
+      },
+      {
+        "bin": "[0.8,0.9)",
+        "n": 2979925,
+        "mean_pred": 0.8513437073854806,
+        "obs_rate": 0.9412015403072225
+      },
+      {
+        "bin": "[0.9,1.0)",
+        "n": 1002571,
+        "mean_pred": 0.9221679799864609,
+        "obs_rate": 0.9910769411842154
+      }
+    ],
+    "per_sample_auroc": {
+      "n_samples": 208,
+      "median": 0.8353140721290141,
+      "p25": 0.8326614184016954,
+      "p75": 0.8373927525350378,
+      "min": 0.740174387702103,
+      "max": 0.8401855333526593
+    },
+    "by_homopolymer": {
+      "homopolymer": {
+        "n": 176,
+        "pos_rate": 0.0,
+        "auroc": null,
+        "auprc": null,
+        "brier": 0.12461994174893026
+      },
+      "other": {
+        "n": 22650957,
+        "pos_rate": 0.6163101629657414,
+        "auroc": 0.8341526308855854,
+        "auprc": 0.8859973231761953,
+        "brier": 0.16726080174142982,
+        "logloss": 0.5052065639352175
+      }
+    },
+    "by_is_pass": {
+      "PASS": {
+        "n": 22645309,
+        "pos_rate": 0.6163365225000904,
+        "auroc": 0.8341536917536043,
+        "auprc": 0.8860084593752011,
+        "brier": 0.1672574382686718,
+        "logloss": 0.505198302627369
+      },
+      "nonPASS": {
+        "n": 5824,
+        "pos_rate": 0.4951923076923077,
+        "auroc": 0.821139738835895,
+        "auprc": 0.8249088115206255,
+        "brier": 0.17905030870563365,
+        "logloss": 0.5352053928461165
+      }
+    }
+  },
+  "importances": {
+    "impurity": [
+      {
+        "feature": "entropy_flank",
+        "impurity_importance": 0.28992320685730033
+      },
+      {
+        "feature": "motif_len",
+        "impurity_importance": 0.15078304844246473
+      },
+      {
+        "feature": "gc_flank",
+        "impurity_importance": 0.11765967510912077
+      },
+      {
+        "feature": "ref_tract_bp",
+        "impurity_importance": 0.09594543197447271
+      },
+      {
+        "feature": "allele_vs_readlen",
+        "impurity_importance": 0.06304989891121958
+      },
+      {
+        "feature": "ref_copynum",
+        "impurity_importance": 0.06281644250839796
+      },
+      {
+        "feature": "gt_repcn_max",
+        "impurity_importance": 0.045375808024477604
+      },
+      {
+        "feature": "gt_repcn_min",
+        "impurity_importance": 0.04503548319154128
+      },
+      {
+        "feature": "flanking_reads",
+        "impurity_importance": 0.04081082547154657
+      },
+      {
+        "feature": "spanning_frac",
+        "impurity_importance": 0.02788421749138721
+      },
+      {
+        "feature": "expansion_over_ref",
+        "impurity_importance": 0.017739812221077934
+      },
+      {
+        "feature": "locus_depth",
+        "impurity_importance": 0.014556405292958223
+      },
+      {
+        "feature": "spanning_reads",
+        "impurity_importance": 0.011672664495590936
+      },
+      {
+        "feature": "in_difficult",
+        "impurity_importance": 0.009656418449637608
+      },
+      {
+        "feature": "gt_hom",
+        "impurity_importance": 0.0024291103645865167
+      },
+      {
+        "feature": "in_segdup",
+        "impurity_importance": 0.001648983588740384
+      },
+      {
+        "feature": "flank_lowmap",
+        "impurity_importance": 0.001477948437034436
+      },
+      {
+        "feature": "repci_width_max",
