KMAL-SP

As a newly discovered PTM through mass spectrometry and protein sequence database searching, lysine malonylation (Kmal) is evolutionarily conserved in both bacterial and mammalian cells. Kmal serve as a chemical reaction to one or more enzymes that catalyze the transfer of malonyl groups from malonyl-CoA to lysine residues, and therefore plays a key role in regulating protein functions. In addition, recent study has initially demonstrates that the enrichment of malonylated proteins have an influence on metabolic pathways, especially those involved in glucose and fatty acid metabolism.

Indeed, we review, analyze and compare 11 different feature encoding methods, with the purpose to extract key patterns and characteristics from residues sequences of Kmal sites. Using the optimal feature sets obtained afterwards, four commonly used (Random Forest (RF), Support Vector Machines (SVM) and K-Nearest Neighbor (KNN)) and one recently proposed (Light Gradient Boosting Machine, LightGBM) machine learning methods are trained and compared in three species, namely Escherichia coli, Mus musculus and Homo sapiens on randomized 10-fold cross-validation tests. Based on the optimal ensemble models, we develop an accessible online predictor, kmal-sp. We hope that this comprehensive survey work and the proposed strategy for building more accurate models can serve as a useful guide for inspiring future development of new computational methods for PTM site prediction, expedite the discovery of new malonylation and other PTM types, and facilitate the hypothesis-driven experimental validation.