Aplicação de autocodificador variacional para agrupamento de sequência de anticorpos

Abstract

The discovery of new antibody-based treatments is an area of growing importance in biotechnology, particularly in immunotherapies for diseases such as cancer and autoimmune disorders. However, the vast diversity in the structure and sequence of antibodies presents a significant challenge in identifying promising therapeutic candidates. This work aims to optimize this process through machine learning techniques, focusing on the clustering of antibodies based on their structural characteristics. By utilizing advanced deep learning models, such as Variational Autoencoders, and clustering methods, we propose an approach that seeks to identify latent patterns and natural clusters within large volumes of antibody data. This strategy can facilitate the screening and prioritization of candidates by gathering and grouping shared characteristics among antibodies. The expectation is that this work will help accelerate research and the development of immunotherapies, enabling a more efficient selection of promising candidates based on their structural properties.