+        "impurity_importance": 0.0012018133362063474
+      },
+      {
+        "feature": "inrepeat_reads",
+        "impurity_importance": 0.00033183288445321743
+      },
+      {
+        "feature": "is_pass",
+        "impurity_importance": 9.729477856164029e-07
+      },
+      {
+        "feature": "motif_is_homopolymer",
+        "impurity_importance": 0.0
+      }
+    ],
+    "permutation": [
+      {
+        "feature": "entropy_flank",
+        "perm_importance_mean": 0.13934060781658777,
+        "perm_importance_std": 0.0006361765279266924
+      },
+      {
+        "feature": "motif_len",
+        "perm_importance_mean": 0.1232472127797279,
+        "perm_importance_std": 0.0005893220011599711
+      },
+      {
+        "feature": "gc_flank",
+        "perm_importance_mean": 0.06320217026546789,
+        "perm_importance_std": 0.00039522027338993824
+      },
+      {
+        "feature": "ref_tract_bp",
+        "perm_importance_mean": 0.056776687651067095,
+        "perm_importance_std": 0.00015236878123781785
+      },
+      {
+        "feature": "ref_copynum",
+        "perm_importance_mean": 0.02267318905161917,
+        "perm_importance_std": 0.00014989102435837524
+      },
+      {
+        "feature": "allele_vs_readlen",
+        "perm_importance_mean": 0.020529595235711205,
+        "perm_importance_std": 0.00017190103816491447
+      },
+      {
+        "feature": "gt_repcn_min",
+        "perm_importance_mean": 0.01731383830567197,
+        "perm_importance_std": 0.000195043199990813
+      },
+      {
+        "feature": "gt_repcn_max",
+        "perm_importance_mean": 0.014405902490600276,
+        "perm_importance_std": 0.00013955774976049523
+      },
+      {
+        "feature": "expansion_over_ref",
+        "perm_importance_mean": 0.008579439049389648,
+        "perm_importance_std": 8.141211169349268e-05
+      },
+      {
+        "feature": "flanking_reads",
+        "perm_importance_mean": 0.005908979701386818,
+        "perm_importance_std": 8.933000723756271e-05
+      },
+      {
+        "feature": "spanning_frac",
+        "perm_importance_mean": 0.005236130437139996,
+        "perm_importance_std": 4.831785228506296e-05
+      },
+      {
+        "feature": "in_difficult",
+        "perm_importance_mean": 0.003852866555695589,
+        "perm_importance_std": 2.129084797378384e-05
+      },
+      {
+        "feature": "spanning_reads",
+        "perm_importance_mean": 0.0029217009056680563,
+        "perm_importance_std": 4.176582259464099e-05
+      },
+      {
+        "feature": "gt_hom",
+        "perm_importance_mean": 0.002172501389781667,
+        "perm_importance_std": 8.3379119655914e-06
+      },
+      {
+        "feature": "locus_depth",
+        "perm_importance_mean": 0.0020709165127682284,
+        "perm_importance_std": 2.549011860464095e-05
+      },
+      {
+        "feature": "in_segdup",
+        "perm_importance_mean": 0.0009386532858458585,
+        "perm_importance_std": 1.750671402431846e-05
+      },
+      {
+        "feature": "flank_lowmap",
+        "perm_importance_mean": 0.0005812032061902617,
+        "perm_importance_std": 1.3028115550094254e-05
+      },
+      {
+        "feature": "repci_width_max",
+        "perm_importance_mean": 0.00026026760399893155,
+        "perm_importance_std": 1.492427417015547e-05
+      },
+      {
+        "feature": "inrepeat_reads",
+        "perm_importance_mean": 5.1632608300478114e-05,
