TB-BIOINFORMATICS FOR THE NON-BIOINFORMATICIAN: A COMPARISON OF FOUR USER-FRIENDLY DRUG-RESISTANCE PREDICTION PIPELINES

Drug-resistant tuberculosis (DRTB) causes the deaths of hundreds of thousands of people annually. Timely and accurate diagnosis of DRTB is key in controlling this global health threat. Conventional drug susceptibility testing of M. tuberculosis is based on culture which can take more than two weeks before a result is obtained. Whole-genome sequencing (WGS) approaches have been used to rapidly detect known gene mutations associated with resistance to all drugs directly from a primary culture. The massive amount of WGS data, however, makes analysis a daunting challenge, especially for the nonbioinformatician. Online tools have been developed to predict antibiotic resistance from WGS data easily, enabling routine use of WGS without bioinformatics expertise. Here, we analyse four such tools, three of which are specific for DRTB, and one comprehensive tool covering all bacteria. Clinical M. tuberculosis cultures with known phenotypic drug susceptibilities were subjected to WGS using the Illumina MiSeq platform. Obtained data were analysed using the following online tools: TB-Profiler, Mykrobe, PhyResSe, and Comprehensive Antibiotic Resistance Database (CARD). The sensitivity and specificity of each tool were calculated for first-line antibiotics rifampicin, isoniazid, ethambutol, and pyrazinamide. Sensitivity ranged from 83.6% to 100% for rifampicin, 82,4% to 97.1% for isoniazid, 69.2% to 92.3% for ethambutol, and 62.5% to 100% for pyrazinamide. For rifampicin and isoniazid, CARD showed highest sensitivity but lowest specificity. CARD was the most sensitive in predicting rifampicin and isoniazid resistance, however, most of the mutations detected were not specific for M. tuberculosis. All the tools provided mutation data, and all except CARD provided strain genotype. TB-Profiler was the most comprehensive and included other TB-specific genetic analysis as well. User-friendly WGS analysis tools can facilitate in timely diagnosis of DRTB. Although bioinformatics expertise is not required, basic knowledge in M. tuberculosis mutations is essential for making informed-interpretations of the analysed data.