Figure 3: PCST-based network reconstruction method identifies active sub-modules in KRAS-dependent cell lines.

(a)Network reconstruction methodology. We built a focused undirected and weighted protein-to-protein interaction network (G) using differential expressed pathways identified by the SPIA algorithm²⁹. We assign weights to both nodes (V) and edges (E). Node weights (bv) correspond to the −log P-value of the combined S score (S) for differential abundance between KRAS-Dep and KRAS-Ind phenotypes, whereas the edge’s (Ce) weight corresponds to the experimental confidence of that interaction as reported for the STRING database. Finally, we used the PCST algorithm to find sub-networks, T, in G that maximized the number of differential expressed proteins recovered as well as the confidence in their interaction. (b)Module M1. This module, identified by the PCST, connects LCK and PAK1 in KRAS-Dep cell lines. The module joins LCK and PAK1 with other proteins that belong to the NF-Kappa β and apoptosis pathways such as NFKBIA, NFKBs, TRAFs and BIRCs. Node size is proportional to the absolute value of the combined S score. Node colour represents over-expressed (red) or under-expressed (green) in KRAS-Dep cells. Edge thickness corresponds to edge’s confidence as calculated from STRING database (methods). (c)Module M2. This module, identified by the PCST, involves KRAS and MET in KRAS-Dep cell lines. Additional targetable proteins such as SYK and LYN are also part of this module. Described as in b. (d)Module M3. This module, identified by the PCST, connects CTNNB1 (β-catenin), CTNNA1, CDH1, TJP2 and other proteins associated cell adhesion complexes and the tight junction pathways. Described as in b.