Fig. 3: Co-regulation analysis maps cooperative protein associations to known protein complexes and pathways.

a, Scatterplots comparing abundances across selected protein pairs across samples. b, Distribution of r_Pearson based on observed (top) and permuted (bottom) data. The observed distribution was obtained by calculating r_Pearson across all possible protein–protein pairs. Permuted distributions were generated by randomly sampling 50,000 protein pairs after randomly shuffling their respective timepoints ten times each before calculating r_Pearson. Colours indicate strongly correlated (≥0.95; blue) or anticorrelated (≤−0.95; red) edges. c, Distribution of protein edge counts across the trimmed correlation network. On average, each protein in the network participated in 174 edges ± 195 edges. d, Ratio of enrichment for the annotated edges in the correlation network (‘observed network’) compared to the expected edge annotation frequencies across Gene Ontology biological process (GOBP), cellular component (GOCC), molecular function (GOMF), localization, pathways, protein–protein interactions (PPIs, BioPlex) or protein complexes. Specifically, we calculated the enrichment for annotated edges as the fraction of annotated edges per category in the observed correlation network divided by the fraction of annotated edges among all possible edges involving the 5,227 proteins in the correlation network. The expected frequency of annotated edges was calculated by generating all possible pairs from 5,227 human proteins (UniProt, July 2024) and computing the number of pairs explained by each functional category. e, Network analysis identifies known associations between proteins for BMP1 and RPL7A. f,g, Network structure of the 26S proteasome (f) and citric acid cycle pathway (g). Magenta nodes indicate known complex members annotated either from CORUM or EMBL ComplexPortal for protein complexes, or from BioCarta, KEGG, the Protein Interaction Database (PID), Reactome and WikiPathways (WP) for biochemical pathways. Blue edges indicate positive correlations between nodes, and red edges indicate anticorrelations. h,i, Cooperative proteins are highly correlated with members of established protein complexes, including the NuA4 chromatin remodelling complex (h) and the Chaperonin-containing T (TRiC/CCT) complex (i). Magenta nodes indicate subunits of a given complex, and orange nodes indicate cooperative proteins, that is, proteins with profiles correlated to proteins constituting a particular protein complex. Cooperative node sizes indicate the negative log₁₀ of the BH-adjusted P value after computing significance from a one-sided Fisher’s exact test to determine the cooperative association of a protein to a particular module. Blue edges indicate correlated edges, and orange edges link cooperative proteins to members of a particular module. j, Bioplex interaction network of the TRiC/CCT complex. Orange nodes are cooperative proteins with profiles correlated to proteins found in the TRiC/CCT complex. Grey edges indicate BioPlex evidence. k, Histogram of protein complexes (x axis) and their respective numbers of cooperative proteins (y axis). l, Heuristic to identify shared cooperative proteins between complexes. m, Heatmap depicting a subset of shared cooperative proteins across manually curated EMBL ComplexPortal protein complexes, namely exosomes, SWI/SNFs, ATAC remodellers, nucleosome remodellers (NuRDs) and histone acetyltransferase (HAT) and deacetylase (HDAC) complexes. The heatmap is coloured by Jaccard similarity coefficients calculated from overlapping sets of cooperative proteins between protein complex pairs, and clustered using Euclidean distances with average linkage.