Known limitations

Core contains substantial native scientific machinery, importers, benchmark packages, and review artifacts. Its breadth is uneven across workflow families and must not be read as universal raw-data processing, external-engine parity, or decision-grade authority.

Current scientific limits

Limitation	Consequence	Responsible interpretation
DDA’s strongest public lane imports external search-engine results	repository review does not reproduce the upstream database search	retain engine, version, parameters, database, and import provenance
DIA execution begins with checked report-level inputs	it does not establish chromatogram-native replay or universal library coverage	bound claims to the declared reports and library context
LFQ evidence is sensitive to design, normalization, missingness, and cohort transfer	repeatability on checked cohorts is not external quantitative truth	publish policy and transfer envelope with the result
multiplex transfer remains fragile	substantive models and executable fixtures still support only internal-use language	do not borrow status from LFQ or targeted workflows
PTM localization is stronger than functional interpretation	a localized site is not automatically causal, occupied, or regulatory	preserve ambiguity and route biological claims through evidence review
targeted workflows remain bounded by calibration, matrix, interference, carryover, and assay burden	executable QC is not universal vendor or laboratory transfer	retain panel, matrix, calibration, and readiness context
format adapters cover declared subsets and mutation fixtures	parsing one producer version does not prove all dialects	name producer, version, unsupported constructs, and rejected rows
checked benchmarks are finite and repository-governed	they cannot represent every instrument, sample, organism, or study design	state the evidence envelope and require outside data for broader claims

Boundary of a passing Core result

flowchart LR
    C["Core result"] --> V["scientific contract satisfied"]
    V --> R["Runtime execution evidence"]
    R --> K["Knowledge grounding"]
    K --> I["Intelligence judgment"]
    I --> L["Lab consequence"]

A passing Core result establishes only its declared scientific contract. It does not establish that execution is reproducible, evidence is well grounded, the recommendation is robust, or the proposed assay is feasible. Those are separate owners and separate proof obligations.

Report the envelope

Name the workflow family, input level, producer and version for imported data, policy and parameters, benchmark assets, negative cases, and untested transfer conditions. Use the current workflow-family status as the public ceiling; never widen it from code volume or test count alone.