Fixtures and derived results for migec

ci/SRR1763769_umi0.5pct.fq.gz is all the reads of 0.5% of the barcodes of a
public HIV-1 Primer ID library -- never 0.5% of the reads, which at 16 reads per
molecule would give molecules seen once each and destroy the size distribution
the fixture exists to preserve. 15.77 reads/barcode against the library's 16.05.

results/ holds the derived tables behind the numbers the documentation quotes,
so a claim can be checked without re-running anything.

Raw ENA runs and simulated corpora are not here: SOURCES.md carries the command
that regenerates each instead of the bytes. Nothing from a private cluster.

.gitattributes

@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+*.fq.gz filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,60 @@
 ---
-license: cc-by-nc-nd-4.0
+license: gpl-3.0
+tags:
+  - umi
+  - molecular-barcodes
+  - sequencing
+  - immunogenomics
+pretty_name: UMI benchmark data for migec
 ---
+
+# umi_data
+
+Fixtures and derived results for [**migec**](https://github.com/antigenomics/migec) — UMI barcode
+extraction, correction and consensus assembly.
+
+This dataset holds **two kinds of thing and no others**:
+
+- `ci/` — small fixtures, cut from public data, that the test suite and the documentation examples
+  run against.
+- `results/` — the derived tables behind the numbers migec's documentation quotes, so that a claim
+  can be checked without re-running anything.
+
+## What is deliberately not here
+
+- **Raw reads that are one command away.** `SRR1763769` is 248 MB on ENA and a `curl` recovers it
+  exactly; `SOURCES.md` carries the command instead of the bytes.
+- **Simulated corpora.** Regenerable from a seed, and the seed is in `SOURCES.md`.
+- **Anything under access control.** Nothing in this dataset came off a private cluster.
+
+## `ci/SRR1763769_umi0.5pct.fq.gz`
+
+2.12 M reads of an HIV-1 Primer ID library, checked out on a 9 nt Primer ID, then cut down to
+**all of the reads of 0.5% of the barcodes** — 9,824 reads over 623 barcodes.
+
+⛔ Not a 0.5% sample of the reads. At 16 reads per molecule, sampling reads gives molecules seen
+once each: the MIG size distribution is destroyed and every consensus is a single read, so the
+fixture silently stops testing the thing it was built for. The barcodes are selected by hashing
+(`splitmix64` of the packed barcode, kept when `hash % 10000 < 50`), which is unbiased, nested —
+a smaller fixture is a subset of a larger one — and reproducible from the definition in
+`include/migec/subsample.hpp`.
+
+The fixture keeps **15.77 reads per barcode** against the full library's 16.05.
+
+```bash
+migec refine  ci/SRR1763769_umi0.5pct.fq.gz -o ref/
+migec assemble ref/CTRL.fq.gz -o cons/
+```
+
+## `results/`
+
+| file | what it backs |
+|---|---|
+| `nulls.json`, `nulls.linkage.tsv` | X3, the three permutation nulls |
+| `nulls_position_pairs.txt` | the pairwise position-dependence map — adjacent pairs only |
+| `nulls_linkage_threshold.txt` | the MIG-split threshold, 8.68 with a bootstrap 95% CI [8.42, 9.14] |
+| `correction_accuracy.tsv` | UMI correction recall/precision against depth |
+| `suggest.cycles.tsv`, `suggest.segments.tsv` | the per-cycle PWM that recovers the barcode layout |
+| `fixture_rank.tsv`, `fixture_bins.tsv` | the barcode-rank curve and per-bin error fraction of the fixture |
+
+Provenance, and the exact command that regenerates each, is in `SOURCES.md`.