+        "perm_importance_std": 5.166444569830962e-06
+      },
+      {
+        "feature": "is_pass",
+        "perm_importance_mean": 5.758337677796987e-08,
+        "perm_importance_std": 3.3427425855204445e-08
+      },
+      {
+        "feature": "motif_is_homopolymer",
+        "perm_importance_mean": 0.0,
+        "perm_importance_std": 0.0
+      }
+    ]
+  }
+}

sv_model_v13_parents.joblib → sv_calibrator.joblib

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3a6ac169d5079ed6dbc5ece294c562ce9595b029147cee5debd48f27b53c2dfe
-size 1809789541
+oid sha256:307cbb00688a5a3b5aa5caac1700cdbd1c2fdd5b4a7e08072a121a4f7526e2e5
+size 7872

sv_config.json

@@ -1,62 +1,67 @@
 {
-  "model_file": "sv_model_v13_parents.joblib",
+  "release_version": "1.0",
+  "model_file": "sv_model.joblib",
+  "calibrator_file": "sv_calibrator.joblib",
+  "calibration": "isotonic regression on out-of-fold scores",
   "features": [
-    "qual",
-    "gq",
-    "pr_ref",
-    "pr_alt",
-    "sr_ref",
-    "sr_alt",
-    "vf_ref",
-    "vf_alt",
-    "total_alt_support",
-    "sr_pr_ratio",
-    "vaf_estimate",
-    "strand_bias_fs",
-    "cipos_width",
-    "ciend_width",
-    "homlen",
-    "svlen_abs",
+    "is_pass",
     "svtype_DEL",
-    "svtype_INS",
     "svtype_DUP",
+    "svtype_INS",
+    "svtype_INV",
     "svtype_BND",
+    "svlen_log",
+    "cipos_width",
+    "ciend_width",
     "is_imprecise",
-    "filter_pass",
-    "n_filter_flags"
+    "pe_support",
+    "sr_support",
+    "total_support",
+    "vaf",
+    "gt_hom",
+    "gq",
+    "qual_norm",
+    "local_depth",
+    "gc_min",
+    "gc_max",
+    "entropy_min",
+    "microhom_max",
+    "in_segdup_either",
+    "in_segdup_both",
+    "in_difficult_either",
+    "in_difficult_both",
+    "in_lowmap_either",
+    "in_tandem_either",
+    "in_Alu_either",
+    "in_L1_either",
+    "in_SVA_either",
+    "in_LTR_either",
+    "frac_span_repeat",
+    "n_neighbors",
+    "nn_log_dist"
   ],
-  "tier_thresholds": {
-    "low": 0.3,
-    "high": 0.7
+  "n_features": 35,
+  "missing_sentinel": -99999.0,
+  "tiers": {
+    "HIGH": "CS>=0.70",
+    "MODERATE": "0.50<=CS<0.70",
+    "WARNING": "0.30<=CS<0.50",
+    "LOW": "CS<0.30"
   },
-  "rules": {
-    "low_qual": {
-      "feature": "qual",
-      "op": "<",
-      "value": 20
-    },
-    "low_gq": {
-      "feature": "gq",
-      "op": "<",
-      "value": 15
-    },
-    "no_alt_support": {
-      "feature": "total_alt_support",
-      "op": "<=",
-      "value": 2
-    },
-    "low_vaf": {
-      "feature": "vaf_estimate",
-      "op": "<",
-      "value": 0.15
-    },
-    "imprecise": {
-      "feature": "is_imprecise",
-      "op": "==",
-      "value": 1
-    }
-  },
-  "sklearn_version_trained": "1.5.1",
+  "tier_edges": [
+    0.3,
+    0.5,
+    0.7
+  ],
+  "score": "CS = isotonic-calibrated P(call concordant with long-read truth)",
   "variant_class": "SV",
-  "override_target": "LOW"
+  "sklearn_version_trained": "1.7.1",
+  "training": {
+    "cohort": "HPRC",
+    "n_samples": 208,
+    "n_train_rows": 2575116,
+    "cv": "5-fold GroupKFold by sample",
+    "oof_auroc": 0.9502,
+    "oof_auprc": 0.9512
+  }
 }

sv_model.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cafd7dd091a0c1f9ceb2629c8bd6315e4967c37c4bf8910fcfea27454be64e7
+size 992442127

sv_model_meta.json

@@ -0,0 +1,583 @@
+{
+  "variant": "sv",
+  "feature_cols": [
+    "is_pass",
