PROTEIN FAMILY CLASSIFICATION USING DEEP LEARNING TECHNIQUES

Abstract

Protein family classification is a fundamental task in bioinformatics, essential for understanding protein functions, structures, and evolutionary relationships. With the rapid growth of biological data, traditional sequence alignment methods have become less efficient and time-consuming. This project proposes a deep learning-based approach for protein family classification using advanced neural network architectures such as Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU). The system utilizes protein sequence data obtained from the SideChainNet dataset, which includes detailed information such as sequences, coordinates, and structural angles. Due to the large size of the dataset, preprocessing and feature extraction are performed using GPU-based platforms like Google Colab, and the processed data is stored in NumPy format for efficient reuse. The dataset is split into training and testing sets in an 80:20 ratio to evaluate model performance. Each deep learning model is trained on the extracted features to learn sequential patterns within protein sequences. The models are evaluated using performance metrics such as accuracy, precision, recall, and F1-score. Experimental results show that the RNN model achieves the highest accuracy of 95%, followed by LSTM with 94% and GRU with 93%. The system also provides a user interface for dataset processing, model training, and family classification of new protein sequences. This approach offers an efficient and scalable solution for protein classification, contributing to advancements in computational biology and drug discovery.