Dataset Information

DeepAMR for predicting co-occurrent resistance of Mycobacterium tuberculosis.

ABSTRACT:

Motivation

Resistance co-occurrence within first-line anti-tuberculosis (TB) drugs is a common phenomenon. Existing methods based on genetic data analysis of Mycobacterium tuberculosis (MTB) have been able to predict resistance of MTB to individual drugs, but have not considered the resistance co-occurrence and cannot capture latent structure of genomic data that corresponds to lineages.

Results

We used a large cohort of TB patients from 16 countries across six continents where whole-genome sequences for each isolate and associated phenotype to anti-TB drugs were obtained using drug susceptibility testing recommended by the World Health Organization. We then proposed an end-to-end multi-task model with deep denoising auto-encoder (DeepAMR) for multiple drug classification and developed DeepAMR_cluster, a clustering variant based on DeepAMR, for learning clusters in latent space of the data. The results showed that DeepAMR outperformed baseline model and four machine learning models with mean AUROC from 94.4% to 98.7% for predicting resistance to four first-line drugs [i.e. isoniazid (INH), ethambutol (EMB), rifampicin (RIF), pyrazinamide (PZA)], multi-drug resistant TB (MDR-TB) and pan-susceptible TB (PANS-TB: MTB that is susceptible to all four first-line anti-TB drugs). In the case of INH, EMB, PZA and MDR-TB, DeepAMR achieved its best mean sensitivity of 94.3%, 91.5%, 87.3% and 96.3%, respectively. While in the case of RIF and PANS-TB, it generated 94.2% and 92.2% sensitivity, which were lower than baseline model by 0.7% and 1.9%, respectively. t-SNE visualization shows that DeepAMR_cluster captures lineage-related clusters in the latent space.

Availability and implementation

The details of source code are provided at http://www.robots.ox.ac.uk/?davidc/code.php.

Supplementary information

Supplementary data are available at Bioinformatics online.

SUBMITTER: Yang Y

PROVIDER: S-EPMC6748723 | biostudies-literature | 2019 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

DeepAMR for predicting co-occurrent resistance of Mycobacterium tuberculosis.

Yang Yang Y Walker Timothy M TM Walker A Sarah AS Wilson Daniel J DJ Peto Timothy E A TEA Crook Derrick W DW Shamout Farah F Zhu Tingting T Clifton David A DA

Bioinformatics (Oxford, England) 20190901 18

<h4>Motivation</h4>Resistance co-occurrence within first-line anti-tuberculosis (TB) drugs is a common phenomenon. Existing methods based on genetic data analysis of Mycobacterium tuberculosis (MTB) have been able to predict resistance of MTB to individual drugs, but have not considered the resistance co-occurrence and cannot capture latent structure of genomic data that corresponds to lineages.<h4>Results</h4>We used a large cohort of TB patients from 16 countries across six continents where wh ...[more]

PMID: 30689732

Similar Datasets

Project description:Background Tuberculosis is one of the deadliest disease caused by Mycobacterium tuberculosis. Its treatment still becomes a burden for many countries including Indonesia. Drug resistance is one of the problems in TB treatment. However, a development in the molecular field through Whole-genome sequencing (WGS) can be used as a solution in detecting mutations associated with TB- drugs. This investigation intended to implement this data for supporting the scientific community in deeply understanding any TB epidemiology and evolution in Papua along with detecting any mutations in genes associated with TB-Drugs. Result A whole-genome sequencing was performed on the random samples from TB Referral Laboratory in Papua utilizing MiSeq 600 cycle Reagent Kit (V3). Furthermore, TBProfiler was used for genome analysis, RAST Server was employed for annotation, while Gview server was applied for BLAST genome mapping and a Microscope server was implemented for Regions of Genomic Plasticity (RGP). The largest genome of M. tuberculosis obtained was at the size of 4,396,040 bp with subsystems number at 309 and the number of coding sequences at 4326. One sample (TB751) contained one RGP. The drug resistance analysis revealed that several mutations associated with TB-drug resistance existed. In details, mutations of rpoB gene which were identified as S450L, D435Y, H445Y, L430P, and Q432K had caused the reduced effectiveness of rifampicin; while the mutases in katG (S315T), kasA (312S), inhA (I21V), and Rv1482c-fabG1 (C-15 T) genes had contributed to the resistance in isoniazid. In streptomycin, the resistance was triggered by the mutations in rpsL (K43R) and rrs (A514C, A514T) genes, and, in Amikacin, its resistance was led by mutations in rrs (A514C) gene. Additionally, in Ethambutol and Pyrazinamide, their reduced effectiveness was provoked by embB gene mutases (M306L, M306V, D1024N) and pncA (W119R). Conclusions The results from whole-genome sequencing of TB clinical sample in Papua, Indonesia could contribute to the surveillance of TB-drug resistance. In the drug resistance profile, there were 15 Multi Drugs Resistance (MDR) samples. However, Extensively Drug-resistant (XDR) samples have not been found, but samples were resistant to only Amikacin, a second-line drug. Supplementary Information The online version contains supplementary material available at 10.1186/s12864-021-08139-3.

