Supplementary Figure 14: Proteins that correlate with PTM substrates share functional properties.

a) Expression level (NSAF) profiles for 15,747 proteins spanning 30 tissues were correlated with averaged PTM profiles across the same tissues, using either estimated stoichiometry or normalized spectral counts (NSC). The representative scatter plot shown here for lysine hydroxylation indicates the extent to which each protein’s tissue profile (points) correlates with lysine hydroxylation across the 30 tissues as measured by estimated stoichiometry (x-axis) or total abundance (y-axis). These data show that the two PTM quantification methods are broadly similar. However, protein correlation (Pearson) ranks may differ greatly between the two quantification methods. Thus, both can produce complementary but similar sets of highly correlated proteins. b) Protein-PTM correlations generally did not indicate specific modified substrates. A protein’s abundance could correlate (Pearson) with a particular PTM because it regulates or directly catalyzes the PTM’s formation on its substrate. Alternatively, proteins could be correlated with a modification because they are themselves heavily-modified substrates of the PTM. Kinases, which both catalyze phosphorylation events and are themselves highly phosphorylated, would be expected to be examples of both conditions, for example. By contrast, collagens would be examples of the latter condition, as abundant proteins in certain tissues that carry a highly degree of hydroxylated prolines. To evaluate these possibilities, we first identified the 20 proteins that most highly correlated with each of the 28 PTMs shown here, as computed using either modification NSC or estimated stoichiometry. Of these, we plotted the number of proteins that were also modified by the indicated PTM. For the most part, however, PTMs were not identified on the same proteins to which they were most highly correlated, suggesting that they may be candidate regulators of PTM transfer. c) Enriched gene ontologies (p values calculated from one-tail Fisher exact test, Benjamini-Hochberg corrected) for the top fifty most correlated proteins for several PTMs suggests either enzymatic activity (that is, oxidoreductase activity is known to be required for lysine hydroxylation to occur) or common functional activity (that is, arginine dimethylation is known to be enriched in RNA splicing proteins, Fig. 4b). As demonstrated in part b, these proteins are not themselves substrates of the PTM of interest. Thus, these ontologies further suggest functional relationships between PTMs and proteins which are highly correlated with them.