What problem does it solve?
Fuse disparate omics-derived patient similarity information into a single, robust network to reveal clinically meaningful patient subtypes and improve downstream clustering and survival analysis. This reduces manual integration work and leverages complementary signals across transcriptomics, methylation, proteomics, and clinical data to strengthen subtype discovery.
Core Features & Use Cases
- Similarity Network Fusion: Compute per-omics distance matrices, convert to affinity matrices, and iteratively fuse them with SNFtool to produce a unified patient similarity network.
- Clustering & Evaluation: Estimate optimal cluster counts, run spectral clustering, compute NMI against known labels, and perform grid search over K and alpha for hyperparameter tuning.
- Clinical Integration & Validation: Add clinical variables as an additional view, rank features by cluster association, and evaluate clinical relevance with Kaplan-Meier survival analysis and log-rank tests.
- Visualization: Export fused matrices to igraph for network plots and heatmaps with cluster annotations to inspect structure and cluster separation.
- Use Case Example: Integrate RNA-seq, methylation, and miRNA profiles from a cancer cohort to identify molecular subtypes and test for survival differences.
Quick Start
Fuse my RNA-seq, methylation, and miRNA sample matrices to identify patient clusters, tune SNF hyperparameters, and evaluate survival differences.
