Abstract
MotivationTimely 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.ResultsCompared 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 implementationThe source code can be found at http://www.robots.ox.ac.uk/ davidc/code.phpSupplementary information Supplementary data are available at Bioinformatics online.
Highlights
Tuberculosis (TB) is one of the leading causes of mortality across the world (World Health Organization et al, 2016)
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
The best performing classifier outperformed the assessed direct association (DA) method in terms of F1-Score, area under curve (AUC) and sensitivity
Summary
Tuberculosis (TB) is one of the leading causes of mortality across the world (World Health Organization et al, 2016). Conventional whole genome sequencing (WGS) methods are based on identifying a number of variants (i.e. single nucleotide polymorphisms, insertions or deletions) and interpreting them as associated (or not) with conferring resistance to each individual drug (Schleusener et al, 2017). They relies on a library of previously identified resistance-associated variants (Coll et al, 2015; Georghiou et al, 2012; Walker et al, 2015). Such techniques can result in lower performance for less-studied drugs e.g. PZA and second-line drugs due to high dimensionality and improper composition of the library
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
