Roadmap: species-scoped RNA, translation and protein abundance data

未关闭
#3 4 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
20/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
python, sql

调研方向

Start with the linked implementation and data issues, especially the unchecked P0 and P1 items, and read the released coverage notes to understand what is already shipped and where gaps remain. Compare each candidate issue with the completion criteria for species and assay coverage, versioned references, provenance, and offline reproducibility. Done means the remaining planned work is implemented and the documented coverage matrix and explicit gaps are accurate.

由索引模型根据 Issue 内容生成。

描述

enhancement

Extend orfdata's PyEnsembl reference/evidence layer with quantitative RNA abundance, ribosome occupancy, protein intensities and calibrated protein copies per cell. Initial species are human, mouse, dog and budding yeast. Any species must fit the same public objects and storage contract; cats, cattle, rats and other organisms are subsequent coverage, not a different architecture.

Structure and ownership

  • PyEnsembl supplies annotations and canonical transcripts/proteins. orfdata supplies experimental observations and additional ORFs; Varcode/Isovar remain consumers of reference and altered-product annotations.
  • Keep ORFModel, ReferenceORF and existence evidence; add a separate quantitative dataset layer with explicit samples, assays, targets, units and provenance. Canonical proteins and ncORFs use the same observation machinery.
  • Namespace every target and sample by dataset and reference context. Species, strain, assembly and annotation release must never disappear from joins, caches or exports.
  • RNA belongs to a gene or transcript as measured; Ribo-seq can support a particular ORF; peptides can belong to several proteins. Preserve those distinctions and mapping ambiguity.
  • Keep direct observations, derived quantities and predictions distinct. Orthology is a comparison relationship, never permission to copy measurements or protein-existence evidence.
  • Retain genuine missingness and assay coverage. No protein observation must not become zero protein or evidence of absence.

Initial parallel coverage

Modality Human Mouse Dog Budding yeast
RNA + relative protein abundance 2026 multispecies atlas + healthy-tissue atlas 2026 multispecies atlas 2026 multispecies atlas + canine proteomics CoPY RNA/protein reference
RNA + ribosome occupancy Chothani smORF atlas Tissue/stage translatome Qualified dog-source audit; enriched pancreas is a lead, not a body atlas Ingolia rich/starvation data
Calibrated protein abundance Targeted human calibration studies Mouse abundance/turnover studies Explicit coverage gap until a calibrated source is verified CoPY isotope-standard measurements
ncORF identity/evidence GENCODE/TransCODE and quantitative microprotein studies OpenProt and source-defined mouse ORFs Verify direct dog sources; do not project human ORFs as observations OpenProt/yeast source-defined ORFs

The leading parallel RNA/proteomics source is Oziolor et al. 2026. The main absolute reference is Lawless et al. 2016. Dataset-specific issues below contain accession and measurement details.

Completion criteria

  • The same query/export path works with real, versioned reference/data slices for all four initial species; no mammal-only data model.
  • Cross-species, cross-assembly and incompatible-assay joins fail explicitly; documented orthology comparison is opt-in.
  • Publish a species-by-assay coverage matrix, including tissue/sample pairing, direct versus inferred values, and known gaps.
  • Importers run reproducibly from pinned inputs and a local cache. Small source-backed fixtures run offline; large downloads are explicit.
  • Prioritize foundations, the parallel atlas and yeast absolute reference, then other importers. An issue tracks planned work; filing it does not mean the data have been imported.

Extends #1 and the draft implementation in #2. The task list links the individual implementation and source-import issues.

Implementation and data issues

P0 — Foundations
  • #4 — Define quantitative observations, sample metadata and versioned dataset adapters
  • #5 — Fix dog species aliases and enforce species/reference isolation for arbitrary organisms
  • #6 — Add explicit RNA/Ribo/protein pairing and orthology comparison views
  • #43 — Query locally built datasets through the offline registry/CLI (released 0.12.0 via PR #44)
  • #46 — Represent source-defined RNA/Ribo regions without mislabelling complete ORFs (released 0.14.0 via PR #47)
  • #48 — Preserve aggregate ORF support separately from peptide/study-resolved observations (released 0.17.0 via PR #53)
  • #49 — Query paired measurements directly by scoped target and native units (released 0.15.0 via PR #51)
  • #54 — Manage optional ORF catalogues through explicit DataCache-backed registry APIs (released 0.18.0 via PR #56)
P1 — First data imports and required coverage audits
  • #7 — Import parallel human, mouse and dog RNA/proteomics from the 2026 nonclinical atlas
  • #23 — Extract canine Pfizer peptide/protein-group evidence with bounded access (focused work under #7)
  • #24 — Publish the compact Pfizer canine artifact and shared consumer contract (focused work under #7; peptide enrichment follows #23)
  • #8 — Import the healthy human tissue RNA/protein atlas and protein-per-mRNA measurements
  • #9 — Import human smORF RNA abundance, ribosome occupancy and translation-efficiency data
  • #10 — Import matched mouse tissue/stage RNA and Ribo-seq with linked proteomics
  • #11 — Import yeast rich/starvation paired RNA and ribosome-footprint measurements
  • #12 — Qualify and import dog RNA/Ribo-seq; track the genuine coverage gap
  • #13 — Import CoPY yeast absolute protein copies and RNA from the same cells
  • #15 — Complete the curated TransCODE MS import and preserve sample-level peptide evidence
P2 — Additional evidence and quantitative references
  • #14 — Add human and mouse absolute-abundance and turnover calibration datasets
  • #16 — Version and import the expanded GENCODE translated-ORF reference catalogue
  • #17 — Import direct quantitative microprotein studies with transcript and peptide mappings
  • #18 — Add a release-pinned OpenProt adapter for multispecies alternative-ORF evidence
  • #55 — Qualify OpenProt coordinate conventions against native references (source discrepancy under #18)
  • #19 — Import CCLE, ProCan and CPTAC quantitative RNA/protein comparator cohorts
  • #20 — Import canine DIA protein measurements and a genuinely paired RNA/protein study
  • #21 — Add optional PaxDb and ProteomicsDB abundance adapters with source-level provenance
P3 — Later RNA/species expansion
  • #22 — Expand RNA context with parallel species adapters; track cat, cattle and rat coverage