SOURCES.md

@@ -0,0 +1,161 @@
+# SOURCES — antigenomics/migec
+
+Every dataset this repo ships, consumes or benchmarks against, where it came from, and how to
+regenerate it.
+
+**Experimental** = measured/sequenced. **Derived** = computed by us from something else. The two
+are never conflated in a table row.
+
+## Shipped in this repo
+
+| Artifact | Origin | Provenance | Regenerate |
+|---|---|---|---|
+| `tests/synthetic/_sim.py` output | none — generated | derived | `SimConfig(seed=...)`; every truth file is a pure function of the seed |
+| `tests/cpp/doctest.h` | [doctest](https://github.com/doctest/doctest) 2.4.11, MIT | vendored | copy from upstream release |
+
+Nothing else is committed. Test corpora live on HuggingFace (below) rather than in git.
+
+## Benchmark data
+
+Nothing here has been fetched yet — this table is the plan of record, and rows gain a "fetched"
+date as they land.
+
+### MIGEC (Shugay et al., Nat Methods 2014, doi:10.1038/nmeth.2960)
+
+| Item | Value |
+|---|---|
+| Public spike-ins | BioProject `PRJNA239303`, runs `SRR1200517`–`SRR1200520`, AMPLICON paired |
+| Ground truth | Supplementary Table 1a: 12 control clonotypes (5 TRA, 5 TRB, 2 IGH) with expected frequencies |
+| Truth source | `41592_2014_BFnmeth2960_MOESM376_ESM.pdf`, extractable with `pdftotext -layout` |
+| Provenance | experimental |
+| Chemistry | MiSeq 2×150; SMART adapter `GTGGTATCAACGCAGAG` |
+
+| Item | Value |
+|---|---|
+| Experiment 1 (internal) | `/projects/cdr3_ngs/2012/12_alvaro_ab_bcr_nnn/` (IGH) and `/projects/cdr3_ngs/2012/08_alvaro_nnnb/` (TCR) on aldan3 |
+| Staged copies | `/projects/tcr_bcr_rnaseq/data_migec_exp1/{IGH_P41,TCR_Project25}_R{1,2}.fastq.gz` |
+| Access | `aldan3 ls`, `aldan3 pull` (see `~/vcs/code/aldan3-client`) |
+| Provenance | experimental, unpublished |
+
+⛔ **Experiment 1 raw reads must not leave the cluster.** Only derived summaries (histograms,
+error-rate tables, consensus statistics) may be published or uploaded to HuggingFace.
+
+⛔ `scratch/spikein/S1_R2_2M.fq` on aldan3 is **corrupt past record 1,742,617**. Do not use it.
+
+### MAGERI (Shugay et al., PLoS Comput Biol 2017, doi:10.1371/journal.pcbi.1005480, PMID 28475621)
+
+| Item | Value |
+|---|---|
+| Error-model datasets | SRA `PRJNA352143` — UMI-tagged sequencing of a known template with 9 polymerases |
+| Duplex sequencing | SRA `SRR1799908`; primer patterns `NNNNNNNNNNNNtgact` / `agtcaNNNNNNNNNNNN` |
+| HIV protease amplicons | SRA `SRP052322`; patterns `NNNNNNNNNcagtttaacttttgggccatccattcc` / `ctatcggctcctgnnnn` |
+| Companion repo | https://github.com/mikessh/mageri-paper (error model PDFs, analysis scripts) |
+| Provenance | experimental |
+
+The patterns above are quoted verbatim from the paper's Methods and are directly reusable as
+`checkout` test cases.
+
+Reference values worth keeping (MAGERI Methods, for comparison rather than reimplementation):
+UMIs below Phred 20 discarded; MIG pairs differing by 1 or 2 substitutions with size ratios above
+20× and 400× treated as error children; MIG size threshold at the square root of the distribution
+peak; consensus core 30 bases with ±5 offset; `CQS = (40/3)·(4f − 1)`; per-substitution error
