To define a unified, confident binding sites for the RNA Binding Proteins (RBPs), we developed a non-negative matrix factorization (NMF) method, RBPgroup.
RBPgroup classifies RBPs and related RNA binding sites into co-binding groups. The idea behind this is that the co-binding of multiple RBPs in a RBP complex (i.e. HNRNP splicing complex) would generate more confident binding sites. To realize this idea, we developed a fine-tuned quality control pipeline followed by NMF, which allowed one RBP grouped into different clusters. We adapted this soft-clustering method because one RBP could co-bind with different co-factors to play different regulatory roles.