MultiOmicsSurvey

A comprehensive survey of computational methods for multi-omic integration in cancer research.

Unsupervised multi-omic integration algorithms benchmarked across 5 data modalities on TCGA breast cancer, validated on TCAG holdout set and the transNEO cohort.

---

Quick Start

# Clone and restore environment
git clone https://github.com/sionaris/MultiOmicsSurvey.git
renv::restore()

Study data can be found in the paper supplement and Zenodo

⚠️ renv notes — Packages not in lockfile

The following packages are not included in renv.lock and must be installed manually if needed:

Package	Reason	Required for
peakRAM	HPC-only dependency	Benchmark memory profiling
Rmosek	Requires MOSEK license	wMKL, KLIC
wMKL	Custom Bioconductor install with C++ code edit	wMKL algorithm
devtools	Development tool	Package installation from GitHub

---

Directory Structure

📁 Scripts — Analysis code

1. automated_scripts/ — Helper functions (sourced by other scripts)
2. Download_TCGA_data.R — Downloads and preprocesses TCGA-BRCA
3. transNEO_pseudocount_determination.R — RNA-seq normalisation for transNEO
4. MOVICS/ — Baseline MOVICS analysis
5. single_algorithm/ — Individual method runs (+HPC scripts)
6. method_comparisons/ — Cross-method comparisons & benchmarks
7. Survival analysis/ — Kaplan-Meier and Cox regression
8. Consensus/ — Consensus clustering pipelines

📁 Python — Python-based methods

• MOFA/ — Multi-Omics Factor Analysis
• MONET/ — Multi Omic clustering by Non-Exhaustive Types
• MSNE/ — Multiple Similarity Network Embedding

📁 Resources — Supporting data

• HPC output/ — Results from cluster computing runs (individual runs)
• Pathways/ — Gene sets for enrichment analysis
• TCGA/ — Clinical data for TCGA
• transNEO/ — Validation cohort (paper)
• Performance/ — Benchmark input data

📁 Results — Output files

• single_algorithm/ — Individual method outputs
• Comparisons/ — Cross-method comparison plots
• Performance_benchmarks/ — Benchmark experiments (robustness, stability, scalability)
• Consensus/ — Consensus clustering results
• Survival_evaluations/ — Survival analysis outputs

📁 docs — Documentation

• R/function_documentation.pdf — Core function reference
• R/benchmark_function_documentation.pdf — Benchmark analysis functions
• Python/function_documentation.pdf — Custom Python functions for MONET and MSNE

---

Methods Included

Category	Algorithms
Similarity Network	ab-SNF, ANF, MDICC, MSNE, NEMO, RWR-F, RWR-NF, SNF, Spectrum
Multiple Kernel Learning	CIMLR, KLIC, wMKL
Matrix Factorisation	MOFA, LRAcluster, MFA
Graph-based	MONET
Bayesian	iClusterBayes
Consensus	COCA

---

Shiny app to explore our results interactively

---

Citation

Publication forthcoming

---

License

Apache 2.0