Search

AbstractMost bacterial genomes contain integrated bacteriophages—prophages—in various states of decay. Many are active and able to excise from the genome and replicate, while others are cryptic prophages, remnants of their former selves. Over the last two decades, many computational tools have been developed to identify the prophage components of bacterial genomes, and it is a particularly active area for the application of machine learning approaches. However, progress is hindered and comparisons thwarted because there are no manually curated bacterial genomes that can be used to test new prophage prediction algorithms.Here, we present a library of gold-standard bacterial genome annotations that include manually curated prophage annotations, and a computational framework to compare the predictions from different algorithms. We use this suite to compare all extant stand-alone prophage prediction algorithms to identify their strengths and weaknesses.We provide a FAIR dataset for prophage identification, and demonstrate the accuracy, precision, recall, and f1 score from the analysis of seven different algorithms for the prediction of prophages. We discuss caveats and concerns in this analysis and how those concerns may be mitigated.

Background: The whole-genome sequencing (WGS) is becoming an increasingly effective tool for rapid and accurate detection of drug resistance in Mycobacterium tuberculosis complex (MTBC). This approach, however, has still been poorly evaluated on strains from Central and Eastern European countries. The purpose of this study was to assess the performance of WGS against conventional drug susceptibility testing (DST) for the detection of multi-drug resistant (MDR) phenotypes among MTBC clinical strains from Poland and Lithuania. Methods: The study included 208 MTBC strains (130 MDR; 78 drug susceptible), recovered from as many tuberculosis patients in Lithuania and Poland between 2018 and 2021. Resistance to rifampicin (RIF) and isoniazid (INH) was assessed by Critical Concentration (CC) and Minimum Inhibitory Concentration (MIC) DST as well as molecular-based techniques, including line-probe assay (LPA) and WGS. The analysis of WGS results was performed using bioinformatic pipeline- and software-based tools. Results: The results obtained with the CC DST were more congruent with those by LPA compared to pipeline-based WGS. Software-based tools showed excellent concordance with pipeline-based analysis in prediction of RIF/INH resistance. The RIF-resistant strains demonstrated a relatively homogenous MIC distribution with the mode at the highest tested MIC value. The most frequent RIF-resistance conferring mutation was rpoB S450L. The mode MIC for INH was two-fold higher among double katG and inhA mutants than among single katG mutants. The overall rate of discordant results between all methods was calculated at 5.3%. Three strains had discordant results by both genotypic methods (LPA and pipeline-based WGS), one strain by LPA only, three strains by MIC DST, two strains by both MIC DST and pipeline-based WGS, and the remaining two strains showed discordant results with all three methods, compared to CC DST. Conclusions: Considering MIC DST results, current CCs of the first-line anti-TB drugs might be inappropriately high and may need to be revised. Both molecular methods demonstrated 100% specificity, while pipeline-based WGS had slightly lower sensitivity for RIF and INH than LPA, compared to CC DST. [ABSTRACT FROM AUTHOR]