Fast and Accurate Epitope Prediction Using GPU-Accelerated ML Techniques

EasyChair Preprint 14169

Abstract

Epitope prediction is a critical component of immunoinformatics, playing a vital role in vaccine development, diagnostic assays, and therapeutic design. Traditional computational methods for epitope prediction often struggle with the challenges of accuracy and speed due to the complexity of protein-antigen interactions and large-scale data processing requirements. This study presents a novel approach to enhance epitope prediction through the integration of GPU-accelerated machine learning (ML) techniques. By leveraging the parallel processing capabilities of GPUs, our method significantly accelerates the training and inference phases of ML models designed to predict epitopes from protein sequences. We employ advanced deep learning architectures, including convolutional neural networks (CNNs) and transformers, optimized for GPU computation to improve prediction accuracy and computational efficiency. Our approach not only reduces the time required to process large datasets but also enhances the precision of epitope identification compared to traditional CPU-based methods. Benchmarking results demonstrate that our GPU-accelerated ML models achieve superior performance in both speed and accuracy, providing a valuable tool for researchers and clinicians in the field of immunology. This work underscores the potential of GPU-accelerated machine learning techniques in advancing the state-of-the-art in epitope prediction and offers a scalable solution for handling the complexities of modern immunoinformatics challenges.

Keyphrases: Convolutional Neural Networks (CNNs), Graphics Processing Unit (GPU), machine learning

@booklet{EasyChair:14169,
  author    = {Abi Litty},
  title     = {Fast and Accurate Epitope Prediction Using GPU-Accelerated ML Techniques},
  howpublished = {EasyChair Preprint 14169},
  year      = {EasyChair, 2024}}