+    "svtype_DEL",
+    "svtype_DUP",
+    "svtype_INS",
+    "svtype_INV",
+    "svtype_BND",
+    "svlen_log",
+    "cipos_width",
+    "ciend_width",
+    "is_imprecise",
+    "pe_support",
+    "sr_support",
+    "total_support",
+    "vaf",
+    "gt_hom",
+    "gq",
+    "qual_norm",
+    "local_depth",
+    "gc_min",
+    "gc_max",
+    "entropy_min",
+    "microhom_max",
+    "in_segdup_either",
+    "in_segdup_both",
+    "in_difficult_either",
+    "in_difficult_both",
+    "in_lowmap_either",
+    "in_tandem_either",
+    "in_Alu_either",
+    "in_L1_either",
+    "in_SVA_either",
+    "in_LTR_either",
+    "frac_span_repeat",
+    "n_neighbors",
+    "nn_log_dist"
+  ],
+  "n_features": 35,
+  "tier_edges": [
+    0.3,
+    0.5,
+    0.7
+  ],
+  "tier_names": [
+    "LOW",
+    "Warning",
+    "Moderate",
+    "High"
+  ],
+  "missing_sentinel": -99999.0,
+  "rf_params": {
+    "bootstrap": true,
+    "ccp_alpha": 0.0,
+    "class_weight": "balanced_subsample",
+    "criterion": "gini",
+    "max_depth": null,
+    "max_features": "sqrt",
+    "max_leaf_nodes": null,
+    "max_samples": null,
+    "min_impurity_decrease": 0.0,
+    "min_samples_leaf": 20,
+    "min_samples_split": 2,
+    "min_weight_fraction_leaf": 0.0,
+    "monotonic_cst": null,
+    "n_estimators": 400,
+    "n_jobs": -1,
+    "oob_score": false,
+    "random_state": 42,
+    "verbose": 0,
+    "warm_start": false
+  },
+  "n_train_rows": 2575116,
+  "n_samples": 208,
+  "qc": {
+    "label_rows_raw": 2782190,
+    "label_dist_raw": {
+      "concordant": 1530286,
+      "discordant": 1251904
+    },
+    "label_rows_usable": 2782190,
+    "ambiguous_keys_dropped": 9462,
+    "ambiguous_feat_rows": 18568,
+    "ambiguous_label_rows": 19114,
+    "dup_keys_feature": 18568,
+    "dup_keys_label": 19114,
+    "merged_rows": 2575116,
+    "match_rate_vs_labels": 0.9255715820989939,
+    "match_rate_vs_features": 1.0,
+    "class_balance": {
+      "concordant": 1511906,
+      "discordant": 1063210
+    },
+    "concordant_rate": 0.5871215121959554
+  },
+  "importances": {
+    "impurity": [
+      {
+        "feature": "svlen_log",
+        "impurity_importance": 0.14231107100051552
+      },
+      {
+        "feature": "svtype_BND",
+        "impurity_importance": 0.13513441637259968
+      },
+      {
+        "feature": "nn_log_dist",
+        "impurity_importance": 0.06888043456708229
+      },
+      {
+        "feature": "svtype_DUP",
+        "impurity_importance": 0.05967552431932549
+      },
+      {
+        "feature": "cipos_width",
+        "impurity_importance": 0.05563805020798381
+      },
+      {
+        "feature": "svtype_DEL",
+        "impurity_importance": 0.053612638587210104
+      },
+      {
+        "feature": "sr_support",
+        "impurity_importance": 0.04950031891627467
+      },
+      {
+        "feature": "vaf",
+        "impurity_importance": 0.04822805852199238
+      },
+      {
+        "feature": "qual_norm",
+        "impurity_importance": 0.04321008187665855
+      },
+      {
+        "feature": "svtype_INS",
+        "impurity_importance": 0.03671741257598638
+      },
+      {
+        "feature": "ciend_width",
+        "impurity_importance": 0.027655467190250325
+      },
+      {
+        "feature": "local_depth",
+        "impurity_importance": 0.024667786835611386
+      },
+      {
+        "feature": "microhom_max",
