TB-DROP: deep learning-based drug resistance prediction of Mycobacterium tuberculosis utilizing whole genome mutations.

The most widely practiced strategy for constructing the deep learning (DL) prediction model for drug resistance of Mycobacterium tuberculosis (MTB) involves the adoption of ready-made and state-of-the-art architectures usually proposed for non-biological problems. However, the ultimate goal is to construct a customized model for predicting the drug resistance of MTB and eventually for the biological phenotypes based on genotypes. Here, we constructed a DL training framework to standardize and modularize each step during the training process using the latest tensorflow 2 API. A systematic and comprehensive evaluation of each module in the three currently representative models, including Convolutional Neural Network, Denoising Autoencoder, and Wide & Deep, which were adopted by CNNGWP, DeepAMR, and WDNN, respectively, was performed in this framework regarding module contributions in order to assemble a novel model with proper dedicated modules. Based on the whole-genome level mutations, a de novo learning method was developed to overcome the intrinsic limitations of previous models that rely on known drug resistance-associated loci. A customized DL model with the multilayer perceptron architecture was constructed and achieved a competitive performance (the mean sensitivity and specificity were 0.90 and 0.87, respectively) compared to previous ones. The new model developed was applied in an end-to-end user-friendly graphical tool named TB-DROP (TuBerculosis Drug Resistance Optimal Prediction: https://github.com/nottwy/TB-DROP ), in which users only provide sequencing data and TB-DROP will complete analysis within several minutes for one sample. Our study contributes to both a new strategy of model construction and clinical application of deep learning-based drug-resistance prediction methods.

Local adaptive evolution of two distinct clades of Beijing and T families of Mycobacterium tuberculosis in Chongqing: a Bayesian population structure and phylogenetic study.

BACKGROUND: Beijing sub-pedigree 2 (BSP2) and T sub-lineage 6 (TSL6) are two clades belonging to Beijing and T family of Mycobacterium tuberculosis (MTB), respectively, defined by Bayesian population structure analysis based on 24-loci mycobacterial interspersed repetitive unit-variable number of tandem repeats (MIRU-VNTR). Globally, over 99% of BSP2 and 89% of TSL6 isolates were distributed in Chongqing, suggesting their possible local adaptive evolution. The objective of this paper is to explore whether BSP2 and TSL6 originated by their local adaptive evolution from the specific isolates of Beijing and T families in Chongqing. METHODS: The genotyping data of 16 090 MTB isolates were collected from laboratory collection, published literatures and SITVIT database before subjected to Bayesian population structure analysis based on 24-loci MIRU-VNTR. Spacer Oligonucleotide Forest (Spoligoforest) and 24-loci MIRU-VNTR-based minimum spanning tree (MST) were used to explore their phylogenetic pathways, with Bayesian demographic analysis for exploring the recent demographic change of TSL6. RESULTS: Phylogenetic analysis suggested that BSP2 and TSL6 in Chongqing may evolve from BSP4 and TSL5, respectively, which were locally predominant in Tibet and Jiangsu, respectively. Spoligoforest showed that Beijing and T families were genetically distant, while the convergence of the MIRU-VNTR pattern of BSP2 and TSL6 was revealed by WebLogo. The demographic analysis concluded that the recent demographic change of TSL6 might take 111.25 years. CONCLUSIONS: BSP2 and TSL6 clades might originate from BSP4 and TSL5, respectively, by their local adaptive evolution in Chongqing. Our study suggests MIRU-VNTR be combined with other robust markers for a more comprehensive genotyping approach, especially for families of clades with the same MIRU-VNTR pattern.

Interaction between rpsL and gyrA mutations affects the fitness and dual resistance of Mycobacterium tuberculosis clinical isolates against streptomycin and fluoroquinolones.

BACKGROUND: The interaction between different drug-resistant mutations is important to the development of drug resistance and its evolution. In this study, we aimed to reveal the potential relationships between mutations conferring resistance to two important antituberculosis drugs streptomycin (STR) and fluoroquinolones (FLQ). MATERIALS AND METHODS: We used an in vitro competitive fitness assay to reveal the interactions between different mutations of rpsL and gyrA in drug-resistant Mycobacterium smegmatis, followed by the analysis of the frequency of rpsL and gyrA mutation combinations in 213 STR-FLQ dual-resistant clinical Mycobacterium tuberculosis isolates from Sichuan region, which was also investigated by the whole genome data from 3,056 global clinical M. tuberculosis isolates. RESULTS: The strains with K43R and K88R mutation in rpsL showed no difference in relative fitness compared with their susceptible ancestor, while K43N, K43M, K43T, and K88E exhibited a significantly lower relative fitness (P<0.05). For the FLQ-resistant mutants, all mutation types showed no difference in their relative fitness. Among STR-FLQ dual-resistant M. smegmatis strains, a lower fitness was detected in those with K43N/M/T and K88E instead of K43R and K88R mutations in rpsL. Among M. tuberculosis isolates harboring rpsL and gyrA dual mutations, the most two frequent combinatorial mutation types were K43R/D94G (n=37) and K43R/A90V (n=24), with the former being the most frequent one by both in vitro tests and clinical survey. CONCLUSION: Our results suggest that the interaction between rpsL and gyrA mutations affects the fitness cost in STR-FLQ dual-resistant M. smegmatis and also the predilection of mutation combinations in clinical M. tuberculosis isolates.