| name | bulk-omics |
| description | Knowledge base from curated multi-omics bioinformatics resources. Use when applying multi-omics data analysis, selecting tools for RNA-seq, ChIP-seq, ATAC-seq, Hi-C, DNA-seq, proteomics pipelines, integrating multiple omics layers, cancer genomics analysis, or designing cross-omics workflows. Covers tool combinations, analysis strategies, and practical decision guides. |
Multi-Omics Bulk Data Analysis — Tools, Pipelines & Integration Strategies
Source: Curated from 15 community-maintained GitHub repositories across 8 omics domains | Chapters: 9 | Generated: 2026-07-12
How to Use This Skill
- Without arguments — load this core reference for multi-omics decision guidance
- With an omics type — ask about "RNA-seq", "ChIP-seq", "ATAC-seq", "Hi-C", etc.; I load the relevant chapter
- With an analysis step — ask about "peak calling", "differential expression", "normalization", "integration"
- With a tool name — ask about "DESeq2", "MACS2", "MOFA", "Juicer"; I find the tool and its context
- With a chapter — ask for
ch01 through ch09; I load that chapter and explain details
- Browse — ask "what chapters do you have?" to see the full index
Core Frameworks & Mental Models
The Multi-Omics Decision Framework
1. Choose your omics layers based on the biological question:
- Gene regulation mechanism → RNA-seq + ChIP-seq (target TF/histone) + ATAC-seq
- Disease subtyping → RNA-seq + DNA methylation + CNV + mutations
- 3D regulatory interactions → Hi-C + ChIP-seq + ATAC-seq + RNA-seq
- Genetic variant interpretation → WGS/WES + RNA-seq + epigenomics
- Drug target discovery → RNA-seq + Proteomics + Phosphoproteomics
2. Preprocess each omics layer independently before integration:
- RNA-seq: Salmon/Kallisto → tximport → DESeq2 normalization
- ChIP-seq/ATAC-seq: MACS2 peak calling → IDR → DiffBind
- Hi-C: Juicer/HiC-Pro → .cool conversion → cooltools normalization
- DNA-seq: GATK/DeepVariant → ANNOVAR → VCF filtering
- Batch correction: ComBat-seq (RNA), MultiBaC (cross-omics)
3. Select integration strategy by goal:
- Unsupervised exploration → MOFA (factor analysis for 2-10 omics types)
- Biomarker discovery with phenotype → DIABLO/mixOmics (supervised multi-block)
- Patient stratification → SNF + consensus clustering → clinical correlation
- Two-layer correlation → sparse CCA / SmCCNet
- Survival prediction → DeepProg / Flexynesis (deep learning)
4. Always use data containers designed for multi-omics:
- R: MultiAssayExperiment (Bioconductor standard)
- Python: muon or Omilayers
The Tool Selection Principle
When choosing among competing tools, prefer: widely-cited > recently-published, with-benchmarks > without-benchmarks, actively-maintained > abandoned. For multi-omics specifically, prefer methods that have been benchmarked in multi-omics comparison studies (Rappoport 2018, Cantini 2020, Herrmann 2020).
The Normalization Principle
"Assumptions matter more than performance." Every normalization method assumes something about the data: most genes don't change (TMM, DESeq2 median-of-ratios), global differences exist (quantile, loess), or spike-ins provide truth (RUV). Know what you're assuming before you normalize. This applies to ALL omics types, not just RNA-seq.
The Batch Effect Precaution
"Always visualize your data by batch before analysis." If samples cluster by processing date, technician, or sequencing run rather than by biological condition, batch correction is mandatory — not optional. Use ComBat-seq for counts, SVA/RUVseq for latent factors, MultiBaC for cross-omics batch.
