Impact of Clinical Parameters in the Intrahost Evolution of HIV-1 Subtype B in Pediatric Patients: A Machine Learning Approach. (18th September 2017)

• Record Type: Journal Article
• Authors: Rojas Sánchez, Patricia; Cobos, Alberto; Navaro, Marisa; Ramos, José Tomas; Pagán, Israel; Holguín, África
• Is Part Of: Genome biology and evolution. Volume 9:Number 10(2017:Oct.)
• Page Start: 2715
• Page End: 2726
• Publication Date: 2017-09-18
• Subjects: HIV-1 -- intrahost evolution -- pediatric patients -- machine learning; Genomics -- Periodicals; Genes -- Periodicals; 572.8605
• DOI: 10.1093/gbe/evx193
• Languages: English
• ISSNs: 1759-6653

Abstract

Abstract: Determining the factors modulating the genetic diversity of HIV-1 populations is essential to understand viral evolution. This study analyzes the relative importance of clinical factors in the intrahost HIV-1 subtype B (HIV-1B) evolution and in the fixation of drug resistance mutations (DRM) during longitudinal pediatric HIV-1 infection. We recovered 162 partial HIV-1B pol sequences (from 3 to 24 per patient) from 24 perinatally infected patients from the Madrid Cohort of HIV-1 infected children and adolescents in a time interval ranging from 2.2 to 20.3 years. We applied machine learning classification methods to analyze the relative importance of 28 clinical/epidemiological/virological factors in the HIV-1B evolution to predict HIV-1B genetic diversity ( d ), nonsynonymous and synonymous mutations ( d N, d S ) and DRM presence. Most of the 24 HIV-1B infected pediatric patients were Spanish (91.7%), diagnosed before 2000 (83.3%), and all were antiretroviral therapy experienced. They had from 0.3 to 18.8 years of HIV-1 exposure at sampling time. Most sequences presented DRM. The best-predictor variables for HIV-1B evolutionary parameters were the age of HIV-1 diagnosis for d, the age at first antiretroviral treatment for d N and the year of HIV-1 diagnosis for d s . The year of infection (birth year) and year of sampling seemed to be relevant for fixation of both DRM at large and, considering drug families, to protease inhibitors (PI). This study identifies, for the first time using machine learning, the factors affecting more HIV-1B pol evolution and those affecting DRM fixation in HIV-1B infected pediatric patients.