Patient (colored by fitted cluster) Factor-space axes (top 3 of K)
⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.

Multi-Omics Patient Subtyping via Joint Matrix Factorization

A cohort of synthetic patients carries three simulated omics layers — genomics, transcriptomics and proteomics — generated from a shared hidden biology plus realistic per-layer batch noise. A live joint non-negative matrix factorization (Lee–Seung multiplicative updates) fuses all three layers into one shared latent space, and k-means clustering on that space groups patients into candidate disease subtypes, rendered as colored points in 3D. Live readouts track reconstruction error, silhouette score and subtype purity against the hidden ground truth as the factorization converges — the same computational pattern behind real multi-omics cancer-subtyping pipelines.