Fig. 1: SignalingProfiler 2.0 pipeline.
From: SignalingProfiler 2.0 a network-based approach to bridge multi-omics data to phenotypic hallmarks
A SignalingProfiler 2.0 input consists of multi-omic data collected from perturbed and control conditions (e.g., disease/ treated vs control). B Coverage of SignalingProfiler 2.0 inferable signaling proteins in human and mouse datasets, categorized by molecular function (TF transcription factors, KIN kinases, PP phosphatases, and OTHER other molecular functions). C SignalingProfiler 2.0 final output illustrates the remodeling of the signal, linking user-defined perturbed nodes (optional) with inferred proteins, and ultimately leading to relevant phenotypes. Node activities are coherent with the sign of the edges (red and blue are active and inactive proteins, respectively). Phosphoproteomics is mapped onto edges (validated interactions with phosphoproteomics). D SignalingProfiler 2.0 is a three-step modular pipeline. Step 1 derives the activity of signaling proteins from regulatory phosphosites (PhosphoScore method) and direct transcripts/phosphopeptides using the VIPER algorithm (footprint-based methods)25 Step 2 A user-defined set of perturbed molecules/receptors (e.g., targets of a treatment or mutated genes in a disease) is connected to the inferred proteins using a prior knowledge network (PKN) exploiting: (i) a shortest-path algorithm to reduce the dimension of the PKN to the neighborhood of the inferred proteins (naïve network); (ii) the CARNIVAL optimization strategy19 that retains only the sign-coherent interactions between proteins (context-specific network). Users can provide custom PKNs. Step 3 The context-specific network is connected to cellular phenotypes using the ProxPath algorithm24 and the phenotype activity is obtained by integrating upstream protein activities.
