Accurate and rapid prediction of tuberculosis drug resistance from genome sequence data using traditional machine learning algorithms and CNN.

Effective and timely antibiotic treatment depends on accurate and rapid in silico antimicrobial-resistant (AMR) predictions. Existing statistical rule-based Mycobacterium tuberculosis (MTB) drug resistance prediction methods using bacterial genomic sequencing data often achieve varying results: high accuracy on some antibiotics but relatively low accuracy on others. Traditional machine learning (ML) approaches have been applied to classify drug resistance for MTB and have shown more stable performance. However, there is no study that uses deep learning architecture like Convolutional Neural Network (CNN) on a large and diverse cohort of MTB samples for AMR prediction. We developed 24 binary classifiers of MTB drug resistance status across eight anti-MTB drugs and three different ML algorithms: logistic regression, random forest and 1D CNN using a training dataset of 10,575 MTB isolates collected from 16 countries across six continents, where an extended pan-genome reference was used for detecting genetic features. Our 1D CNN architecture was designed to integrate both sequential and non-sequential features. In terms of F1-scores, 1D CNN models are our best classifiers that are also more accurate and stable than the state-of-the-art rule-based tool Mykrobe predictor (81.1 to 93.8%, 93.7 to 96.2%, 93.1 to 94.8%, 95.9 to 97.2% and 97.1 to 98.2% for ethambutol, rifampicin, pyrazinamide, isoniazid and ofloxacin respectively). We applied filter-based feature selection to find AMR relevant features. All selected variant features are AMR-related ones in CARD database. 78.8% of them are also in the catalogue of MTB mutations that were recently identified as drug resistance-associated ones by WHO. To facilitate ML model development for AMR prediction, we packaged every step into an automated pipeline and shared the source code at https://github.com/KuangXY3/MTB-AMR-classification-CNN .

High-throughput prescreening of pharmaceuticals using a genome-wide bacterial bioreporter array.

We assessed the applicability of multi-strain bacterial bioreporter bioassays to drug screening. To this end, we investigated the reactions of a panel of 15 luminescent recombinant Escherichia coli bacterial bioreporters to a library of 420 pharmaceuticals. The panel included bacterial bioreporters associated with oxidative stress, DNA damage, heat shock, and efflux of excess metals. Eighty nine drugs elicited a response from at least one of the panel members and formed distinctive clusters, some of which contained closely related drugs. In addition, we tested a group of selected nine drugs against a collection of about 2000 different fluorescent transcriptional reporters that covers the great majority of gene promoters in E. coli. The sets of induced genes were in accord with the in vitro toxicity of the tested drugs, as reflected by the response patterns of the 15-member panel, and provided more insights into their toxicity mechanisms. Facilitated by microplates and robotic systems, all assays were conducted in high-throughput. Our results thus suggest that multi-strain assemblages of bacterial bioreporters have the potential for playing a significant role in drug development alongside current in vitro toxicity tests.

Whole genome characterization of the mechanisms of daptomycin resistance in clinical and laboratory derived isolates of Staphylococcus aureus.

BACKGROUND: Daptomycin remains one of our last-line anti-staphylococcal agents. This study aims to characterize the genetic evolution to daptomycin resistance in S. aureus. METHODS: Whole genome sequencing was performed on a unique collection of isogenic, clinical (21 strains) and laboratory (12 strains) derived strains that had been exposed to daptomycin and developed daptomycin-nonsusceptibility. Electron microscopy (EM) and lipid membrane studies were performed on selected isolates. RESULTS: On average, six coding region mutations were observed across the genome in the clinical daptomycin exposed strains, whereas only two mutations on average were seen in the laboratory exposed pairs. All daptomycin-nonsusceptible strains had a mutation in a phospholipid biosynthesis gene. This included mutations in the previously described mprF gene, but also in other phospholipid biosynthesis genes, including cardiolipin synthase (cls2) and CDP-diacylglycerol-glycerol-3-phosphate 3-phosphatidyltransferase (pgsA). EM and lipid membrane composition analyses on two clinical pairs showed that the daptomycin-nonsusceptible strains had a thicker cell wall and an increase in membrane lysyl-phosphatidylglycerol. CONCLUSION: Point mutations in genes coding for membrane phospholipids are associated with the development of reduced susceptibility to daptomycin in S. aureus. Mutations in cls2 and pgsA appear to be new genetic mechanisms affecting daptomycin susceptibility in S. aureus.