The Integration Hierarchy
"Integrate at the appropriate level." Not all integration methods are equal:
- Early integration: Concatenate features → single model (simple but ignores data heterogeneity)
- Intermediate integration: Joint dimension reduction → shared latent space (MOFA, JIVE, iCluster)
- Late integration: Per-omics models → ensemble prediction (SNF, DeepProg ensemble)
- Hierarchical integration: Domain knowledge guides the integration structure (RGCCA, multi-block PLS)
Chapter Index
| # | Title | Key Topics |
|---|
| ch01 | RNA-seq Analysis Ecosystem | Quantification, normalization, DESeq2/edgeR/limma, alignment-free, batch correction, databases |
| ch02 | ChIP-seq Analysis Ecosystem | Peak calling (MACS2/SICER), IDR, differential binding, motif analysis, super-enhancers, CUT&RUN/CUT&Tag |
| ch03 | ATAC-seq Analysis Workflow | ENCODE pipeline, csaw/differential accessibility, footprinting, HMMRATAC, nucleosome positioning |
| ch04 | Hi-C & 3D Genomics Tools | Juicer/HiC-Pro/cooltools, TAD/loop calling, differential interactions, Capture Hi-C, HiChIP |
| ch05 | DNA-seq: Variant Calling & Methylation | GATK/DeepVariant, somatic/germline, CNV/SV, mutational signatures, GWAS, methylation analysis |
| ch06 | Proteomics Tools & Resources | MaxQuant/FragPipe, DDA/DIA, DEP differential analysis, CPTAC |
| ch07 | Cancer Genomics & TCGA Analysis | Survival analysis, driver genes, mutational signatures, clonal architecture, cancer databases |
| ch08 | Multi-omics Integration Strategies | MOFA, mixOmics/DIABLO, SNF, JIVE, CCA, autoencoders, method selection guide |
| ch09 | Cross-omics Workflows & Tool Combinations | 5 end-to-end workflow templates, tool pairing matrices, batch effect correction |
Topic Index
- ATAC-seq → ch03, ch02 (footprinting), ch09 (workflow)
- Batch correction → ch01 (ComBat-seq), ch08 (MultiBaC), ch09 (cross-omics)
- Cancer genomics → ch07 (TCGA, survival), ch05 (somatic variants), ch08 (subtyping)
- Capture Hi-C → ch04 (CHiCAGO, Peaky), ch02 (CHiCAGO protocol)
- ChIP-seq → ch02 (core), ch09 (RNA-seq integration)
- Chromatin state → ch02 (ChromHMM, Segway)
- CNV → ch05 (CNVnator, GISTIC), ch04 (CNV-aware Hi-C)
- CUT&RUN/CUT&Tag → ch02 (protocols, SEACR)
- Deep learning → ch08 (autoencoders), ch04 (Hi-C enhancement), ch05 (DeepVariant)
- Differential analysis → ch01 (RNA-seq), ch02 (ChIP-seq), ch03 (ATAC-seq), ch04 (Hi-C)
- eQTL → ch05 (MatrixeQTL), ch07 (TCGA), ch09 (workflow)
- Footprinting → ch03 (HINT-ATAC, TOBIAS), ch02 (msCentipede)
- GWAS → ch05 (pipelines, annotation, FUMA)
- Hi-C → ch04 (pipelines, tools, formats)
- HiChIP → ch04 (MAPS, hichipper, FitHiChIP)
- Integration methods → ch08 (comprehensive catalog), ch09 (workflows)
- Motif analysis → ch02 (HOMER, MEME, gimmemotifs)
- MultiAssayExperiment → ch08 (R container), ch09 (usage)
- Mutational signatures → ch05 (deconstructSigs, MutationalPatterns), ch07 (COSMIC)
- Normalization → ch01 (RNA-seq), ch04 (Hi-C), ch02 (cross-sample)
- Peak calling → ch02 (MACS2, SICER, epic2), ch03 (ATAC-seq specific)
- Proteomics → ch06 (MaxQuant, DIA-NN, DEP)
- RNA-seq → ch01 (core), ch09 (integration workflows)
- SNF → ch08 (method), ch07 (cancer subtyping), ch09 (workflow)
- SNP/variant calling → ch05 (GATK, DeepVariant, Strelka2)
- Super-enhancer → ch02 (ROSE, HOMER, CREAM)
- Survival analysis → ch07 (TCGA pipeline, GEPIA)
- TADs → ch04 (Arrowhead, TopDom, Armatus, insulation score)
- TCGA → ch07 (access, analysis, tools)
Supporting Files
- glossary.md — 50 key terms with precise definitions and chapter references
- patterns.md — 12 reusable analysis patterns with when/how/trade-offs
- cheatsheet.md — Decision tables, thresholds, tells & smells for rapid reference
Scope & Limits
This skill covers bulk omics data analysis tools, pipelines, and integration strategies from 15 curated community repositories. Focus is on practical tool selection, workflow design, and multi-omics integration. For single-cell omics, see dedicated single-cell resources (referenced in ch01). For specific protocol details, consult original tool documentation and publications.