+        "impurity_importance": 0.023198612318248754
+      },
+      {
+        "feature": "is_imprecise",
+        "impurity_importance": 0.02244493530457544
+      },
+      {
+        "feature": "frac_span_repeat",
+        "impurity_importance": 0.02223685870094091
+      },
+      {
+        "feature": "entropy_min",
+        "impurity_importance": 0.02149966456515826
+      },
+      {
+        "feature": "pe_support",
+        "impurity_importance": 0.018807543132767727
+      },
+      {
+        "feature": "gc_min",
+        "impurity_importance": 0.018609267758191137
+      },
+      {
+        "feature": "gq",
+        "impurity_importance": 0.017999043161167707
+      },
+      {
+        "feature": "gc_max",
+        "impurity_importance": 0.01691329606031783
+      },
+      {
+        "feature": "total_support",
+        "impurity_importance": 0.01639193545906872
+      },
+      {
+        "feature": "gt_hom",
+        "impurity_importance": 0.014227746565587479
+      },
+      {
+        "feature": "n_neighbors",
+        "impurity_importance": 0.013414404592739407
+      },
+      {
+        "feature": "in_difficult_both",
+        "impurity_importance": 0.01186932037949349
+      },
+      {
+        "feature": "is_pass",
+        "impurity_importance": 0.011331347970027637
+      },
+      {
+        "feature": "in_tandem_either",
+        "impurity_importance": 0.006678373764632643
+      },
+      {
+        "feature": "in_lowmap_either",
+        "impurity_importance": 0.004352768263433798
+      },
+      {
+        "feature": "in_Alu_either",
+        "impurity_importance": 0.004250882794571496
+      },
+      {
+        "feature": "in_difficult_either",
+        "impurity_importance": 0.004067397992242813
+      },
+      {
+        "feature": "in_segdup_either",
+        "impurity_importance": 0.001790844691729305
+      },
+      {
+        "feature": "in_segdup_both",
+        "impurity_importance": 0.001567017346771133
+      },
+      {
+        "feature": "in_L1_either",
+        "impurity_importance": 0.0014931685060102253
+      },
+      {
+        "feature": "in_LTR_either",
+        "impurity_importance": 0.001243669659616864
+      },
+      {
+        "feature": "in_SVA_either",
+        "impurity_importance": 0.00038064004121650836
+      },
+      {
+        "feature": "svtype_INV",
+        "impurity_importance": 0.0
+      }
+    ],
+    "permutation": [
+      {
+        "feature": "svlen_log",
+        "perm_importance_mean": 0.04194058803806115,
+        "perm_importance_std": 0.0007089186317830076
+      },
+      {
+        "feature": "nn_log_dist",
+        "perm_importance_mean": 0.019778079687185944,
+        "perm_importance_std": 0.0002553374546546104
+      },
+      {
+        "feature": "svtype_DEL",
+        "perm_importance_mean": 0.018689927317770305,
+        "perm_importance_std": 0.0002501511162311443
+      },
+      {
+        "feature": "cipos_width",
+        "perm_importance_mean": 0.017400205941047363,
+        "perm_importance_std": 0.00021672163272084185
+      },
+      {
+        "feature": "svtype_BND",
+        "perm_importance_mean": 0.015007739432103828,
+        "perm_importance_std": 0.00026380208237979455
+      },
+      {
+        "feature": "qual_norm",
+        "perm_importance_mean": 0.013299944084231186,
+        "perm_importance_std": 0.00013420815060181905
+      },
+      {
+        "feature": "gc_min",
