mcclain-thiel/omnisci/bioinformatics

Bioinformatics

Start with the biological question, sampling design, assay, raw data shape, and target estimand. Inspect metadata and quality before selecting a pipeline.

Record:

• sample and feature identifiers, groups, covariates, and repeated measures;
• reference assembly/database and version;
• tool versions, parameters, random seeds, and compute environment;
• filtering, normalization, transformations, exclusions, and missingness;
• train/validation/test boundaries where models are involved.

Treat contamination, batch effects, compositionality, multiple testing, population structure, leakage, and pseudoreplication as first-class failure modes. Prefer effect sizes and uncertainty alongside significance. Confirm that the statistical unit matches the biological replicate.

Keep raw data immutable and produce a machine-readable run manifest plus intermediate QC artifacts. Make every derived table traceable to inputs and a command, script, or notebook cell.