v1 · processed omics
Omics Readout v1
The processed-first omics method for scoring target expression and gene-set signals from public matrices before any raw FASTQ/BAM reprocessing.
Omics Readout v1 starts with processed public matrices. It computes reproducible target and pathway scores first, then lets specialist agents interpret what the numbers can and cannot support.
Deterministic math first. LLM interpretation second.
Processed matrices are small enough for local or Dagster CPU execution and let TWOG test whether a signal is worth deeper raw-data reprocessing later.
The first built-in panels cover vimentin target expression, mesenchymal/ECM state, angiogenesis/endothelial signal, and coagulation/vascular injury signal.
If sample labels support tumor/control comparison, the method can report cohort differences. If labels are missing, the output stays descriptive.
The Modal GPU lane is reserved for later raw-data work: FASTQ/BAM reprocessing, large single-cell jobs, spatial analysis, or containerized workflows that exceed local CPU scope.
- Dataset source
- Accession, source URL, artifact hash, and matrix type.
- Parsing status
- Supported format, skipped raw file reason, missing labels, and unsupported columns.
- Normalization
- Count-like CPM/log handling or dataset-level z-scoring for processed expression.
- Scores
- Target expression and gene-set score distributions by sample or cohort.
- Interpretation limits
- Whether labels support tumor/control comparison or only descriptive evidence.
- Processed-matrix readouts are reproducible summaries, not final biological proof.
- Missing labels prevent differential claims.
- Unsupported raw files are recorded as negative coverage, not hidden failures.
- Agent review should explain how a signal changes candidate confidence.
Omics Readout v1 does not reprocess raw sequencing data or prove mechanism. It produces bounded, reproducible expression evidence for specialist review.