+        "perm_importance_mean": 0.012263946411167393,
+        "perm_importance_std": 0.0001367518392425017
+      },
+      {
+        "feature": "entropy_min",
+        "perm_importance_mean": 0.01175411732205851,
+        "perm_importance_std": 0.00010827028744123835
+      },
+      {
+        "feature": "pe_support",
+        "perm_importance_mean": 0.011431666717043187,
+        "perm_importance_std": 0.0002522902431662052
+      },
+      {
+        "feature": "vaf",
+        "perm_importance_mean": 0.010811854209153338,
+        "perm_importance_std": 0.000206278576011294
+      },
+      {
+        "feature": "svtype_DUP",
+        "perm_importance_mean": 0.010256370264851777,
+        "perm_importance_std": 0.00011705190707458093
+      },
+      {
+        "feature": "frac_span_repeat",
+        "perm_importance_mean": 0.009963778341341434,
+        "perm_importance_std": 0.0003054705160430916
+      },
+      {
+        "feature": "local_depth",
+        "perm_importance_mean": 0.009591312749510661,
+        "perm_importance_std": 9.987579143240777e-05
+      },
+      {
+        "feature": "gc_max",
+        "perm_importance_mean": 0.009472834004097907,
+        "perm_importance_std": 3.73573580102502e-05
+      },
+      {
+        "feature": "gq",
+        "perm_importance_mean": 0.009203283859545164,
+        "perm_importance_std": 0.00010130110270414088
+      },
+      {
+        "feature": "sr_support",
+        "perm_importance_mean": 0.008346937601715121,
+        "perm_importance_std": 9.766321902716232e-05
+      },
+      {
+        "feature": "microhom_max",
+        "perm_importance_mean": 0.0074945231035745685,
+        "perm_importance_std": 3.858741211472431e-05
+      },
+      {
+        "feature": "svtype_INS",
+        "perm_importance_mean": 0.007191278115888,
+        "perm_importance_std": 6.25382140306982e-05
+      },
+      {
+        "feature": "total_support",
+        "perm_importance_mean": 0.0071782291459881135,
+        "perm_importance_std": 9.233685802608801e-05
+      },
+      {
+        "feature": "is_pass",
+        "perm_importance_mean": 0.006096510294509483,
+        "perm_importance_std": 0.00018164506103297094
+      },
+      {
+        "feature": "n_neighbors",
+        "perm_importance_mean": 0.005750802897597151,
+        "perm_importance_std": 5.2407508065749236e-05
+      },
+      {
+        "feature": "in_difficult_both",
+        "perm_importance_mean": 0.005015233708107925,
+        "perm_importance_std": 0.00018000694143725676
+      },
+      {
+        "feature": "ciend_width",
+        "perm_importance_mean": 0.004891217221742616,
+        "perm_importance_std": 8.693371224059806e-05
+      },
+      {
+        "feature": "in_tandem_either",
+        "perm_importance_mean": 0.0043522978742952965,
+        "perm_importance_std": 0.00013096877220331791
+      },
+      {
+        "feature": "gt_hom",
+        "perm_importance_mean": 0.00323902471619224,
+        "perm_importance_std": 6.580316549205161e-05
+      },
+      {
+        "feature": "in_lowmap_either",
+        "perm_importance_mean": 0.002848785493209416,
+        "perm_importance_std": 6.035487488575199e-05
+      },
+      {
+        "feature": "in_Alu_either",
+        "perm_importance_mean": 0.002534492148327061,
+        "perm_importance_std": 8.942851729207139e-05
+      },
+      {
+        "feature": "in_difficult_either",
