Skip to content

Entrypoints and worked examples

The package root exposes evidence models and high-value biological resolution functions. Specialized memory, reconciliation, reference, and review builders remain in their owner modules. Knowledge has no standalone CLI or HTTP API.

Create a contextual evidence bundle

from bijux_proteomics_knowledge import EvidenceBundle, EvidenceRecord
from bijux_proteomics_knowledge.memory.models.evidence import (
    EvidenceExtractionMethod,
    EvidenceKind,
    EvidenceOrigin,
    EvidenceSourceType,
    EvidenceStrength,
    QuantitativeSupport,
)

record = EvidenceRecord(
    evidence_id="evidence:akt1-phospho-17",
    kind=EvidenceKind.PHOSPHOPROTEOMICS,
    title="AKT1 site signal after treatment",
    source="run-17 differential PTM result",
    source_type=EvidenceSourceType.LAB_ASSAY,
    origin=EvidenceOrigin.OBSERVED,
    extraction_method=EvidenceExtractionMethod.AUTOMATED_IMPORT,
    biological_system="treated human cell line",
    endpoint="site abundance",
    claim="AKT1 site abundance increased after treatment",
    related_targets=["target:akt1"],
    decision_tags=["mechanism-review"],
    quantitative_support=QuantitativeSupport(
        effect_size=1.4,
        q_value=0.008,
        replicate_count=3,
        site_localization_probability=0.97,
        scale_type="log2-ratio",
    ),
    confidence=0.88,
    strength=EvidenceStrength.SUPPORTING,
)

bundle = EvidenceBundle(
    bundle_id="evidence:akt1-review-bundle",
    target_id="target:akt1",
    records=[record],
)
print(bundle.to_stable_json())

The record is intentionally verbose: context and quantitative limitations are part of the evidence, not optional commentary.

Resolve protein identifiers

from bijux_proteomics_knowledge import (
    render_protein_id_resolution_tsv,
    resolve_protein_ids,
)

# `annotation_pack` is a curated AnnotationPack from bijux-proteomics-core.
rows = resolve_protein_ids(("P31749", "AKT1", "unknown-protein"), annotation_pack)
print(render_protein_id_resolution_tsv(rows))

Inspect resolution_status and ambiguity_count before using a resolved accession. An alias match can be biologically unsafe when multiple entries share the same symbol.

Resolve pathway coverage

from bijux_proteomics_knowledge import (
    PathwayCoveragePolicy,
    resolve_pathway_members,
)

report = resolve_pathway_members(
    ("P31749", "P27361"),
    pathway_pack,
    policy=PathwayCoveragePolicy(minimum_coverage_fraction=0.6),
)
for entry in report.entries:
    print(entry.pathway_id, entry.coverage_fraction, entry.unresolved_inputs)

Coverage confidence describes curated member resolution only. Use core quantitative evidence and an explicit interpretation policy to assess activity.

Choose the next surface

  • memory.models owns evidence, claims, lineage, dossiers, and knowledge state.
  • memory.integrity owns graph structure and integrity validation.
  • memory.reconciliation owns conflict resolution and escalation.
  • root resolution functions own protein, feature, pathway, complex, kinase, drug, disease, ortholog, and coverage reports.
  • references owns grounding rules, scientific literature, comparators, benchmarks, risk, and reproducibility packs.
  • reviews owns provenance, contradiction stress, and decision briefs.

Persist the typed report and reference identity before handing any knowledge artifact to intelligence or lab workflows.