+rates fitted as Beta, counts as Beta-Binomial, `Q = −10 log10 P`, capped at 100.
+
+### 10x Genomics
+
+| Item | Value |
+|---|---|
+| Datasets | four "Connect-generated GEX+VDJ" sets: human PBMC, mouse PBMC, human melanoma, mouse splenocytes (Cell Ranger 6.0.1) |
+| Whitelists | barcode whitelists shipped with Cell Ranger (`10XGenomics/cellranger`) |
+| Reference calls | the published `filtered_feature_bc_matrix` — used as a comparator, not re-run |
+| Provenance | experimental (10x), reference calls derived |
+
+**X1 (read-start dispersion) used `pbmc_1k_v3`**, Cell Ranger 3.0.0, GRCh38-3.0.0:
+
+| Item | Value |
+|---|---|
+| BAM | `https://cf.10xgenomics.com/samples/cell-exp/3.0.0/pbmc_1k_v3/pbmc_1k_v3_possorted_genome_bam.bam` (4.79 GB) + `.bai` (4.6 MB) |
+| Fetch | not fetched — the server serves HTTP range requests (verified 206, 2026-08-13), so `pysam.AlignmentFile(url)` reads only the regions asked for |
+| Regions | `11:65497688-65508073` (MALAT1), `7:5527151-5530601` (ACTB), `15:44711477-44718877` (B2M) — chosen for expression, since an unexpressed locus has no UMI with more than one read |
+| Contig names | plain `1`, `2`, … `X` — **not** `chr1`. Cell Ranger's GRCh38-3.0.0 reference is Ensembl-styled |
+| Regenerate | `python scripts/read_start_dispersion.py --bam <url> --region … ` |
+| Provenance | experimental (10x); the dispersion statistics are derived |
+
+### HIV-1 Primer ID — the X2 quality-floor control
+
+| Item | Value |
+|---|---|
+| Run | `SRR1763769`, 2,122,456 read pairs, study `PRJNA272736` |
+| Fetch | `curl -O ftp://ftp.sra.ebi.ac.uk/vol1/fastq/SRR176/009/SRR1763769/SRR1763769_2.fastq.gz` (248 MB; R1 is not needed — the Primer ID is on the cDNA primer in **R2**) |
+| Layout | 9 nt Primer ID, then `CAGTTTAACTTTTGGGCCAT`; recovered from the data by per-cycle entropy, not from the protocol |
+| Paper | Zhou, Jones, Mieczkowski & Swanstrom, *J Virol* 89(16):8540–8555, 2015, [doi:10.1128/JVI.00522-15](https://doi.org/10.1128/JVI.00522-15) — reports a residual error rate of ~1 in 10,000 |
+| Regenerate | `python scripts/quality_floor.py --reads SRR1763769_2.fastq.gz --out x2/ --window 180` (X2, the quality floor) and `python scripts/permutation_nulls.py --reads SRR1763769_2.fastq.gz --out x3/ --cycles 32 --window 180` (X3, the three permutation nulls) |
+| Provenance | experimental (ENA); the floor and its interval are derived |
+| ⚠ | ENA's metadata gives every run in the study the same title, so it does **not** identify which runs are controls. This is HIV plasma — a quasispecies — so the estimator restricts to monomorphic positions rather than assuming clonality. The library is also 49.6% occupied on its 9 nt barcode, which `checkout` flags as saturated; the measured floor is an upper bound. |
+
+### Calib (github.com/vpc-ccg/calib)
+
+| Item | Value |
+|---|---|
+| Use | comparator for UMI grouping accuracy — it clusters on barcode *and* sequence, we (today) on barcode alone |
+| Get it | `git clone https://github.com/vpc-ccg/calib && cd calib && make` → `calib`, `calib_cons` |
+| Run | `calib -f R1.fq -r R2.fq -l <barcode_len> -o prefix` → `prefix.cluster` |
