Discovering B/T Cell-Cancer Antigen Affinity for Targeted Cancer Drug Therapy and Diagnosis Through Deep Neural Network Models

Abstract

Cancer, one of the leading causes of death worldwide, derives from an uncontrolled division of abnormal cells in a given part of the body. Targeted immunotherapy is a promising avenue of cancer treatments, galvanizing the body's own immune system to marshal B and T cells against abnormal cell growth, pathologically inhibiting antigen function and proliferation. In the current landscape of cancer immunotherapy, however, medicine industries are belabored with the need to use painstaking trial and error and numerous wet-lab investigations to test amino acid sequences' affinities to cancer antigens. Furthermore, most cancer antigens and the structures of their epitopes are unknown, and although most malignancies can be cured when diagnosed early, organ-specific assessments cannot be used for early-stage cancers. To bridge this gap, I innovate a deep convolutional neural network (CNN) model pipeline, which analyzes the complex amino acid makeup of tumor infiltrating B/T cell receptors based on the relationships of biochemical properties among adjacent amino acids predictive of how these receptor polypeptides fold in three-dimensional space, computing high affinity amino acid sequences to revolutionize both targeted drug discovery and early diagnosis.