Fig. 1

a Workflow used for the identification of protein complex variants from large-scale proteomics dataset. Protein complex definitions were assembled from different resources and the literature (Additional file 2, Methods). Protein complex-based normalization [10] was used to investigate compositional changes in two published datasets: a time course experiment of fibroblast reprogramming to induced pluripotent stem cells [12] and an 11 cancer cell line dataset [13]. b The abundance of members of protein complexes is correlated across proteome profiles. A total of 824 proteins from 123 complexes were quantified in the 11 cell lines dataset [13]. Distributions of Pearson correlation coefficients were plotted for pairwise comparisons between members of the same complex (light blue) and between proteins from different complexes (orange). c Similar to b, except co-expression between members of protein complexes was calculated using gene expression profiles derived from 10 cancer cell lines (missing GAMG cell line from Geiger et al. [13]). Co-expression analysis was limited to the set of proteins from b that were also quantified in the microarray experiment (HG-U133A, Additional file 3). Pearson correlation coefficients were calculated between pairs of 796 proteins from 117 complexes. Members of protein complexes show higher co-expression at the mRNA level, but to a much lesser extent than the protein co-expression