Genotypic prediction of TB drug resistance and treatment management - latest enrolment updates at July 2020

The databases

TB phenotypic resistance data

Phenotypic analysis of MDR strains from the laboratories in St. Petersburg and Chisinau is performed, including the collection of basic patient characteristics from the hospital records. This collection will form the basis of a genomic epidemiological study of MDR strains in Eastern Europe, and contributes to an analyisis of treatment outcomes of therapies with the new MDR-TB drug bedaquiline. Eventually this cohort will be employed to test the geno2pheno[TB] performance. The DB data come from centers in Russia and Moldova

TB WGS data

Whole genome sequencing (WGS) of MDR strains in the laboratories in Borstel and St. Petersburg. The study includes re-cultivation (if needed) and isolation of mycobacterial DNA as well as harmonizatons of protocols between Borstel, Chisinau and St. Petersburg. WGS data will also be pocessed with the geno2pheno prototype. Resistance predictions will be evaluated to the gold standard pDST previously retrieved. The DB data come from centers in Russia and Moldova

TB Geno2Pheno

Development of a machine learning algorithm to predict drug resistances from WGS data (geno2pheno). Starting from about 1200 published WGS datasets and available pDST results pre-liminary models are trained and different algorithms are compared. To this end, different supervised learning techniques have been evaluated. Finally, a version of geno2pheno[TB] will be evaluated with the cohort from "TB phenotypic drug susceptibility data" and "TB WGS data" to predict the best therapy regimen under consideration of recent TB treatment guidelines.