+        "perm_importance_mean": 0.002091988241603948,
+        "perm_importance_std": 5.002148003217409e-05
+      },
+      {
+        "feature": "is_imprecise",
+        "perm_importance_mean": 0.001979962861476592,
+        "perm_importance_std": 6.0060264184685146e-05
+      },
+      {
+        "feature": "in_L1_either",
+        "perm_importance_mean": 0.0011150058316057754,
+        "perm_importance_std": 2.43935928349277e-05
+      },
+      {
+        "feature": "in_LTR_either",
+        "perm_importance_mean": 0.0006375153501665843,
+        "perm_importance_std": 3.5907563733425047e-05
+      },
+      {
+        "feature": "in_segdup_either",
+        "perm_importance_mean": 0.0006168866779678206,
+        "perm_importance_std": 2.936960388188349e-05
+      },
+      {
+        "feature": "in_segdup_both",
+        "perm_importance_mean": 0.0005383371585652164,
+        "perm_importance_std": 3.300168720103858e-05
+      },
+      {
+        "feature": "in_SVA_either",
+        "perm_importance_mean": 0.00017570394306039018,
+        "perm_importance_std": 3.211137729753612e-06
+      },
+      {
+        "feature": "svtype_INV",
+        "perm_importance_mean": 0.0,
+        "perm_importance_std": 0.0
+      }
+    ]
+  },
+  "finalized_unix": 1782044045,
+  "cv_report": {
+    "overall": {
+      "n": 2575116,
+      "pos_rate": 0.5871215121959554,
+      "auroc": 0.9501778843150939,
+      "auprc": 0.9511739760033938,
+      "brier": 0.07706030782330338,
+      "logloss": 0.2628003224452437
+    },
+    "calibration": [
+      {
+        "bin": "[0.0,0.1)",
+        "n": 701606,
+        "mean_pred": 0.018278732155845624,
+        "obs_rate": 0.008842569761376044
+      },
+      {
+        "bin": "[0.1,0.2)",
+        "n": 77383,
+        "mean_pred": 0.1443051591972469,
+        "obs_rate": 0.10029334608376517
+      },
+      {
+        "bin": "[0.2,0.3)",
+        "n": 53690,
+        "mean_pred": 0.24934861469329547,
+        "obs_rate": 0.19802570311044887
+      },
+      {
+        "bin": "[0.3,0.4)",
+        "n": 57370,
+        "mean_pred": 0.35138865912279116,
+        "obs_rate": 0.31784905002614605
+      },
+      {
+        "bin": "[0.4,0.5)",
+        "n": 70890,
+        "mean_pred": 0.4521905700361762,
+        "obs_rate": 0.45171392297926366
+      },
+      {
+        "bin": "[0.5,0.6)",
+        "n": 98358,
+        "mean_pred": 0.5530042682672152,
+        "obs_rate": 0.5941153744484434
+      },
+      {
+        "bin": "[0.6,0.7)",
+        "n": 158080,
+        "mean_pred": 0.6545151929144646,
+        "obs_rate": 0.7394863360323887
+      },
+      {
+        "bin": "[0.7,0.8)",
+        "n": 264297,
+        "mean_pred": 0.7534285409368959,
+        "obs_rate": 0.8542737904705691
+      },
+      {
+        "bin": "[0.8,0.9)",
+        "n": 407540,
+        "mean_pred": 0.8554391726312989,
+        "obs_rate": 0.922547480001963
+      },
+      {
+        "bin": "[0.9,1.0)",
+        "n": 685902,
+        "mean_pred": 0.9410329493878745,
+        "obs_rate": 0.962179728299378
+      }
+    ],
+    "per_sample_auroc": {
+      "n_samples": 208,
+      "median": 0.9812468623755227,
+      "p25": 0.9541872315128539,
+      "p75": 0.9849148282479339,
+      "min": 0.8241255348590714,
+      "max": 0.9884727126524584
+    },
+    "by_svtype": {
+      "BND": {
+        "n": 545562,
+        "pos_rate": 0.05967241120165994,
+        "auroc": 0.9578724482896165,
