Fig. 1

Overall workflow of Sparse Network-regularized Singular Value Decomposition (SNSVD). SNSVD integrates both a gene expression and a normalized Laplacian matrix L encoding a protein-protein interaction (PPI) network to identify gene functional modules. Based on the output of SNSVD (i.e., sparse singular vectors u and v), we can identify a gene module whose members are from the nonzero elements of u and v. Herein, we show a toy example to explain how SNSVD works. The gene module identified by SNSVD contains four genes (g₁,g₂,g₃,g₄) and five samples (s₁,s₂,s₃,s₄,s₅), where the four genes are correlated across the five samples and the four genes correspond to a dense subnetwork of PPI network