Impacto de uso de expressões gênicas com maior precisão e com aplicação de modelo de redução de ruído em diferentes embeddings para geração de moléculas

Abstract

The generation of new molecules and molecular compounds is a task that has a wide range of applications in various areas of science, and is highly relevant for the development of drugs for diseases that still have no cure. With the ``de novo'' design, machine learning algorithms are capable of creating new molecules using Generative Artificial Intelligence. However, the gene expression used for this process has data with a lot of noise and that is difficult to map. This work seeks to explore the way in which gene expressions are represented, exploring the representation of gene expressions by different embeddings, with different decimal points of precision, and with the application of a noise reduction model. The results obtained show a good result with gene expressions of three precision points, and a wide variety in the application of the noise reduction model for different embeddings.