Fig. 3

Data-driven exploration of protein signaling network modules in response to PARP1i. Proteins were grouped into functional modules using Netbox algorithm (FDR corrected p-value ≤0.05), which combines prior-knowledge of protein network interactions with a clustering algorithm to identify functional protein modules across the boundaries of curated and pre-defined lists of proteins [28, 29] (Methods). As a next step, protein modules were characterized by assigning module protein members to pathways based on the Reactome database using g:profiler analysis (adj. p-value ≤0.05) (STable2). (a) Depicted are annotated protein modules with > 3 protein members. Nodes represent proteins and are colored based on protein expression change (log2ratio PARP1i/DMSO). Edges represent protein interactions. (b) Heatmap showing annotated protein modules and protein expression changes of corresponding protein members (STable2). Based on their function and protein expression changes, annotated protein modules were evaluated to be associated with PARP1i-induced sensitivity or resistance (STable2)