Implementação de funcionalidades para uma plataforma de análise de variantes e novos métodos para prover melhor acurácia na identificação de mutações patogênicas

Abstract

Current scientific advances in genomics have been provided due to increasing extraction of significant DNA information owing to use of new technologies available for the analysis of genetic data. A current challenge of precision medicine is identify which of the mutations detected by the sequencing process play a role in responding to a treatment, in tumorigenesis, or in diagnosis. Considering that one of current challenges of precision medicine is identify which mutations detected by sequencing process have a possible role in response to a treatment, in tumorigenesis or in diagnosis. We propose that through this study an improvement component of software product (ViVa) was implemented, responsible for providing assistance to data collected. It has been improved, in order to make analyzes more efficient and their visualization more accurately. This work proposes the implementation of new functionalities that add value to the product, contributing directly to automation and improvement of the processes performed by the analysis tools of variants available in the market. Aiming at a practical applicability about was developed, an analysis of public data used to annotate the variants from this system was proposed. For this, a study was carried out regarding the data of existing predictors, through which it was identified that mean accuracy of predictors turns around 85%. However, although this rate is considerably high, it was also possible to observe that there is a high degree of disagreement between the predictors regarding identification of the mutational impact and its pathogenicity. In order to improve this accuracy, we describe the creation of a decision tree, and the discretization of characteristics (attributes coming from the integration of databases). In tests performed, when we compared the results obtained in our decision tree with the predictors, our decision tree reached the highest precision in all tested variables: true neutral 87%, false neutral 6%, false pathogenic 13% and, true pathogenic 94%.