On preprosessing of protein sequences for neural network prediction of polyproline type II secondary structure

Motivation: Polyproline type II stretches are rather rare among proteins, and, therefore, it is a very challenging task to try to find them computationally. In the present study our aim was to consider especially the preprocessing phase, which is important for any machine learning method. Preprocessing includes selection of relevant data from Protein Data Bank and investigation of learnability properties. These properties show whether the material is suitable for neural network computing. In addition, algorithms in connection with data selection and other preprocessing steps were considered. Results: We found that feedforward perceptron neural networks were appropriate for the prediction of polyproline type II as well as relatively effective in this task. The problem is very difficult because of high similarity of the two classes present in the classification. Still neural networks were able to recognize and predict about 75 % of secondary structures. Contact: [email protected]