Optimization based Long Short Term Memory Network for Protein Structure Prediction

Abstract

One of the challenging tasks in computational biology is the anticipation of protein secondary structure (PSS) from amino acid sequences. Numerous computational and statistical methods are used for this purpose. With the growing attention of deep learning, models such as convolutional neural network and recurrent neural network are also used for this prediction. But, these strategies require a lot of hyperparameters tuning to accomplish the best outcome. In this paper, we proposed a bidirectional embedded recurrent deep neural system using long short term memory (LSTM) cells with continuous coin betting optimizer (COCOB) to tune the hyperparameters for the prediction of PSS. We have performed this experiment on Nvidia DGX station. We assessed our model on a FASTA-formatted file which consists of Protein Data Bank (PDB) sequences and their relative secondary structure. We report better performance (Q3=79.01% and Q8=82.38%) than best in class (Q3=64.9% and Q8=68.2%) methods.

Keywords

• proteomic
• bioinformatics
• deep learning
• protein secondary structure
• cocob