Scope And Non-Goals¶
bijux-proteomics-core is the scientific engine and contract library. Its
scope follows proteomics meaning rather than application workflow: an operation
belongs here when its result and failure boundary can be defined without
provider selection, execution state, evidence policy, recommendation posture,
or laboratory capacity.
In Scope¶
- parsing and validation of scientific formats and exported results;
- sequence, peptide, modification, mass, spectrum, and database operations;
- identification confidence, FDR, protein inference, and ambiguity;
- quantitative normalization, missingness, comparison, and QC;
- DIA, LFQ, multiplex, PTM, targeted, and DDA scientific surfaces;
- deterministic reports and portable scientific artifacts;
- benchmark packages, comparator evidence, acceptance bars, and challenge corpora;
- Python and file-oriented CLI interfaces over the same scientific owners.
Explicit Non-Goals¶
| Concern | Owner |
|---|---|
| universal identifiers, document metadata, canonical JSON | Foundation |
| processes, providers, state, retries, checkpoints, replay | Runtime |
| claims, citations, contradictions, biological evidence custody | Knowledge |
| candidate ranking, confidence, recommendation, refusal | Intelligence |
| assay feasibility, scheduling, handoff, observation, promotion | Lab |
| legacy Runtime namespace compatibility | Agentic Proteins |
| repository quality, generated governance, release automation | Maintainer tooling |
Core is not a vendor pipeline emulator. Import adapters can preserve declared semantics and field loss without proving external-engine parity. Benchmark fixtures can challenge a method without proving universal transfer. Biological interpretation can compute a result without becoming a curated evidence store or a recommendation.
Growth Rule¶
New scientific capability requires an owned domain boundary, explicit input and output contracts, units and orientation, rejected-input evidence, QC, limitations, focused tests, and a public navigation route. Avoid adding thin peer modules when the capability belongs to an existing scientific domain.
The Core public surface should grow by scientific responsibility, not by one file per delivery request or one wrapper per caller.