This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 825673.
Genotypic prediction of tuberculosis drug resistance and implementation of effective MDR-TB management
Leading partner: FZB (Germany)
An effective approach to the management of MDR-TB is to tailor drug therapy regimens on the basis of comprehensive resistance testing. CARE contributes to this effort by:
Helping elucidate MDR-TB epidemiology in Russia and Moldova
Developing a genotype-to-phenotype prediction framework based on whole genome sequencing to infer second-line TB drug resistance from genotypic data and ease the choice of personalised treatment
Designing and developing a software tool for phenotypic drug resistance prediction by genotype
Characterizing host genetic traits that predispose to extrapulmonary TB and are known to be involved in higher susceptibility to HIV
Defining and testing a pilot model intervention of personalised effective MDR-TB treatment in Russia and other countries of the former Soviet Union, and designing associated clinical training modules
We developed the first version of the software tool to predict drug-resistances from whole genome sequencing (WGS). The tool is called “Geno2Pheno[TB]”, is based on Machine Learning (ML) and we identified Random Forests (RFs) as most suitable method so far. We trained and evaluated a variety of RF-models on different genetic features for 9 important drugs (Isoniazid, Rifampicine, Streptomycin, Ethambutol Pyrazinamide, Amikacin, Capreomycin, Flouroquinolones, Thioamides) using a set of up to 2,000 Mycobacterium tuberculosis (Mtb) genome sequences. The resulting models showed overall significantly improved performance (e.g., sensitivity and specificity) compared to a state-of-the-art mutation catalogue, including both known resistance mediating and benign mutations.
In parallel, we were collecting over 400 clinical MDR-TB isolates for whole genome sequencing and extended phenotypic characterization from the Republic of Moldova and St. Petersburg, creating the MDR-TB cohort. With this newly generated dataset we will eventually describe the bacterial population structure and drug resistance development of MDR-TB isolates in Eastern Europe, and evaluate the geno2pheno[TB] tool.
A preliminary analysis of the newly established MDR-TB cohort permitted to identify genomic variants implicated in Bedaquiline (BDQ) resistance, a new hallmark drug for the treatment of MDR-TB drug therapy.