Load a blood-transcript delivery from your laboratory: three CSVs — the run and assay identity, one row per sample, and the expression matrix. A targeted panel or a whole transcriptome (Ensembl gene ids, raw counts). Nothing is written until you have seen what was understood, every sample is resolved to an animal or explicitly rejected, and the QC gate has run. Full file specification: API & formats.
1The three files
run.csv carries the assay identity, version, provider and — for a whole transcriptome — the reference assembly, annotation version and quantification tool. samples.csv carries one row per sample with its animal identifier, sampling session and QC metrics. expression.csv is genes down, sample barcodes across.
2What was understood
Read this before committing. Nothing has been written yet.
—
Samples in file
—
Resolved to an animal
—
Sampling sessions
—
Genes / targets
—
Expression values
—
Assay & version
—
QC passed / scorable
Resolution worksheet
Every sample's match is recorded as evidence. A sample that matches no animal, or more than one, must be resolved in the registry or explicitly rejected here before anything is committed. A rejected sample is stored as rejected — no animal, no result — so the fact that it arrived is kept.
QC gate
Thresholds come from Genemap's QC table, each naming its source. A result below a threshold is stored as failed and excluded from scoring; a result the gate cannot judge (no mapping rate, a sex mismatch) is held with the reason. Only passed results with raw counts are scorable.
Expression data recorded here is a physiological state reading. Scores derived from it are not breeding values, are not used in any EBV or index, and no genetic parameter is estimated for any expression trait in this phase.
3Deliveries
Every assay run in your herd, what the QC gate made of it, and whether it has been processed (normalised, batch-estimated, scored). Reprocess re-runs the same chain; it writes nothing twice. Declare anchor samples where reference material was re-assayed across plates.
Phase E.0: every whole-transcriptome delivery in this organisation, oldest first, through the same chain — writes nothing twice; one log row per delivery, all sharing one batch id.