Prediction of protein networks through data integration

The document discusses methods for predicting protein-protein interaction networks through integrating diverse data sources. It describes the STRING database which predicts interactions between proteins in 373 genomes using genomic context methods, co-expression data, experiments, and literature mining. It also discusses NetworKIN, a method that predicts phosphorylation sites and potential kinase-substrate relationships through integrating phosphoproteomics data, sequence motifs, and network context. Benchmarking shows NetworKIN can predict interactions with over 2.5-fold greater accuracy compared to sequence-based methods alone by incorporating network context.