Project description:RationaleMinority drug-resistant Mycobacterium tuberculosis subpopulations can be associated with phenotypic resistance but are poorly detected by Sanger sequencing or commercial molecular diagnostic assays.ObjectivesTo determine the role of targeted next-generation sequencing in resolving these minor variant subpopulations.MethodsWe used single molecule overlapping reads (SMOR), a targeted next-generation sequencing approach that dramatically reduces sequencing error, to analyze primary cultured isolates phenotypically resistant to rifampin, fluoroquinolones, or aminoglycosides, but for which Sanger sequencing found no resistance-associated variants (RAVs) within respective resistance-determining regions (study group). Isolates also underwent single-colony selection on antibiotic-containing agar, blinded to sequencing results. As a positive control, isolates with multiple colocalizing chromatogram peaks were also analyzed (control group).Measurements and main resultsAmong 61 primary culture isolates (25 study group and 36 control group), SMOR described 66 (49%) and 45 (33%) of 135 total heteroresistant RAVs at frequencies less than 5% and less than 1% of the total mycobacterial population, respectively. In the study group, SMOR detected minor resistant variant subpopulations in 80% (n = 20/25) of isolates with no Sanger-identified RAVs (median subpopulation size, 1.0%; interquartile range, 0.2-3.9%). Single-colony selection on drug-containing media corroborated SMOR results for 90% (n = 18/20) of RAV-containing specimens, and the absence of RAVs in 60% (n = 3/5) of isolates. Overall, Sanger sequencing was concordant with SMOR for 77% (n = 53/69) of macroheteroresistant (5-95% total population), but only 5% of microheteroresistant (<5%) subpopulations (n = 3/66) across both groups.ConclusionsCryptic minor variant mycobacterial subpopulations exist below the resolving capability of current drug susceptibility testing methodologies, and may explain an important proportion of false-negative resistance determinations.

Project description:Pyrazinamide (PZA) is a prodrug that is converted to pyrazinoic acid by the enzyme pyrazinamidase, encoded by the pncA gene in Mycobacterium tuberculosis. Molecular identification of mutations in pncA offers the potential for rapid detection of pyrazinamide resistance (PZA(r)). However, the genetic variants are highly variable and scattered over the full length of pncA, complicating the development of a molecular test. We performed a large multicenter study assessing pncA sequence variations in 1,950 clinical isolates, including 1,142 multidrug-resistant (MDR) strains and 483 fully susceptible strains. The results of pncA sequencing were correlated with phenotype, enzymatic activity, and structural and phylogenetic data. We identified 280 genetic variants which were divided into four classes: (i) very high confidence resistance mutations that were found only in PZA(r) strains (85%), (ii) high-confidence resistance mutations found in more than 70% of PZA(r) strains, (iii) mutations with an unclear role found in less than 70% of PZA(r) strains, and (iv) mutations not associated with phenotypic resistance (10%). Any future molecular diagnostic assay should be able to target and identify at least the very high and high-confidence genetic variant markers of PZA(r); the diagnostic accuracy of such an assay would be in the range of 89.5 to 98.8%. Importance: Conventional phenotypic testing for pyrazinamide resistance in Mycobacterium tuberculosis is technically challenging and often unreliable. The development of a molecular assay for detecting pyrazinamide resistance would be a breakthrough, directly overcoming both the limitations of conventional testing and its related biosafety issues. Although the main mechanism of pyrazinamide resistance involves mutations inactivating the pncA enzyme, the highly diverse genetic variants scattered over the full length of the pncA gene and the lack of a reliable phenotypic gold standard hamper the development of molecular diagnostic assays. By analyzing a large number of strains collected worldwide, we have classified the different genetic variants based on their predictive value for resistance which should lead to more rapid diagnostic tests. This would assist clinicians in improving treatment regimens for patients.

Dataset Information

DeepAMR for predicting co-occurrent resistance of Mycobacterium tuberculosis.

Motivation

Results

Availability and implementation

Supplementary information

Publications

DeepAMR for predicting co-occurrent resistance of Mycobacterium tuberculosis.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets