Predicting Intrinsically Disordered Proteins Based on Different Feature Teams

The characteristics of intrinsically disordered proteins depend on their length. An obvious fact is that the composition of amino acid sequences is different for different length disordered regions. In order to improve the performance of the predicting model, a new method was proposed to predict disordered regions of diverse length disordered regions in proteins by using different feature teams. Taking into account the relevance between their characteristics and length of intrinsically disordered proteins, different feature teams were constructed for different length disordered regions. In every feature team, the selection of window sizes and features could meet the demand of the corresponding length disordered region. Comparing with the traditional method, this method could consider not only the influence of the window sizes but also the effect of the feature information. According to every feature team, a basic predictor was required to built by SVM. By integrating these basic predictors, the final decision could be made by the majority voting method. Subsequent simulation suggests that the proposed method can consider the information from the long and short disordered regions simultaneously and get a good predicting accuracy for IDPs, especially for short disordered regions.