DESCRIPTION: document neoag_tested_{mmu,hsa} + flag 6186 mouse CEDAR rows mislabelled as human

DESCRIPTION.md

@@ -15,6 +15,29 @@ omits the feature columns IEDB doesn't populate. In `neoag_tested`/`neoag_candid
 primary datasets take precedence over the aggregated compendia (CEDAR, Neopep): a duplicate
 peptide–MHC already present in a primary dataset is dropped from CEDAR/Neopep.
 
+The `neoag_tested_{mmu,hsa}` pair (`dataset_origin=IEDB_neoag`) is split on **`mhc_species`**,
+not host: H-2 restrictions go to `_mmu`, HLA restrictions (HLA-transgenic mice and human hosts)
+to `_hsa`; `host_species`/`source_species` preserve the distinction within each file. They share
+the schema above and append 25 extra columns (`peptide_wt`, `strain`, `tumor_model`, `gene_name`,
+`uniprot_id`, `structure_id`, assay counts, and NetMHCpan `*_pred` fields). **Canonical columns
+hold as-reported values only**; every predicted value lives in a `pred_*` column, so `affinity`
+stays experimental (nM, 123 rows) while `affinity_wt`/`agretopicity` are null (IEDB assays the
+neo-epitope, not its WT counterpart — predictions are in `pred_affinity_wt_nm`/
+`pred_agretopicity`). `expression` (805 rows) is a **reference proxy and is not comparable to
+TESLA's**: TESLA reports the source gene's expression in the patient's own tumour, whereas here
+it is that gene's mean TPM in the same syngeneic model measured by an unrelated study
+(GEO GSE281579, untreated tumours only; `expression_model`/`expression_sd`/`expression_n` record
+the join). These two tables **overlap `neoag_tested.tsv.gz` via CEDAR by design**
+(both are IEDB-derived): all 2,273 peptides also appear there, so dedupe on peptide × MHC when
+joining. The precedence rule above was *not* applied — `neoag_tested.tsv.gz` is unchanged.
+
+**Caveat — mouse rows mislabelled as human in `neoag_tested.tsv.gz`:** 6,186 CEDAR rows carry
+mouse MHC alleles (`H2-Db` 1,020, `H2-Kb` 932, `H2-b class I` 719, `H2-Kd` 525, `H2-d class II`
+505, …) but are labelled `mhc_species=HomoSapiens`, `host_species=human`. `SIINFEKL`, the
+canonical H-2Kb ovalbumin epitope, is among them. Filtering that table on `mhc_species` therefore
+silently admits mouse pMHC. Pre-existing and left unfixed; `neoag_tested_mmu.tsv.gz` is the
+correctly-labelled murine set.
+
 The `thymus/` tables use the reference-peptide schema (`peptide, mhc_a, mhc_class, mhc_species,
 source_protein, species, dataset_origin`; `mhc_a` is null — the HLA Ligand Atlas gives
 donor-level candidate alleles, not a per-peptide deconvolution, and the seqtree self-similarity
@@ -30,8 +53,10 @@ the mouse thymus proteome (PXD007288) is a 12 GB MaxQuant archive, and the mTEC-
 | `pmhc/pmhc_full.tsv.gz` | 6 | IEDB epitope–MHC | positive IEDB MHC-ligand binding records (human IMGT/HLA + mouse H-2) |
 | `pmhc/pmhc_shortlist.tsv.gz` | 6 | IEDB epitope–MHC (≥2 refs) | high-confidence pairs (≥2 references) |
 | `immunogenicity/iedb_labeled.tsv.gz` | 7 | IEDB immunogenicity — **ipred training set** | immunogenic (T-cell+, 34,362) vs non-immunogenic (MHC-ligand+ in healthy self, 756,506); `host_species` (human/mouse) is the cross-validation axis; ipred featurizes `peptide` (MJ/hydropathy/Kidera) |
-| `immunogenicity/neoag_tested.tsv.gz` | 9 | Neoantigens — tested (workhorse) | tested patient-derived neoantigens, immunogenicity 0/1 (TESLA, Neopep, CEDAR, Gfeller, ITSNdb, GBM, VACCIMEL, Bjerregaard) |
+| `immunogenicity/neoag_tested.tsv.gz` | 9 | Neoantigens — tested (workhorse) | tested patient-derived neoantigens, immunogenicity 0/1 (TESLA, Neopep, CEDAR, Gfeller, ITSNdb, GBM, VACCIMEL, Bjerregaard). **Human only by label — see the mouse-contamination caveat below.** |
 | `immunogenicity/neoag_candidates.tsv.gz` | 9 | Neoantigens — candidates | untested neoantigen candidates (Neopep `not_tested`) |
+| `immunogenicity/neoag_tested_mmu.tsv.gz` | 12 | Neoantigens — tested, **mouse MHC** | 1,977 murine tested neo-epitopes (IEDB), immunogenicity 0/1 (740 immunogenic); `mhc_species=MusMusculus` |
+| `immunogenicity/neoag_tested_hsa.tsv.gz` | 12 | Neoantigens — tested, **human MHC in mouse models** | 717 rows from the same IEDB export whose restriction is HLA — HLA-transgenic mice plus human-host epitopes (KRAS G12D, CDK4 R24C, PMEL); `mhc_species=HomoSapiens` |
 | `tcga/tcga_neoantigens.tsv.gz` | 10 | TCGA neoantigens | 2.24M expressed mutant-peptide–HLA binders across 8,505 TCGA donors (derived from open-access TCGA somatic + expression; per-barcode HLA/neoantigens as in TCIA / Thorsson 2018) |
 | `immunogenicity/hla_pop_freqs.tsv.gz` | — | HLA population freqs | EUR/ASN/AFR class-I (A/B/C) allele frequencies (AFND) |
 | `proteome/human.fasta.gz` | 4 | Human proteome (UP000005640) | self-reference proteome for similarity search |