Novel Protein Structure Prediction Model Using Fused Pipit Adapted Deep Convolutional Neural Network Classifier

Abstract

One of the main objectives of computational biology is protein structure prediction. This is important for the development of novel enzymes in biotechnology and medicine. Each protein has a certain shape and structure and our life is supported by the complex and coordinated interaction of proteins. Hence identifying the protein structure possesses various challenges and various researches are performed relying upon various classifiers. In this research a fused pipit adapted deep convolutional neural network classifier (CNN) is used for the detection of the PSS with higher accuracy. The feature extraction is made using the fused Natural language processing (NLP) based pretrained models that efficiently extracted the features and is developed using the pretrained models T5XLuniref and XLnet model. The pretrained and the deep CNN classifier is optimized effectively using the pipit optimization that mimics the foraging and the safeguarding behavior of the pipits. The enabled optimization aids in the tuning of the fusion parameters and the hyper parameters of the classifier. By measuring the improvement, the model's dominance is demonstrated, and suggested method attained an progress of 1.77 %for accuracy, 1.01 % for sensitivity and 6.90 % for specificity, which proves the efficacy of the model.