Using Immune Footprints in a Novel Deep Learning Model to Detect Human Diseases

Overview

In response to diseases and regardless of their characteristics, humans have developed an all - encompassing defense mechanism known as the immune system, where antibodies are a key player. For antibodies to be effective, they must bind to the surfaces of di sease-related antigens with highly variable shapes, doing so through recombination processes that make each antibody unique. This uniqueness allows for a correlation to be established from an antibody to its corresponding disease. While there have been previous attempts to correlate diseases using feature-based machine learning, the direct use of amino acid sequences in a deep learning model remains to be explored. Here, we propose a language modeling- based approach for classifying disease -specific antibodies against a healthy control set. Using the pre - trained ProtBERT-BFD model from Rostlab, we were able to generate an embedding vector with 1024 values for each amino acid in an antibody sequence. These values were then averaged across every amino acid to obtain a single “sentence -embedding vector” that was passed to a feedforward neural network of progressively smaller layers. The final output con sisted of probability matches for each disease, with the highest probability becoming the predicted class. This multi -class model was then implemented in antibodies associated with COVID-19, HIV, influenza, Dengue fever, CLL, and a healthy control, achieving an overall accuracy of 90.50%. We believe this high accuracy demonstrates the potential for a rapid, multi- disease diagnosis method supporting hundreds of diseases, carrying with it immense ramifications for the future of medical testing.

Competition history

• JSHS 2024 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.