A Discrete Hidden Markov Model for Detecting Histone Crotonyllysine Sites

Crotonyllysine is a new type of post-translational modifications that is responsible for promoter and enhancer region of gene transcription. Due to little knowledge about its sophisticated mechanism, accurate identification of crotonyllysine still remains challenging. We presented a discrete hidden Markov model to address this problem. We reached a predictive sensitivity of 0.7941 by the leave-one-out cross validation, more than those predicted by the representation-based support vector machine and random forest. The large-scale prediction confirmed most of computer-annotated crotonyllysine sites of five protein sequences in the Uniprot database. We demonstrated that disorder, physicochemical properties and position-specific distribution of amino acids around lysine appeared not to be strongly linked to crotonylation. These results and analysis indicated that it is effective for the presented method to detect crotonyllysine sites. The predicting tool is freely available for academic research at http://yun.baidu.com/share/link?shareid=442733655&uk=1460570570.