immunogenicity: the MHC class-II corpus the shipped class-II model is fitted on

Built from the same staged IEDB export as the class-I arm, by the same rules, so
the two are comparable by construction. 1,096,034 rows over 643,744 peptides,
77,943 immunogenic; human 603,781 peptides / 65,486 positive, mouse 50,258 /
12,457.

Restriction is parsed rather than imputed -- 50.8% of rows resolve, the rest are
kept as an explicit `unresolved` group rather than assigned a guessed allele.
That is why class II has one arm where class I has two: an HLA-matched variant
needs a resolved allele on both sides. The complementarity fit reads only peptide
and label, so the unresolved rows still carry their weight.

Deposited with the generator's run log; the build is deterministic and every count
was reproduced exactly from a clean checkout on 2026-08-19.

immunogenicity/SOURCES.md

@@ -319,3 +319,51 @@ Regenerate (needs `~/hf/pmhc_data/dump`, which is gitignored — the raw zips ar
 ```zsh
 python immunogenicity/staged/fetch.py
 ```
+
+## `corpus_iedb_mhc2.parquet` — the MHC class-II arm (2026-08-19)
+
+**Derived/computed**, from `staged/iedb_tcell.parquet` and `staged/iedb_ligand.parquet` — the same
+IEDB build 2026-08-11 export the class-I corpus is built from, by the same rules, so the two arms are
+comparable by construction rather than by assertion. This is what the shipped class-II
+complementarity model (`mhcmatch.complement`, `cls="mhc2"`, v0.16.0) is fitted on.
+
+| field | value |
+|---|--:|
+| rows (peptide, allele group, host) | 1,096,034 |
+| distinct peptides | 643,744 |
+| immunogenic peptides | 77,943 |
+| human rows | 1,036,041 |
+| human positives | 65,486 |
+| human distinct peptides | 603,781 |
+| human allele groups | 121 |
+| mouse rows | 59,993 |
+| mouse positives | 12,457 |
+| mouse distinct peptides | 50,258 |
+| mouse allele groups | 40 |
+| peptide length | 11–25, median 15 |
+
+Three rules differ from the class-I arm and nothing else does:
+
+- **Length** is 11–25 rather than 8–11.
+- **Restriction is parsed, never imputed.** Class-II restriction is written a dozen ways across IEDB
+  (`HLA-DRB1*15:01`, `HLA-DPA1*01:03/DPB1*04:01`, `HLA-DR15`, serotypes, `H2-IAb`), so the corpus
+  carries a parser and keeps whatever it cannot resolve as an explicit `unresolved` group rather than
+  guessing an allele. It parses **1,003,554 / 1,974,413 rows (50.8 %)**; the remainder are retained,
+  not dropped, because the complementarity fit reads only peptide and label. There is therefore **one
+  class-II arm where class I has two** — no HLA-matched variant is built, since matching needs a
+  resolved allele on both sides.
+- **Rule 8** (pathogen epitopes) is applied by parsing the source organism rather than by set-match.
+
+`evidence` records why each row is what it is: `eluted_but_positive_elsewhere=5,305`,
+`eluted_foreign=103,351`, `eluted_self=1,156,574`, `eluted_unknown=440,458`, `tcell_negative=110,312`,
+`tcell_positive=158,413`.
+
+`corpus_iedb_mhc2.log` is the generator's own run record, deposited beside it; the counts in the table
+above are read from it. The build is deterministic — regenerated 2026-08-19 from a clean checkout, every
+count reproduced exactly.
+
+Regenerate (needs this mirror at `$MHCMATCH_PMHC_DIR`, and the benchmark repo):
+
+```zsh
+python bench/neoag/corpus_iedb.py --cls mhc2
+```

immunogenicity/corpus_iedb_mhc2.log

@@ -0,0 +1,24 @@
+# tcell: 570,434 staged rows
+#   class II: 247,456
+#   host human or mouse: 239,196
+#   11-25mer, canonical: 221,787
+#   after artefact exclusions: 221,787  (115,263 peptides)
+# ligand: 5,773,495 staged rows
+#   class II: 1,966,913
+#   host human or mouse: 1,905,785
+#   11-25mer, canonical: 1,752,626
+#   after artefact exclusions: 1,752,626  (538,321 peptides)
+# T-cell assayed (peptide, host) pairs: 119,615
+#   positive: 39,818  |  measured negative: 79,797
+#   peptides positive in any host: 38,320
+# elution pool by origin: foreign=145,186, self=1,165,489, unknown=441,951
+# rows entering allele resolution: 1,974,413 (1,324,898 distinct peptide/host/restriction)
+# class-II restriction parsed on 1,003,554/1,974,413 rows (50.8%); the rest are kept as `unresolved`
+#   top groups: unresolved=970,859, HLA-DRB1*15=174,134, HLA-DRB1*13=163,275, HLA-DRB1*01=105,737, HLA-DRB1*03=83,487, HLA-DRB1*04=52,007, HLA-DRB1*07=39,065, HLA-DPA1*01/DPB1*04=37,066
+# evidence: eluted_but_positive_elsewhere=5,305, eluted_foreign=103,351, eluted_self=1,156,574, eluted_unknown=440,458, tcell_negative=110,312, tcell_positive=158,413
+# out: 1,096,034 (peptide, allele group, host) rows over 643,744 peptides, 77,943 immunogenic
+#   human: 1,036,041 rows, 65,486 positive, 121 allele groups, 603,781 peptides
+#     length 11-25, median 15
+#   mouse: 59,993 rows, 12,457 positive, 40 allele groups, 50,258 peptides
+#     length 11-25, median 15
+# -> /Users/mikesh/vcs/projects/2026-mhcmatch-benchmark/bench/neoag/corpus_iedb_mhc2.parquet  (3s)

immunogenicity/corpus_iedb_mhc2.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63fe69c10d635f5c17bebae135bd390dc475ee7cffbde59e04c2dcdfca639708
+size 8509426