| name | research-analysis |
| description | Route downstream research analysis for shotgun metagenomics: QC reports (MultiQC), diversity/figures (visualization, microeco, phyloseq, STAMP), differential abundance (MaAsLin2/3, ANCOM-BC, LEfSe), association networks (SparCC, SPIEC-EASI), and workflow managers (Nextflow, Snakemake). Use after taxonomy/function tables exist. Child skills: visualization, microeco, phyloseq, maaslin2, ancombc, lefse, stamp, sparcc, spieceasi, multiqc, nextflow, snakemake.
|
| license | MIT |
| category | orchestration |
| tags | ["research-analysis","statistics","visualization","differential-abundance","network","Nextflow","Snakemake"] |
| stage | report |
Research analysis (stats · figures · reports · workflows)
Downstream scientific delivery after profiling/annotation tables exist:
QC dashboards, diversity/composition figures, differential abundance, microbial
networks, and reproducible pipeline wrappers. Does not replace mag-qc.
Analytical thinking
| Claim / deliverable | Prefer | Do not treat as |
|---|
| Multi-tool QC HTML | multiqc | Biological DA |
| Modular R paper figures | microeco | Causal mechanism |
| Classic R data class / ggplot2 | phyloseq | Sole DA method |
| Multivariable associations | maaslin2 | Unadjusted LEfSe alone |
| Compositional bias-corrected DA | ancombc | Network edges |
| Classic LDA biomarkers | lefse | Modern multivariable default |
| GUI exploratory profile tests | stamp | Covariate-heavy models |
| Compositional correlations | sparcc | Direct interactions |
| Sparse conditional networks | spieceasi | DA between groups |
| Figure checklist by stage | visualization | Table generation |
| Portable pipeline (containers) | nextflow | Statistics Methods |
| Rule-based DAG (conda-centric) | snakemake | Statistics Methods |
Hard rules:
- Every figure: stage + method + DB/software version + normalization
- Inspect batch / library size before group tests
- Report rarefaction (if used), transform, test, and FDR/FWER
- Do not mix MetaPhlAn, Kraken, and MAG abundances unlabeled
- Network edges and DA hits are associations — not validated biology
- Workflow managers still must pin each process tool + database
Decision tree
Research downstream need?
│
├─ Aggregate QC logs → multiqc
├─ Diversity / composition figures
│ ├─ Modular R6 workflow → microeco
│ ├─ phyloseq object + classic API → phyloseq
│ └─ Stage checklist → visualization
├─ Differential abundance
│ ├─ Covariates / longitudinal → maaslin2
│ ├─ Sampling-fraction bias correction → ancombc
│ ├─ Classic LDA → lefse
│ └─ GUI exploratory → stamp
├─ Association networks
│ ├─ SparCC correlations → sparcc
│ └─ SPIEC-EASI graphs → spieceasi
├─ Reproducible pipelines
│ ├─ Nextflow / nf-core → nextflow
│ └─ Snakemake → snakemake
├─ Interactive MAG UI → anvio
└─ Unsure → this hub
Typical order
study-design → (± nextflow|snakemake wrappers)
→ QC (+ multiqc)
→ taxonomy / function tables
→ phyloseq|microeco explore
→ maaslin2|ancombc (± lefse|stamp)
→ (± sparcc|spieceasi)
→ Methods: pin versions, transforms, FDR
Related skills
visualization · microeco · phyloseq · maaslin2 · ancombc · lefse ·
stamp · sparcc · spieceasi · multiqc · nextflow · snakemake ·
study-design · setup · metaphlan · kraken2 · humann · anvio ·
metagenomics-workflow · tool-selection
