2.5. Construction of the Human Protein–Protein Interaction Network

A network of human protein–protein interactions was constructed based on curating high confidence, experimentally verified interactions from multiple protein–protein interaction sources and pathway databases, as well as from primary literature. This network is termed as human protein–protein interaction network (hPPiN) (Sambarey et al., in press). Briefly, the Search Tool for The Retrieval of Interacting Genes/Proteins (STRING) version 10 (Szklarczyk et al., 2014) was mined to extract all human interactions with a combined score > 900, and the functional nature of these interactions was identified from the protein actions file. Based on these annotations, the edges were assigned directions, however, interactions describing a ‘binding’ event were represented as bidirectional edges. SignaLink v 2.0 (Fazekas et al., 2013) was mined to identify regulatory interactions of transcriptional, post-transcriptional and pathway regulators. Additional interactions present in non-disease conditions were identified from the Cancer Cell Map (Krogan et al., 2015), and the BioGRID database (Chatr-Aryamontri et al., 2015) was mined to identify unique interactions not reported by the other resources used. In addition to PPI databases and resources, primary literature was explored to identify experimentally verified interacting proteins in the human proteome. Interactions were extracted from Multinet (Khurana et al., 2013), and from the macrophage interaction network (Sambarey et al., 2013). From this constructed hPPiN, interactions were extracted for all protein-coding genes present in the RNA-Seq dataset.