Logistic Regression Classifiers in the Analysis of HIV-1 Use of CCR5 and CXCR4 Coreceptors

Resumo

The development of HIV coreceptor antagonists—such as maraviroc—has made coreceptor use diagnosis mandatory for each patient in the clinical setting. Technical issues presented by Trofile™, the gold standard phenotypic assay, hindered its use as a routine diagnostic tool. This issue has lead to the development of genotypic algorithms, whose evaluations are based on V3 region sequences of the viral envelope. These algorithms proved to be cheaper, easier to use, and less time consuming than the phenotypic method. One of them, geno2pheno has also gained widespread use since it showed 86.5% concordance with the gold-standard phenotypic assay. In this regard, the present project aims to develop accurate predictive models based on V3 sequence information. For this, 2,109 DNA sequences from HIV-1 subtype B were used. Data labeled with geno2pheno’s results were used to construct and evaluate the logistic regression models. Three parsimonious models were obtained, and their evaluations showed promising results. These outcomes encourage us to continue the project of an easy to use and more accurate alternative for HIV coreceptor use diagnosis, capable of guiding clinical decision making regarding the use of CCR5 antagonists in HIV treatment.