+| `.cluster` format | 9 TSV columns: `cluster_id, node_id, read_id, f_name, f_seq, f_qual, r_name, r_seq, r_qual` (verified against the upstream README, 2026-08-13) |
+| Truth used here | **our** simulator, `tests/synthetic/_sim.py`, which writes `truth_reads.tsv` (`read_id`, `molecule_id`). Calib's own simulator emits no read→molecule map |
+| Compared by | `scripts/compare_calib.py` — adjusted Rand index, plus split and merge fractions separately |
+| Storage | ⛔ do not store simulated reads — record the exact command and seed here instead |
+| Provenance | derived (simulated) |
+
+## HuggingFace — `isalgo/umi_data`
+
+Does not exist yet. When created: a git + git-lfs mirror at `~/hf/umi_data`, written by
+committing and pushing **in the mirror**, one commit per change set. Never through the HTTP API —
+that writes remotely only, leaves the mirror silently stale, and lands one commit per call.
+
+Planned layout:
+
+```
+umi_data/
+  ci/            small slices for CI, subsampled by WHOLE UMIs (all reads of N UMIs)
+  truth/         ground-truth tables (spike-in clonotypes, simulated molecule tables)
+  whitelists/    barcode whitelists + a README recording their upstream and license
+  SOURCES.md
+```
+
+⛔ Not in this dataset: aldan3 Experiment 1 raw reads, and anything regenerable by a one-line
+command (record the command here instead of storing gigabytes in LFS).
+
+---
+
+## What this HuggingFace dataset contains
+
+Appended when `isalgo/umi_data` was first populated (2026-08-13). The sections above describe every
+dataset migec uses; this section says which of them are shipped here and how each file was made.
+
+### `ci/SRR1763769_umi0.5pct.fq.gz`
+
+| Item | Value |
+|---|---|
+| Origin | derived from `SRR1763769` (ENA, PRJNA272736) — see the section above for the fetch command |
+| Made by | `migec suggest` → `migec checkout` → `migec subsample --keep 0.5` |
+| Regenerate | `migec suggest SRR1763769_2.fastq.gz` gives `NNNNNNNNNcagtttaacttttgggccatcca`; `migec checkout SRR1763769_2.fastq.gz -b barcodes.txt -o co/`; `migec subsample co/CTRL.fq.gz -o SRR1763769_umi0.5pct.fq.gz --keep 0.5` |
+| Selection | `splitmix64(packed barcode) % 10000 < 50` — deterministic, unbiased, nested |
+| Provenance | experimental (ENA), subset by whole barcodes |
+| ⚠ | 9,824 reads over 623 barcodes, 15.77 reads/barcode against the full library's 16.05 |
+
+### `results/`
+
+| Item | Value |
+|---|---|
+| Provenance | **derived** — every file is computed, none is measured |
+| `nulls*` | `python scripts/permutation_nulls.py --reads SRR1763769_2.fastq.gz --out x3/ --cycles 32 --window 180` |
+| `correction_accuracy.tsv` | `python scripts/correction_accuracy.py` (simulated, seed 5) |
+| `suggest.*` | `migec suggest SRR1763769_2.fastq.gz --cycles 32` |
+| `fixture_*` | `migec refine ci/SRR1763769_umi0.5pct.fq.gz -o ref/` |
+
+### ⛔ Not shipped here
+
+- **aldan3 Experiment 1 raw reads.** They do not leave the cluster; only derived summaries may.
+- **Britanova et al aging (bulk TCR, shallow).** On aldan3, not yet pulled.
+- **Raw ENA runs.** One `curl` away, so the command is shipped and the bytes are not.
+- **Simulated corpora.** Regenerable from the seed recorded above.

