Combining random forest with multi-amino acid features to identify protein palmitoylation sites

• Obtained satisfaction results by using seven amino acid properties
• Accurate prediction of palmitoylation sites by using random forests
• Free online service for other researchers has been provided.

As one of the most important and ubiquitous post-translational lipid modifications, protein palmitoylation plays significant roles in a variety of biological processes, including signaling, neuronal transmission, and membrane trafficking. Protein palmitoylation is a highly dynamic process, which regulates various protein functions. The dynamic nature of palmitoylation makes it very difficult to identify such kind modification by experimental assay methods. Therefore, using computational approaches to identify palmitoylation sites is of highly important. In this study, a new method was proposed to predict palmitoylation sites based on multi-amino acid properties and random forest (RF) algorithm. The prediction accuracy, sensitivity, specificity, Matthews correlation coefficient (MCC) and area under the curve values (AUC) for current method were 91.85%, 88.89%, 94.67%, 0.8377, and 0.9595, respectively. These results indicated that the current method was a powerful and effective tool for identifying palmitoylation sites, which would be a complement to protein palmitoylation research. Furthermore, a free online service was established in http://sysbio.yznu.cn/Research/RandomForcast.aspx.