RS-ExpNet-CRNMF

RS-ExpNet-CRNMF applies a robust and sparse co-regularized nonnegative matrix factorization (NMF) framework to integrate mRNA expression patterns and protein-protein interaction networks for identification of mutated driver genes from somatic mutation data.

Key Features:

• Integration of Prior Information: Incorporates protein-protein interaction network data and mRNA expression patterns as complementary prior information.
• Robust and Sparse Co-Regularization: Employs co-regularized NMF with Frobenius norm regularization and sparsity-inducing penalties to mitigate overfitting and enforce sparse representations.
• Top-Scoring Gene Identification: Produces sparse gene representations that enable selection of top-scoring genes as candidate driver genes.
• Performance Enhancement: Recovers known benchmarking genes and detects novel driver genes, showing improved performance relative to existing network-based methods.

Scientific Applications:

• Cancer driver gene discovery: Identification and prioritization of mutated driver genes in cancer genomics using integrated expression and interaction data.
• Low-frequency driver detection: Detection of low-frequency mutated driver genes from somatic mutation data that are difficult to find with traditional approaches.

Methodology:

Co-regularized nonnegative matrix factorization integrating interaction network and mRNA expression data, using Frobenius norm regularization and sparsity-inducing penalties to produce sparse gene representations and rank top-scoring genes.