Data-driven learning models for protein folding analysis

Abstract

In this paper, the protein folding prediction problem is modelled and solved using reinforcement learning. Deep learning methods have been commonly used to predict protein structure and folding in recent years. However, it has been found that this method may not take into account the energy function and hydrophobic-polar nature of proteins. This paper attempted to use the Q-learning approach using ℇ-greedy policy to predict the secondary structure of protein given its amino acid sequence and optimal energy. This paper can be improved by incorporating pretraining of the agent or using more advanced deep Q-learning algorithm.