ci/SRR1763769_barcodes.txt

@@ -0,0 +1 @@
+CTRL	NNNNNNNNNcagtttaacttttgggccatcca

ci/SRR1763769_umi0.5pct.fq.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b8b0a3d38ae805cacc730e73a60bea6c495cff5ca9e2deb1cec054f05bc08b0
+size 1065751

results/correction_accuracy.tsv

@@ -0,0 +1,8 @@
+coverage	reads_per_umi	children	merged	recall	precision	molecules_kept	epsilon
+1.3	1.1056	820	173	0.1098	0.5202	0.9958	6.056515e-04
+2.0	1.5140	1092	671	0.4203	0.6841	0.9893	1.729467e-03
+3.0	2.3235	1768	1572	0.7376	0.8295	0.9868	2.588426e-03
+4.0	3.1167	2536	2443	0.8916	0.9255	0.9914	2.943061e-03
+6.0	4.6152	3923	3855	0.9674	0.9844	0.9973	2.870565e-03
+10.0	7.1180	6700	6575	0.9782	0.9968	0.9993	2.877537e-03
+25.0	13.2987	17016	16745	0.9831	0.9990	0.9997	2.872429e-03

results/fixture_bins.tsv

@@ -0,0 +1,10 @@
+min_reads	max_reads	barcodes	reads	merged	fraction_erroneous	payload_entropy_bits
+1	1	380	380	3	0.007895	0.1950
+2	3	143	323	1	0.006993	0.2052
+4	7	18	86	0	0.000000	0.0290
+8	15	7	75	0	0.000000	0.0185
+16	31	7	163	0	0.000000	0.0000
+32	63	6	289	0	0.000000	0.0203
+64	127	32	3171	0	0.000000	0.0815
+128	255	29	5059	0	0.000000	0.0068
+256	511	1	278	0	0.000000	0.0000

results/fixture_rank.tsv

@@ -0,0 +1,103 @@
+rank	reads	cumulative_reads	cumulative_fraction
+1	278	278	0.028298
+2	242	520	0.052932
+3	232	752	0.076547
+4	230	982	0.099959
+5	225	1207	0.122862
+6	223	1430	0.145562
+7	213	1643	0.167243
+8	203	1846	0.187907
+9	199	2045	0.208164
+10	189	2234	0.227402
+11	189	2423	0.246641
+12	187	2610	0.265676
+13	178	2788	0.283795
+14	177	2965	0.301812
+15	163	3128	0.318404
+16	161	3289	0.334792
+17	159	3448	0.350977
+18	159	3607	0.367162
+19	154	3761	0.382838
+20	151	3912	0.398208
+21	151	4063	0.413579
+22	148	4211	0.428644
+23	145	4356	0.443404
+24	145	4501	0.458164
+25	145	4646	0.472923
+26	143	4789	0.487480
+27	141	4930	0.501832
+28	138	5068	0.515879
+29	137	5205	0.529825
+30	133	5338	0.543363
+31	127	5465	0.556291
+32	126	5591	0.569116
+33	125	5716	0.581840
+34	125	5841	0.594564
+35	124	5965	0.607186
+36	120	6085	0.619401
+37	119	6204	0.631515
+38	118	6322	0.643526
+39	117	6439	0.655436
+40	114	6553	0.667040
+41	113	6666	0.678542
+43	106	6882	0.700529
+45	105	7092	0.721906
+47	103	7299	0.742976
+49	95	7494	0.762826
+51	91	7678	0.781555
+53	89	7857	0.799776
+55	80	8023	0.816673
+57	72	8168	0.831433
+59	71	8311	0.845989
+61	66	8445	0.859629
+64	59	8630	0.878461
+67	34	8765	0.892203
+70	27	8853	0.901160
+73	22	8925	0.908489
+76	15	8977	0.913783
+79	11	9013	0.917447
+82	8	9039	0.920094
+86	6	9065	0.922740
+90	5	9085	0.924776
+94	4	9102	0.926507
+98	4	9118	0.928135
+102	3	9132	0.929560
+107	3	9147	0.931087
+112	3	9162	0.932614
+117	3	9177	0.934141
+122	3	9192	0.935668
+128	3	9210	0.937500
+134	3	9228	0.939332
+140	2	9243	0.940859
+146	2	9255	0.942081
+153	2	9269	0.943506
+160	2	9283	0.944931
+167	2	9297	0.946356
+175	2	9313	0.947985
+183	2	9329	0.949613
+192	2	9347	0.951445
+201	2	9365	0.953278
+211	2	9385	0.955314
+221	2	9405	0.957349
+232	2	9427	0.959589
+243	1	9448	0.961726
+255	1	9460	0.962948
+267	1	9472	0.964169
+280	1	9485	0.965493
+293	1	9498	0.966816
+307	1	9512	0.968241
+322	1	9527	0.969768
+338	1	9543	0.971397
+354	1	9559	0.973025
+371	1	9576	0.974756
+389	1	9594	0.976588
+408	1	9613	0.978522
+428	1	9633	0.980558
+449	1	9654	0.982695
+471	1	9676	0.984935
+494	1	9699	0.987276
+518	1	9723	0.989719
+543	1	9748	0.992264
+570	1	9775	0.995012
+598	1	9803	0.997862
+619	1	9824	1.000000

