Avaliação do uso de inteligência artificial aplicada à mineração de dados termodinâmicos de proteínas.

Abstract

The alignment of modern biotechnology with bioinformatics has provided important information for the discovery and development of new drugs. Mutagenesis studies from computational approaches have attempted to predict the effects of missense mutations on proteins that are related to serious diseases through their three-dimensional structures. For this purpose, computational stability predictors which evaluate the effects of mutation need a large volume of thermodynamic data to be able to predict the structural effects caused to the protein. One of the recurring problems is the lack of structuring and standardization of the data used, which takes a lot of human time to solve. Thus, the use of artificial intelligence makes data mining and management possible in less time, assisting in the design process of new drugs. This research presents the training of a machine learning model, on the LitSuggest platform, to retrieve references containing thermodynamic data of proteins, deposited in the PubMed repository. A total of 14 references were classified by the model and selected in manual curation, totaling 283 new mutations and 2,901 new data added to ThermoMutDB.