Aarya Upadhyay
Research Mentor(s): Cristian Minoccheri
Mentor Department: Computational Medicine and Bioinformatics
Authors: Cristian Minoccheri, Aarya Upadhyay, Junkuan Liu
Session: Session 3 (11:00am – 11:50am)
Presentation Type: Poster 85
Abstract
The process of developing effective antibodies is often slow and resource-intensive these days. This project explores the use of Conditional Variational Autoencoders(CVAEs) to generate new antibody sequences based on specified characteristics, opening the door to a new innovative way of creating useful antibodies. By applying machine learning techniques, we aim to create a more data-driven and efficient approach that allows for the controlled generation of antibody sequences with optimized properties. Our approach involves training a CVAE model on a database of antibody sequences, each labeled with key properties. The model learns a latent space representation of these sequences, enabling the generation of novel antibodies that meet predefined conditions. This allows for more targeted exploration of potential drug candidates, reducing the need for the trial-and-error experimentation that is commonly used these days. Preliminary results show that the model is capable of producing antibody sequences that align with specified properties, however, as we get more data, the model could continue to improve. By leveraging generative modeling techniques, this research highlights how machine learning can accelerate the drug/antibody design process and reduce the time and cost associated with antibody discovery. This approach has the potential to contribute to more efficient therapeutic development and improve the overall pipeline for biologic drug design
