Year: 2024
Type: other
Abstract: Identifying and classifying protein kinds and activities within biological systems is complex and challenging. Traditional classification methods often offer limited precision due to the complexity of proteins.This chapter explores the use of deep learning (DL) models, specifically variational autoencoders (VAE) that incorporate Kullback-Leibler (KL) divergence, to understand and solve this intricate problem. Proteins, which have various physical characteristics, are frequently classified broadly. However, a more detailed classification is achieved by considering the distinct properties of individual amino acids. By examining the amino acid sequences, proteins are accurately classified, as the specific arrangement within the chain intimately determines their functioning. Based on this idea, a framework that combines recurrent neural networks (RNN) and long short-term memory (LSTM) models were established. This merger demonstrated a remarkable accuracy of 96.19% during thorough experimentation. The performance of the proposed work exceeded the current cutting-edge methods, indicating significant progress in precisely understanding and grasping the complex parameters and classifications of proteins. The root acknowledges that although general classifications based on physical characteristics provide insight, genuine comprehension arises from the detailed arrangement of amino acids, facilitating accurate categorization, and interpretation of protein function.
Authors: Swathi Gowroju, M. Sudhakar, Mohit, Turki Aljrees
Institutions: Maharshi Dayanand University, University of Hafr Al-Batin
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: gold
OpenAlex ID: W4402987703