Place: VIVA Room, Level 7, School of Computer Sciences
Predicting secondary structure from a protein sequence provides an important role in the prediction of protein three-dimensional structure and function. The prediction accuracy of the secondary structure has been stagnant for many years largely due to the challenge of considering non-local interactions between amino acid residues in the protein sequence. Recent advances in computational techniques along with the increased availability of solved protein structures have attracted many researchers to develop secondary structure predictors. Deep learning is a rapidly evolving field which excels at problems where there are complex relationships between input features and desired outputs. More powerful deep learning methods with improved capability of capturing long-range interactions have become popular tools for machine learning. In this work, we propose a deep learning based method to predict secondary structure. We aim to utilize deep learning neural network architecture to predict secondary structure by using physiochemical and evolutionary information. The performance of our method will be evaluated in terms of three-state prediction accuracy using the most common data sets used by several researchers to develop secondary structure prediction methods. The proposed method is expected to significantly improve the overall secondary structure prediction accuracy.