Application of machine learning techniques to tuberculosis drug resistance analysis.

MOTIVATION: Timely identification of Mycobacterium tuberculosis (MTB) resistance to existing drugs is vital to decrease mortality and prevent the amplification of existing antibiotic resistance. Machine learning methods have been widely applied for timely predicting resistance of MTB given a specific drug and identifying resistance markers. However, they have been not validated on a large cohort of MTB samples from multi-centers across the world in terms of resistance prediction and resistance marker identification. Several machine learning classifiers and linear dimension reduction techniques were developed and compared for a cohort of 13 402 isolates collected from 16 countries across 6 continents and tested 11 drugs. RESULTS: Compared to conventional molecular diagnostic test, area under curve of the best machine learning classifier increased for all drugs especially by 23.11%, 15.22% and 10.14% for pyrazinamide, ciprofloxacin and ofloxacin, respectively (P < 0.01). Logistic regression and gradient tree boosting found to perform better than other techniques. Moreover, logistic regression/gradient tree boosting with a sparse principal component analysis/non-negative matrix factorization step compared with the classifier alone enhanced the best performance in terms of F1-score by 12.54%, 4.61%, 7.45% and 9.58% for amikacin, moxifloxacin, ofloxacin and capreomycin, respectively, as well increasing area under curve for amikacin and capreomycin. Results provided a comprehensive comparison of various techniques and confirmed the application of machine learning for better prediction of the large diverse tuberculosis data. Furthermore, mutation ranking showed the possibility of finding new resistance/susceptible markers. AVAILABILITY AND IMPLEMENTATION: The source code can be found at http://www.robots.ox.ac.uk/ davidc/code.php. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Microbial Translocation Does Not Drive Immune Activation in Ugandan Children Infected With HIV.

Objective: Immune activation is associated with morbidity and mortality during human immunodeficiency virus (HIV) infection, despite receipt of antiretroviral therapy (ART). We investigated whether microbial translocation drives immune activation in HIV-infected Ugandan children. Methods: Nineteen markers of immune activation and inflammation were measured over 96 weeks in HIV-infected Ugandan children in the CHAPAS-3 Trial and HIV-uninfected age-matched controls. Microbial translocation was assessed using molecular techniques, including next-generation sequencing. Results: Of 249 children included, 142 were infected with HIV; of these, 120 were ART naive, with a median age of 2.8 years (interquartile range [IQR], 1.7-4.0 years) and a median baseline CD4+ T-cell percentage of 20% (IQR, 14%-24%), and 22 were ART experienced, with a median age of 6.5 years (IQR, 5.9-9.2 years) and a median baseline CD4+ T-cell percentage of 35% (IQR, 31%-39%). The control group comprised 107 children without HIV infection. The median increase in the CD4+ T-cell percentage was 17 percentage points (IQR, 12-22 percentage points) at week 96 among ART-naive children, and the viral load was <100 copies/mL in 76% of ART-naive children and 91% of ART-experienced children. Immune activation decreased with ART use. Children could be divided on the basis of immune activation markers into the following 3 clusters: in cluster 1, the majority of children were HIV uninfected; cluster 2 comprised a mix of HIV-uninfected children and HIV-infected ART-naive or ART-experienced children; and in cluster 3, the majority were ART naive. Immune activation was low in cluster 1, decreased in cluster 3, and persisted in cluster 2. Blood microbial DNA levels were negative or very low across groups, with no difference between clusters except for Enterobacteriaceae organisms (the level was higher in cluster 1; P < .0001). Conclusion: Immune activation decreased with ART use, with marker clustering indicating different activation patterns according to HIV and ART status. Levels of bacterial DNA in blood were low regardless of HIV status, ART status, and immune activation status. Microbial translocation did not drive immune activation in this setting. Clinical Trials Registration: ISRCTN69078957.

Inferring risks of coronavirus transmission from community household data.

The response of many governments to the COVID-19 pandemic has involved measures to control within- and between-household transmission, providing motivation to improve understanding of the absolute and relative risks in these contexts. Here, we perform exploratory, residual-based, and transmission-dynamic household analysis of the Office for National Statistics COVID-19 Infection Survey data from 26 April 2020 to 15 July 2021 in England. This provides evidence for: (i) temporally varying rates of introduction of infection into households broadly following the trajectory of the overall epidemic and vaccination programme; (ii) susceptible-Infectious transmission probabilities of within-household transmission in the 15-35% range; (iii) the emergence of the Alpha and Delta variants, with the former being around 50% more infectious than wildtype and 35% less infectious than Delta within households; (iv) significantly (in the range of 25-300%) more risk of bringing infection into the household for workers in patient-facing roles pre-vaccine; (v) increased risk for secondary school-age children of bringing the infection into the household when schools are open; (vi) increased risk for primary school-age children of bringing the infection into the household when schools were open since the emergence of new variants.

TETyper: a bioinformatic pipeline for classifying variation and genetic contexts of transposable elements from short-read whole-genome sequencing data.

Much of the worldwide dissemination of antibiotic resistance has been driven by resistance gene associations with mobile genetic elements (MGEs), such as plasmids and transposons. Although increasing, our understanding of resistance spread remains relatively limited, as methods for tracking mobile resistance genes through multiple species, strains and plasmids are lacking. We have developed a bioinformatic pipeline for tracking variation within, and mobility of, specific transposable elements (TEs), such as transposons carrying antibiotic-resistance genes. TETyper takes short-read whole-genome sequencing data as input and identifies single-nucleotide mutations and deletions within the TE of interest, to enable tracking of specific sequence variants, as well as the surrounding genetic context(s), to enable identification of transposition events. A major advantage of TETyper over previous methods is that it does not require a genome reference. To investigate global dissemination of Klebsiella pneumoniae carbapenemase (KPC) and its associated transposon Tn4401, we applied TETyper to a collection of over 3000 publicly available Illumina datasets containing blaKPC. This revealed surprising diversity, with over 200 distinct flanking genetic contexts for Tn4401, indicating high levels of transposition. Integration of sample metadata revealed insights into associations between geographic locations, host species, Tn4401 sequence variants and flanking genetic contexts. To demonstrate the ability of TETyper to cope with high-copy-number TEs and to track specific short-term evolutionary changes, we also applied it to the insertion sequence IS26 within a defined K. pneumoniae outbreak. TETyper is implemented in python and is freely available at https://github.com/aesheppard/TETyper.

Instrumental variables and interactions in the causal analysis of a complex clinical trial.

We consider the application of instrumental variable techniques in a longitudinal clinical trial in paediatric HIV/AIDS, with a substantial degree of non-compliance to randomized treatment (Nelfinavir versus placebo) and with left censoring of the outcome variable (HIV RNA concentration). We consider in detail the assumptions and implications behind the inclusion and exclusion of interactions between randomized arm and baseline covariates in modelling actual treatment received, and between treatment and baseline covariates in modelling outcome. Estimated treatment effects were sensitive to inclusion of interactions, and we show how such sensitivity can be explored and explained.