On predicting secondary structure transition

A function of a protein is dependent on its structure; therefore, predicting a protein structure from an amino acid sequence is an active area of research. To improve the accuracy of validation of structures, we are studying the predictability of secondary structure transitions using the following machine learning algorithms: naive Bayes, C4.5 decision tree, and random forest. The annotated data sets from PDB that have agreement with DSSP and STRIDE are used for training and testing. We have demonstrated that predicting structure transition with high degree of certainty is possible and we were able to get as high as 97.5% of prediction accuracy.