Released coverage (2026-10-13)

Foundations #4–#6 are released through ORFdata 0.2–0.5. ORFdata 0.6 ships complete human/mouse/dog Pfizer native protein-group summaries (#24, part of #7), with 2,509,912 measurements. ORFdata 0.7 ships 69,125 CoPY yeast RNA/protein/peptide observations (part of #13); its historical genomic reference and individual cross-assay replicate map remain unresolved. Both use the same explicit download/build/query APIs and source-scoped identities. These releases do not yet add nonhuman experimental ncORF catalogues or the remaining RNA/Ribo-seq importers.

ORFdata 0.8–0.10 are also merged, deployed and verified on PyPI: compact lossless storage (#40), native unit/scale selectors (#41), and complete pinned human/mouse/dog search-reference sequences (#42). Search membership remains distinct from detection and group attribution. ORFdata 0.11.0 is merged, deployed and verified on PyPI via PR #50. It ships native dog/mouse RNA counts under #7: 47,395,300 source-exact observations across 434 dog and 606 mouse libraries, including noncoding gene targets and measured zeros. Both public artifacts passed full-source reconciliation, fresh installation and offline queries. Missing human native atlas counts (#37) and library-to-protein-pool correspondence remain explicit gaps.

The Wang human tissue data artifact is published and independently verified: 6,273,425 source-exact observations across 29 specimens, including native RNA/protein measurements and explicitly calibrated copy estimates. Fresh public download, hashes, installation and offline queries pass. Software integration is merged and deployed as ORFdata 0.13.0 via PR #45, verified on PyPI with matching wheel/sdist digests. All 333 checkout/sdist tests and six CI jobs pass. The separate Eraslan reprocessing view remains under #8.

OpenProt 2.2 human/mouse/yeast source inventories, native sequences/aliases, coordinate conventions and evidence-resolution caveats are audited in #18. Dog/cat are absent from its current species list; no nonhuman ncORF catalogue is claimed as released. The native yeast RNA/Ribo tables are now released as region measurements in ORFdata 0.16.0; historical source-region coordinates and exact aliquot pairing remain to be qualified under #11.

ORFdata 0.12.0 is merged, deployed and verified on PyPI: explicit offline open_build/inspect_build and CLI --built queries (#44/#43). Source and artifact cache origins stay separate; full dog/mouse/CoPY builds were queried offline. All 309 checkout/sdist tests and six CI checks pass.

ORFdata 0.14.0 is merged, deployed and verified on PyPI: explicit source-region targets (#47/#46), using the same species/reference identity and storage/query APIs. Native yeast RNA/Ribo fixture observations pass round-trip checks; the source importer subsequently shipped in 0.16.0, with remaining reference/pairing work under #11. All 336 checkout/sdist tests and six CI jobs pass.

ORFdata 0.15.0 is merged, deployed and verified on PyPI: scoped left/right paired target selectors and native per-side filters (#51/#49). Narrow queries deserialize only selected relationships/observations; native references, pairing strength, group ambiguity and missing mates are retained. All 353 checkout/sdist tests and six CI jobs pass.

ORFdata 0.16.0 is merged, deployed and verified on PyPI via PR #52. The pinned Ingolia yeast release contains 329,952 source-exact observations for 13,748 regions across eight libraries, including RNA and ribosome occupancy with native normalization inputs. All values and 219,968 input links independently reconcile; local and public installation/offline queries pass. All 370 checkout/sdist tests and nine exact-commit CI checks pass. Exact genomic windows, aliquot pairing and complete ORF models remain under #11. Data release.

ORFdata 0.17.0 is merged, deployed and verified on PyPI via PR #53: typed aggregate existence-evidence counts, shared source identity and explicit resolved-evidence filters (#48). All 396 checkout/sdist tests, six CI jobs, 29 Varcode and 48 Isovar compatibility tests pass. Four-species modified-product probes preserve base support without inventing retained peptides or mutant translation. OpenProt native reference qualification continues under #18; optional catalogue download/build/query management is tracked in #54.

OpenProt yeast qualification now validates all three native Ensembl106 assignments against original GTF/transcript sequences and PyEnsembl binding, retaining S288c strain identity. The three parallel RefSeq assignments remain unresolved. Source genomic endpoint conventions disagree with the README; #55 records the validated yeast +1 conversion and requires independent human/mouse qualification. These local audits are not yet a released experimental catalogue.

ORFdata 0.18.0 is merged, deployed and verified on PyPI via PR #56: explicit optional catalogue discovery/download/build/offline-query APIs through DataCache (#54). All 409 checkout/sdist tests and six CI jobs pass. The registry remains empty until source artifacts are qualified and published; a local managed three-model yeast audit is not counted as released coverage.

主要语言
Python
星标
1
派生
1
平均合并
10 小时 11 分钟
30 天内合并 PR
51

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南,环境需要你自己搭建:先看它的 README,通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

openvax/orfdata 的其他 Issue

查看 openvax/orfdata 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。