Why Trust DIA¶
This page is about the current flagship dia result surface.
The right trust level is bounded. The repository now ships two public DIA packages plus one published cross-package report, so a reviewer can inspect whether DIA trust survives beyond one library-conditioned package. The authority still stops at library-conditioned review rather than vendor-execution parity.
The meaningful change since v0.3.7 is that DIA now sits on a clearer chain of
runtime, grounding, recommendation, and consequence surfaces. The boundary is
still real: stronger public packets do not yet become chromatogram-level or
vendor-parity authority.
Open First¶
benchmark:dia_library_extraction_consistencypackages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/package_manifest.jsonpackages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/README.mdpackages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/primary/spectronaut_report.tsvpackages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/primary/spectronaut_pipeline_export.tsvpackages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_library_review_package/comparator/diann_pipeline_export.tsvpackages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_matrix_shift_review_package/package_manifest.jsonpackages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_matrix_shift_review_package/cross_package_generalization.jsonpackages/bijux-proteomics-runtime/src/bijux_proteomics_runtime/workflows/benchmark_runs.py
Current Trust Earned¶
outsider_review:diais complete enough to count as an outsider-auditable flagship family.- benchmark evidence tier is
external_reproduction_package. - public claim support is
advisory. - the runtime package
dia-diann-pipeline-corpusis real and currentlyraw_executable. - the companion family-transfer report is currently
highly_stableat0.83, but protein-level absence language still weakens on the second package. - the recommendation posture is
recommend_with_downgrade. - the lab posture is
exploratory_only.
Best Cross-Surface Audit¶
- open Workflow Families first to see how DIA is released beside DDA, LFQ, PTM, and targeted
- open Benchmark Assets and the DIA package files above when the question is whether library-shaped proof is truly inspectable
- open Black-Box Benchmark Dashboard when the question becomes how much rerun trust the raw-executable lane actually earns
- open Workflow Claim Grounding, Workflow Recommendation Confidence, and Workflow Consequence Maps when the dispute shifts from extraction success to release-grade language
What The Reader Is Really Auditing¶
- whether library-conditioned DIA evidence stays inspectable across the primary and companion packages
- whether the raw-executable runtime lane preserves rerun trust instead of outsourcing it to maintainer narration
- whether absent-peptide and matrix-shift pressure remain visible in the final sentence
Why This Packet Is Scientifically Real¶
- DIA now has a real paired review surface rather than one convenient library-shaped package being asked to carry the whole trust story
- the raw-executable runtime lane makes rerun realism part of the public product rather than an internal promise
- the scientific value is bounded transition- and library-conditioned review, not vague workflow optimism
Exact Claims¶
- DIA adapter normalization preserves library-conditioned transition semantics across the pinned export corpus
- DIA review surfaces keep capability limits explicit instead of implying vendor-pipeline parity
Why This Surface Is Stronger Now¶
- DIA now has a companion matrix-shift packet instead of one primary package being asked to carry the full trust story alone
- the runtime lane is raw-executable enough to make rerun trust part of the public story rather than a maintainer-only claim
- the surrounding grounding, recommendation, and consequence pages now make the library-conditioned downgrade easier to defend and harder to overstate
What You Can Trust Right Now¶
- the repo keeps DIA partial-support language explicit
- the DIA-NN and Spectronaut confrontation is visible instead of implied
- the runtime lane is raw-executable enough to preserve lineage and artifact browsing without stopping at an import-only bridge
- the tracked public package makes artifact inventory, quality posture, and lifecycle boundaries inspectable without reading internal code
- the companion matrix-shift package publishes a second family-transfer check at
packages/bijux-proteomics-core/benchmark-assets/flagship-public-packages/dia_matrix_shift_review_package/cross_package_generalization.json
Why The Public Sentence Still Narrows¶
- library incompleteness still limits broader biological confidence
- absent-peptide reasoning weakens under the companion matrix-shift package
- chromatogram-level and vendor-execution parity remain outside the current proof boundary
Honest Reader Outcome¶
- a reviewer should trust the bounded DIA review packet more than the older docs implied
- the same reviewer should still stop short of chromatogram-native or vendor-parity authority
What You Should Not Trust Yet¶
- vendor-library parity is not earned
- chromatogram-level and vendor-execution parity are not earned
- the current reproduction story still depends on execution steps outside the repository proof boundary
- protein-level absence claims stay downgrade-heavy because the second package weakens them materially
Consequence Boundary¶
- open Workflow Consequence Maps before widening DIA language from library-conditioned support to broader biological confidence
- open What Changed The Recommendation before claiming that one repeat or one matrix-shift outcome changed the full public call
- open Workflow Refusal Handbook when the honest next move may still be stop, rerun, narrow, or refuse
Evidence Grounding¶
- sentence grounding and unsupported-claim review: Workflow Claim Grounding
- citation freshness, bibliography export, and gap audits: Workflow Literature Audits