Identifying Human SIRT1 Substrates by Integrating Heterogeneous Information from Various Sources

Most proteins undergo different kinds of modification after translation. Protein acetylation is one of the most crucial post-translational modifications, which causes direct or indirect impact on various biological activities in vivo. As a member of Class III HDACs, SIRT1 was the closest one to the yeast sir2 and drew most attention, while a small number of known SIRT1 substrates caused difficulties to clarify its function. In this work, we designed a novel computational method to screen SIRT1 substrates based on manually collected data and Support Vector Machines (SVMs). Unlike other approaches, we took both primary sequence and protein functional features into consideration. Through integrating functional features, the Matthews correlation coefficient (MCC) for the prediction increased from 0.10 to 0.65. The prediction results were verified by independent dataset and biological experiments. The validation results demostrated that our classifier could effectively identify SIRT1 substrates and filter appropriate candidates for further research. Furthermore, we provide online tool to support SIRT1 substrates prediction, which is freely available at http://bioinfo.bjmu.edu.cn/huac/.