Omics Workflow
Intake contract
Identify assay, platform, experimental unit, biological replicates, groups, batches, pairing, reference/annotation version, raw/processed inputs, and primary biological contrast. Stop if group and batch are fully confounded or replication cannot support inference.
Stage gates
Raw QC: integrity, depth, quality, contamination, adapters, duplication, mapping/assignment expectations.
Metadata: unique sample IDs, factor levels, units, missingness, batch and subject structure.
Quantification: versioned reference, parameters, multi-mapping policy, feature definition.
Exploration: sample-level PCA/MDS, library size, outliers, and batch patterns without deleting samples silently.
Inference: explicit design matrix and contrasts; multiple-testing control; effect sizes with uncertainty.
Interpretation: correct background universe, annotation version, redundancy control, and evidence-level language.
Delivery: source tables, code, environment, QC report, exclusions, and failed checks.
Domain red lines
- Do not treat cells as biological replicates; use sample-aware inference or pseudobulk where appropriate.
- Treat microbiome abundance as compositional and evaluate sparsity and prevalence.
- Do not infer mechanism from enrichment, network centrality, classifier importance, or differential abundance alone.
- Never overwrite raw data or package user data into a reusable plugin.
1---2name: omics-workflow3description: Plan, audit, or interpret public or user-authorized sequencing and omics workflows, including bulk RNA-seq, single-cell, amplicon, metagenomic, variant, and metabolomic analyses. Use only after design and metadata fit are checked.4metadata:5 source: https://skillmd.com/skills/minimax-ai/omics-workflow6---78# Omics Workflow910## Intake contract1112Identify assay, platform, experimental unit, biological replicates, groups, batches, pairing, reference/annotation version, raw/processed inputs, and primary biological contrast. Stop if group and batch are fully confounded or replication cannot support inference.1314## Stage gates15161. `Raw QC`: integrity, depth, quality, contamination, adapters, duplication, mapping/assignment expectations.172. `Metadata`: unique sample IDs, factor levels, units, missingness, batch and subject structure.183. `Quantification`: versioned reference, parameters, multi-mapping policy, feature definition.194. `Exploration`: sample-level PCA/MDS, library size, outliers, and batch patterns without deleting samples silently.205. `Inference`: explicit design matrix and contrasts; multiple-testing control; effect sizes with uncertainty.216. `Interpretation`: correct background universe, annotation version, redundancy control, and evidence-level language.227. `Delivery`: source tables, code, environment, QC report, exclusions, and failed checks.2324## Domain red lines2526- Do not treat cells as biological replicates; use sample-aware inference or pseudobulk where appropriate.27- Treat microbiome abundance as compositional and evaluate sparsity and prevalence.28- Do not infer mechanism from enrichment, network centrality, classifier importance, or differential abundance alone.29- Never overwrite raw data or package user data into a reusable plugin.