+        "auprc": 0.7205291737443336,
+        "brier": 0.03197870532515856,
+        "logloss": 0.11268068083789429
+      },
+      "DEL": {
+        "n": 1168706,
+        "pos_rate": 0.7785747655954535,
+        "auroc": 0.8919672267722084,
+        "auprc": 0.9557196435784474,
+        "brier": 0.0922761808544916,
+        "logloss": 0.3149783791403593
+      },
+      "DUP": {
+        "n": 156534,
+        "pos_rate": 0.023419832113151136,
+        "auroc": 0.9936692829891886,
+        "auprc": 0.8545445334945798,
+        "brier": 0.009443816981288416,
+        "logloss": 0.03587112293286734
+      },
+      "INS": {
+        "n": 704314,
+        "pos_rate": 0.8032780833548673,
+        "auroc": 0.8549802852596324,
+        "auprc": 0.9494948614191011,
+        "brier": 0.1017598124373945,
+        "logloss": 0.3429363119373415
+      }
+    },
+    "by_size": {
+      "1-10kb": {
+        "n": 171605,
+        "pos_rate": 0.7607121004632732,
+        "auroc": 0.9296938425020418,
+        "auprc": 0.967200346507,
+        "brier": 0.06806650111282099,
+        "logloss": 0.2455626050274475
+      },
+      "10-100kb": {
+        "n": 42645,
+        "pos_rate": 0.21177160276703014,
+        "auroc": 0.9920116885561145,
+        "auprc": 0.9676826349150152,
+        "brier": 0.029729192191828648,
+        "logloss": 0.10986227030572825
+      },
+      "100bp-1kb": {
+        "n": 906987,
+        "pos_rate": 0.7867466678133204,
+        "auroc": 0.8976865414893258,
+        "auprc": 0.9590485952985468,
+        "brier": 0.08302330275831495,
+        "logloss": 0.28853539649876037
+      },
+      "<100bp": {
+        "n": 733616,
+        "pos_rate": 0.7488359032518375,
+        "auroc": 0.867906684781433,
+        "auprc": 0.93767984701013,
+        "brier": 0.11261501240225187,
+        "logloss": 0.3687373180284347
+      },
+      ">100kb": {
+        "n": 54796,
+        "pos_rate": 0.00839477334111979,
+        "auroc": 0.9897200830900804,
+        "auprc": 0.788119068363777,
+        "brier": 0.004907638846204342,
+        "logloss": 0.029204468344063542
+      },
+      "NA": {
+        "n": 665467,
+        "pos_rate": 0.16371360262792894,
+        "auroc": 0.976441017719166,
+        "auprc": 0.9070303608956529,
+        "brier": 0.04103092730468184,
+        "logloss": 0.1444199784010817
+      }
+    }
+  }
+}

tier_thresholds.json

@@ -1,11 +1,12 @@
 {
+  "release_version": "1.0",
   "tiers": {
-    "HIGH": "CS >= 0.70",
-    "MODERATE": "0.50 <= CS < 0.70",
-    "WARNING": "0.30 <= CS < 0.50",
-    "LOW": "CS < 0.30"
+    "HIGH": "CS>=0.70",
+    "MODERATE": "0.50<=CS<0.70",
+    "WARNING": "0.30<=CS<0.50",
+    "LOW": "CS<0.30"
   },
-  "override_sv": "any of {QUAL<20, GQ<15, total_alt_support<=2, vaf_estimate<0.15, is_imprecise==1} -> LOW (see sv_config.json 'rules')",
-  "override_str": "any of {locus_coverage<20, is_low_depth==1, support_type_a2<=1, total_support_a2<5, ci_width_a2>3, allele_balance<0.3} -> WARNING (see str_config.json 'rules')",
-  "note": "Only MODERATE-to-HIGH is a monotone precision ladder; LOW/WARNING are not rank-ordered by precision. Use HIGH as a candidate-triage filter."
+  "score": "CS = isotonic-calibrated probability of concordance with long-read truth",
+  "calibration": "isotonic on OOF; SV Brier 0.0771->0.0744, STR 0.1673->0.1589",
+  "note": "Tiers are buckets of the calibrated CS. HIGH is the candidate-triage tier."
 }