results/nulls.json

@@ -0,0 +1,166 @@
+{
+  "pattern": "NNNNNNNNNcagtttaacttttgggccatcca",
+  "umi_length": 9,
+  "distinct_umis": 124562,
+  "independence": {
+    "marginal_collision": [
+      0.25127042763836727,
+      0.2511799694569492,
+      0.2510680325893868,
+      0.250685142818615,
+      0.25119883257646525,
+      0.25152659822676793,
+      0.2512721571271776,
+      0.2512702648341363,
+      0.25150845180679615
+    ],
+    "rows": [
+      {
+        "k": 1,
+        "windows": 9,
+        "log10_excess": 0.0
+      },
+      {
+        "k": 2,
+        "windows": 8,
+        "log10_excess": 0.00014325516660969345
+      },
+      {
+        "k": 3,
+        "windows": 7,
+        "log10_excess": 0.00027521939620349
+      },
+      {
+        "k": 4,
+        "windows": 6,
+        "log10_excess": 0.0002586863203026871
+      },
+      {
+        "k": 5,
+        "windows": 5,
+        "log10_excess": -0.0004941386106713668
+      }
+    ],
+    "log10_excess_per_position": 6.0889746369456446e-05,
+    "predicted_excess_full_length": 1.0011222598412741,
+    "p_coll_independent": 3.985529065688231e-06,
+    "p_coll_dependent": 3.990001864904884e-06
+  },
+  "jsd": [
+    {
+      "k": 2,
+      "windows": 8,
+      "jsd_observed": 8.336592740810704e-05,
+      "jsd_null_mean": 9.603944583045343e-05,
+      "jsd_null_sd": 2.009250742502297e-05,
+      "excess": -1.2673518422346385e-05,
+      "z": -0.6307584292124205
+    },
+    {
+      "k": 3,
+      "windows": 7,
+      "jsd_observed": 0.000211494001152272,
+      "jsd_null_mean": 0.00018819096572001093,
+      "jsd_null_sd": 2.569862458976419e-05,
+      "excess": 2.3303035432261065e-05,
+      "z": 0.9067814252418283
+    },
+    {
+      "k": 4,
+      "windows": 6,
+      "jsd_observed": 0.00048692927616137533,
+      "jsd_null_mean": 0.00044161520624727194,
+      "jsd_null_sd": 3.4547138974011904e-05,
+      "excess": 4.531406991410339e-05,
+      "z": 1.311659120258581
+    },
+    {
+      "k": 5,
+      "windows": 5,
+      "jsd_observed": 0.0013089503289189577,
+      "jsd_null_mean": 0.0013597114810166623,
+      "jsd_null_sd": 7.53217397502907e-05,
+      "excess": -5.076115209770452e-05,
+      "z": -0.6739243180785479
+    }
+  ],
+  "graph": {
+    "pairs_observed": 844243,
+    "pairs_by_chance": 817358.2165055806,
+    "pairs_excess": 26884.78349441942,
+    "epsilon_permutation": 0.0014428240591377324,
+    "epsilon_all_pairs": 0.08102151371605056,
+    "count_ratio": [
+      {
+        "ratio": 2,
+        "observed": 521137,
+        "null_mean": 511118.2,
+        "null_sd": 413.0281754120541,
+        "excess": 10018.799999999988,
+        "z": 24.256940800720958
+      },
+      {
+        "ratio": 5,
+        "observed": 242467,
+        "null_mean": 224578.5,
+        "null_sd": 334.65182220520353,
+        "excess": 17888.5,
+        "z": 53.45406423345586
+      },
+      {
+        "ratio": 10,
+        "observed": 205158,
+        "null_mean": 186908.1,
+        "null_sd": 464.13529107476046,
+        "excess": 18249.899999999994,
+        "z": 39.320216219154936
+      },
+      {
+        "ratio": 20,
+        "observed": 182195,
+        "null_mean": 164592.35,
+        "null_sd": 348.7176923348863,
+        "excess": 17602.649999999994,
+        "z": 50.47822461240518
+      },
+      {
+        "ratio": 50,
+        "observed": 139949,
+        "null_mean": 127536.75,
+        "null_sd": 321.3884147848126,
+        "excess": 12412.25,
+        "z": 38.62071384343674
+      }
+    ]
+  },
+  "linkage": {
+    "migs": 1642,
+    "randomisations": 8210,
+    "thresholds": [
+      {
+        "target_fp": 0.05,
+        "threshold": 4.846130887164428,
+        "called": 123,
+        "fraction": 0.07490864799025579
+      },
+      {
+        "target_fp": 0.01,
+        "threshold": 11.656788052112361,
+        "called": 21,
+        "fraction": 0.012789281364190013
+      },
+      {
+        "target_fp": 0.001,
+        "threshold": 21.10364618458584,
+        "called": 3,
+        "fraction": 0.0018270401948842874
+      }
+    ],
+    "null_quantiles": {
+      "0.5": 0.7066568330023071,
+      "0.9": 3.44729372716644,
+      "0.99": 11.656788052112361,
+      "0.999": 21.10364618458584
+    }
+  }
+}

