Enhancing Protein Disorder Detection By Refined Secondary Structure Prediction

More and more proteins have been observed to display functions through intrinsic disorder. Such structurally flexible regions are shown to play important roles in biological processes and are estimated to be abundant in eukaryotic proteomes. Previous studies largely use evolutionary information and combinations of physicochemical properties of amino acids to detect disordered regions from primary sequences. In our recent work DisPSSMP, it is demonstrated that the accuracy of protein disorder prediction is greatly improved if the disorder propensity of amino acids is considered when generating the condensed PSSM features. This work aims to investigate how the information of secondary structure can be incorporated in DisPSSMP to enhance the predicting power. We propose a new representation of secondary structure information and compare it with three naive representations that have been discussed or employed in some related works. The experimental results reveal that the refined information from secondary structure prediction is of benefit to this problem.