The bundled uorf-evidence catalogue duplicates the uORF records already in orf-evidence

オープン
#201 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
4/5
見積もり時間
3〜5日
初心者へのやさしさ
45/100
issue の種類
リファクタリング
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
python
領域
data

調査の方向性

Start with orfdata/catalog.py to trace _parse_catalog, _load_catalog, and load_uorf_catalog, then inspect orfdata/build.py and pyproject.toml for the default build output and packaged data. Check how tests and downstream users rely on the legacy catalogue before changing its handling. Done means the bundled uORF data is no longer duplicated, the default build produces the full catalogue, and legacy file support or its removal is addressed.

索引モデルが issue の本文から書いたものです。

説明

Problem

orfdata ships two human GRCh38 catalogues, built by the same build.py from the same pinned sources:

  • data/orf-evidence: all 7,264 ORFs;
  • data/uorf-evidence: 3,771 uORF and uoORF models in the legacy isovar.uorf_evidence.v1 schema.

Every record in the second catalogue is also in the first. Drop the four fields that the legacy loader fills in with fixed values, and the 3,771 uORF/uoORF records in orf-evidence/catalog.json.gz are identical to the 3,771 records in uorf-evidence/catalog.json.gz. As a result:

  • the wheel carries 672 KB of duplicate data;
  • the catalogue parser keeps a legacy-schema branch to read it;
  • python -m orfdata.build without --all-orfs still writes the legacy catalogue by default.

Evidence

Read at d74c9c2.

The records match. I decompressed both files and compared the records with jq -S, after applying del(.source, .species, .taxon_id, .translation_table) to the orf-evidence side. Across the 3,083 uORF and 688 uoORF records there are no differences.

  • Every uORF/uoORF record in orf-evidence has source="gencode_phase1", species="homo_sapiens", taxon_id=9606 and translation_table=1. These are the values that _record(row, legacy=True) fills in (orfdata/catalog.py:971-972).
  • The metadata differs only in dataset_version (2026-10-08.2 against 2026-10-11.1), model_scope, and the legacy file's sources list and manifest schema.

Code kept for the second copy:

  • orfdata/catalog.py:951-962: _parse_catalog accepts both schemas and handles the legacy one separately.
  • orfdata/catalog.py:1009-1035: UORFRecord, UORFCatalog, UORFPeptide = PeptideEvidence and load_uorf_catalog. load_uorf_catalog loads resource="uorf-evidence" and then filters to uORF/uoORF records anyway (:1033-1034).
  • orfdata/build.py:563,574,595,599: the schema, loader and manifest all depend on all_orfs. :617 sets the default --output to orfdata/data/uorf-evidence.

Packaging and docs:

  • pyproject.toml:61 packages both directories. orf-evidence/catalog.json.gz is 1,304,648 bytes and uorf-evidence/catalog.json.gz is 671,566 bytes.
  • docs/reference.md:327-328: "load_uorf_catalog() preserves Isovar's upstream-only snapshot ... New code should use the broader load_catalog() API."

Users of the legacy API:

  • isovar, read at 064cd02: isovar/uorf_catalog.py, isovar/cli/isovar_uorf_evidence.py and isovar/uorf_annotation.py. isovar's tests/test_uorf_catalog.py:181-190 reads data/uorf-evidence/manifest.json and catalog.json.gz directly.
  • varcode, read at 96dc77b: varcode/uorf.py:13-20.

Suggested fix

Keep load_uorf_catalog() and UORFCatalog as the uORF view, but build them from the one bundled catalogue. Load orf-evidence (as _load_catalog(..., resource="orf-evidence") does now) and keep the uORF/uoORF records:

def load_uorf_catalog(path=None, *, expected_sha256=None):
    """Upstream ORFs (uORF, uoORF) from the bundled catalogue, or from a local file."""
    catalog = _load_catalog(path, expected_sha256=expected_sha256)
    return UORFCatalog(
        [r for r in catalog.records if r.biotype in ("uORF", "uoORF")], catalog.metadata)

Then:

  1. Delete orfdata/data/uorf-evidence/ and remove it from package-data.
  2. Make build.py build the full catalogue into orfdata/data/orf-evidence by default.
  3. Keep the isovar.uorf_evidence.v1 reader only for legacy files that users supply, or drop it in a major release.
  4. Ask isovar to point tests/test_uorf_catalog.py at orf-evidence. Its record counts come from the records, which are identical.

https://claude.ai/code/session_01JHJhwj29oDqoeu7TYd34Pi

主要言語
Python
スター
1
フォーク
1
平均マージ
10時間 11分
マージ済み PR(30日)
51

環境構築

このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

openvax/orfdata のほかの issue

openvax/orfdata の issue をすべて見る

似ている issue

Python の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。