results/nulls_linkage_threshold.txt

@@ -0,0 +1,8 @@
+3,312 MIGs, 82,800 randomisations
+  FP 0.100   threshold   3.64   calls   457 (13.80%)
+  FP 0.050   threshold   4.97   calls   249 (7.52%)
+  FP 0.010   threshold   8.68   calls    53 (1.60%)
+  FP 0.001   threshold  29.27   calls     4 (0.12%)
+
+1% point 8.68  bootstrap 95% CI [8.42, 9.14]  over 82,800 null scores
+nominal p<0.01 (score>2) calls 1,014 (30.62%)

results/nulls_position_pairs.txt

@@ -0,0 +1,16 @@
+125,373 distinct barcodes, 9 nt
+marginal m_j: 0.2512 0.2508 0.2500 0.2504 0.2506 0.2489 0.2489 0.2466 0.2489
+
+ separation  pairs  mean ratio  max ratio    argmax
+          1      8     1.00438    1.01128       7,8
+          2      7     1.00325    1.01088       5,7
+          3      6     1.00309    1.00930       5,8
+          4      5     1.00155    1.00251       4,8
+          5      4     1.00172    1.00424       2,7
+          6      3     1.00204    1.00436       2,8
+          7      2     1.00087    1.00175       1,8
+          8      1     0.99999    0.99999       0,8
+
+adjacent  mean 1.00438   distant mean 1.00228
+all 36 pairs, summed log10 excess: 1.1035x
+the adjacent-window null extrapolated: 1.041x

