PSSM-PROREP: A Flexible Web Server for Generating Various Position Specific Score Matrix-derived Protein Representations

With the number of protein sequences growing exponentially in the post-genome era, machine learning methods that sufficiently extract protein sequence information have been introduced for predicting protein functions on a large scale. Having protein representations with some biological meanings can depict the essential correlation with protein functions, which offers the most effective strategy to enhance the prediction accuracy. Position Specific Scoring Matrix (PSSM) has been commonly applied to convert protein sequences with changeable lengths into feature vectors with a specified length while preserving evolutionary information to the utmost extent. However, there’s currently no web-server providing PSSM-derived protein representations for public usage, thereby possibly decreasing research efficiency. To get rid of this situation, we develop a web server called PSSM-PROREP(at http://prorep.weka.cc) for the first time to calculate the extensively used PSSM-derived protein representations effectively and flexibly. Researchers can evaluate the prediction performance of different protein representations calculated by PSSM-PROREP, and then select optimal feature sets to construct more robust protein function prediction systems. It is expected that PSSM-PROREP may become a practical tool in protein sequence analysis and protein function predictions.