Improve signal peptide prediction by using functional domain information

Signal peptides are significant important in targeting the translocation of integral membrane proteins and secretory proteins. Due the high similarity between the transmembrane helices and signal peptides, classifiers have limit ability to discriminate the signal peptides from the transmembrane helices. To solve this problem, the protein functional domain information is applied in this method. For accurately identify the cleavage sites along the sequence, a subset of potential cleavage sites was firstly screened out by statistical machine learning rules, and then the final unique site was picked out according to its evolution conservation score. This method has been benchmarked on multiple datasets and the experimental results have shown its superiority.