Omics quality control that reviewers expect

Why reviewers ask about QC first

Omics datasets fail peer review more often on traceability than on novelty. Reviewers expect to see how low-quality reads were removed, how batch structure was assessed, and how every figure maps back to a filtered table.

CERA Medical documents these steps in the analysis plan before compute work begins, then mirrors them in the results package so sponsors can answer methodology questions without reconstructing the pipeline from memory.

Minimum QC documentation

A defensible QC appendix typically includes software names and versions, parameter choices, and the order of filtering steps. Ambiguous phrases such as “standard filtering” are avoided in favour of explicit thresholds.

• Read-level QC with documented cut-offs (quality scores, adapter trimming, duplicate handling)
• Sample-level QC plots retained (depth, complexity, contamination screens where relevant)
• Batch or run-order effects tested and either modelled or ruled out with rationale
• Figure panels keyed to table row identifiers or sample codes agreed in the scope

Batch awareness without overfitting

Batch correction is applied only when diagnostics support it and the study design allows. When correction is inappropriate, that decision is recorded so reviewers understand why raw or partially adjusted models were reported.

Sensitivity analyses—running key contrasts with and without batch terms—are often included so sponsors can show robustness without hiding uncertainty.

Handover that survives audit

Deliverables include normalized count or variant tables, session logs or workflow reports, and README-style notes for non-bioinformatician collaborators. The goal is that a new analyst could reproduce the headline findings from the archived materials within the agreed retention period.