SUMO Substrates and Sites Prediction Combining Pattern Recognition and Phylogenetic Conservation

Motivation: Small Ubiquitin-related modifier (SUMO) proteins are widely expressed in eukaryotic cells, which are reversibly coupled to their substrates by motif recognition, called sumoylation. Two interesting questions are 1) how many potential SUMO substrates may be included in mammalian proteomes, such as human and mouse, 2) and given a SUMO substrate, can we recognize its sumoylation sites? To answer these two questions, previous prediction systems of SUMO substrates mainly adopted the pattern recognit ion methods, which could get high sensitivity with relatively too many potential false positives. So we use phylogenetic conservation between mouse and human to reduce the number of potential false positives. Results: We first use two major patterns of a potential SUMO substrate: an NLS (Nuclear Localization Signal) and a consensus motif ψ-K-X-E, where ψ meant a hydrophobic amino acid. So we followed a simple rule to predict the SUMO substrates: the sub-cellular localization of a given protein must be predicted as nuclear by PSORT II, and the protein must have at least one consensus motif ψ(A, F, I, L, M, V, W)-K-X-E. After the above methods, there are still too many predicted positives. To eliminate the potential false positives, we used the orthology information between mouse and human, and followed the rule that at least one consensus motif should be at the same position after sequence alignment of the ortholog pair. We got 2,683 potential SUMO substrates in both mouse and human, with 58 out of 79 known SUMO substrates in human predicted correctly. For the sumoylation sites prediction, our method got nearly the same sensitivity as the existed tool SUMOplot, with 42 against 42 (high probability) or 44 (all) true positives of 54 known sumoylation sites respectively. But our method outperformed SUMOplot significantly by specificity, with 74 against 152 (high probability) or 324 (all) predicted sumoylation sites respectively. So our method greatly reduced the number of potential false positives while still kept a satisfying sensitivity. Availability: The software SSP (SUMO Substrate and Site Prediction) 1.0 for Windows system, written by Delphi, is available from http://973-proteinweb.ustc.edu.cn/sumo/. Contact: [email protected], [email protected]