results/suggest.cycles.tsv

@@ -0,0 +1,33 @@
+cycle	A	C	G	T	entropy_bits	collision	consensus	consensus_fraction	deviation_from_uniform	mean_phred
+0	0.270923	0.205301	0.257113	0.266663	1.991699	0.252764	A	0.270923	0.044699	32.80
+1	0.271191	0.207799	0.252794	0.268216	1.992327	0.252570	A	0.271191	0.042201	32.88
+2	0.260145	0.214829	0.251519	0.273507	1.994404	0.251895	T	0.273507	0.035171	32.86
+3	0.264594	0.226742	0.249786	0.258878	1.997561	0.250833	A	0.264594	0.023471	32.85
+4	0.258525	0.219768	0.261271	0.260436	1.996371	0.251223	G	0.261271	0.030232	32.87
+5	0.258896	0.205309	0.263363	0.272432	1.991717	0.252758	T	0.272432	0.044691	36.28
+6	0.263860	0.213038	0.253609	0.269493	1.994197	0.251951	T	0.269493	0.036962	36.31
+7	0.266021	0.211668	0.248465	0.273846	1.993202	0.252297	T	0.273846	0.039867	36.37
+8	0.266939	0.207021	0.252163	0.273877	1.991917	0.252709	T	0.273877	0.042979	36.33
+9	0.019594	0.970866	0.003522	0.006018	0.225672	0.943013	C	0.970866	0.720866	37.08
+10	0.968345	0.014162	0.010935	0.006558	0.250719	0.938056	A	0.968345	0.718345	37.05
+11	0.008490	0.002787	0.969377	0.019346	0.235673	0.940146	G	0.969377	0.719377	37.14
+12	0.005530	0.004560	0.014301	0.975608	0.199325	0.952067	T	0.975608	0.725608	36.73
+13	0.004907	0.006492	0.013414	0.975187	0.203603	0.951236	T	0.975187	0.725187	36.96
+14	0.019737	0.006090	0.006386	0.967787	0.248862	0.937080	T	0.967787	0.717787	37.11
+15	0.973334	0.005460	0.013971	0.007235	0.216524	0.947656	A	0.973334	0.723334	37.01
+16	0.967139	0.015541	0.006885	0.010435	0.258128	0.935755	A	0.967139	0.717139	36.90
+17	0.006016	0.964210	0.003140	0.026634	0.260500	0.930456	C	0.964210	0.714210	37.14
+18	0.003991	0.015479	0.005562	0.974968	0.202207	0.950849	T	0.974968	0.724968	37.06
+19	0.012195	0.008205	0.004315	0.975285	0.203498	0.951415	T	0.975285	0.725285	37.03
+20	0.004711	0.005571	0.012721	0.976997	0.191027	0.954739	T	0.976997	0.726997	37.15
+21	0.012056	0.004225	0.018752	0.964967	0.267394	0.931676	T	0.964967	0.714967	37.13
+22	0.007805	0.009895	0.972664	0.009635	0.223964	0.946328	G	0.972664	0.722664	36.95
+23	0.007580	0.007000	0.978660	0.006760	0.182688	0.957927	G	0.978660	0.728660	37.08
+24	0.007455	0.021115	0.960465	0.010965	0.297493	0.923115	G	0.960465	0.710465	37.04
+25	0.006305	0.971880	0.013765	0.008050	0.227187	0.944845	C	0.971880	0.721880	37.14
+26	0.020306	0.962394	0.003345	0.013955	0.280894	0.926821	C	0.962394	0.712394	37.14
+27	0.955170	0.012170	0.008650	0.024010	0.329070	0.913149	A	0.955170	0.705170	37.03
+28	0.008540	0.029820	0.006570	0.955070	0.320772	0.913164	T	0.955070	0.705070	37.09
+29	0.005350	0.972415	0.002905	0.019330	0.214142	0.946002	C	0.972415	0.722415	37.06
+30	0.022790	0.964955	0.003100	0.009155	0.261817	0.931751	C	0.964955	0.714955	36.95
+31	0.951595	0.017345	0.004080	0.026980	0.342575	0.906578	A	0.951595	0.701595	36.89

results/suggest.segments.tsv

@@ -0,0 +1,3 @@
+kind	begin	end	length	consensus	mean_deviation
+umi	0	9	9		0.037808
+constant	9	32	23	CAGTTTAACTTTTGGGCCATCCA	0.718232