Explore a cutting-edge deep learning approach for predicting multi-class peptides using T-cell receptor (TCR) sequences in this conference talk from the International Conference on Bioinformatics and Systems Biology (ICBS) 2024. Delve into the innovative methodology that combines TCR sequence vectorization, similarity networks, and V/J gene information to overcome limitations of existing binary outcome models. Learn how this novel framework enhances prediction performance by integrating encoded TCR sequences, node embeddings, and V/J genes. Discover the process of training a sequence embedding model, constructing TCR networks using NAIR, and employing node embedding techniques to capture structural information. Gain insights into the four-fully-connected-layer prediction model used to determine TCR-peptide binding probabilities. Examine the effectiveness of various graph neural networks in identifying potential antigen targets, and understand the implications of this research for advancing our understanding of the immune system and developing new treatments for diseases like cancer.
Overview
Syllabus
Li Zhang: Prediction of Multi-Class Peptides by T-cell Receptor Sequences withDeepLearning #ICBS2024
Taught by
BIMSA