Wellcome Trust Introduces Mykrobe Predictor for Genetic Analysis of Antibiotic Resistance

By Bio-IT World Staff

December 23, 2015 | A large team of researchers led by Zamin Iqbal of the Wellcome Trust Centre for Human Genetics has published a new software tool that rapidly identifies antibiotic-resistant bacteria using DNA data. The program, called Mykrobe Predictor, has shown very similar results to standard phenotypic tests―in which bacteria are cultured and subjected to different antibiotics in the lab―in head-to-head comparisons on Staphylococcus aureus and Mycobacterium tuberculosis. The research appeared this week in Nature Communications.

As DNA sequencing becomes faster and cheaper, researchers and hospitals have explored using the technology as a diagnostic for infectious diseases. Genetic data can uniquely identify species and strains of bacteria, viruses, and other pathogens, as well as flag genes and mutations known to protect these pathogens from the most commonly used drugs. Mykrobe Predictor, trained on hundreds of samples of Staphylococcus and Mycobacterium species, quickly teases apart closely-related bacteria, allowing it to accurately find S. aureus and M. tuberculosis even in mixed samples. The authors report that their program can call species and antibacterial resistance genes from raw DNA data within three minutes.

Evaluating both live samples and in silico data, Mykrobe Predictor was able to identify antibiotic resistance factors with high accuracy, comparable to phenotypic testing; in particular, it classified over 99% of S. aureus samples correctly in the published study. However, the program is not perfect, limited mainly by current knowledge about the genetic causes of antibiotic resistance. For M. tuberculosis, false negative rates in Mykrobe Predictor exceeded 15%, as some strains of the bacterium had acquired drug resistance through unknown genetic mechanisms. This problem could be even more acute in less studied pathogens. On the other hand, the program can quickly be improved as more research is done into antibiotic resistance.

While the Nature Communications paper only looked at retrospective samples, three U.K. hospitals are now partnering with the Wellcome Trust to test out Mykrobe Predictor in the clinic. Even with 16-hour sequencing runs on the Illumina MiSeq to generate the DNA data, the Mykrobe pipeline is faster than current methods of screening for drug resistance―especially with M. tuberculosis, which reproduces slowly and can take weeks to culture in phenotypic screens. Iqbal’s team also used Mykrobe Predictor with a second sequencer, the Oxford Nanopore MinION, which yielded results in just seven hours, although they have not tested this pipeline as extensively.

(Oxford Nanopore itself has released a program, What’s in my Pot?, to quickly call microbial species in mixed samples, but it is not trained or evaluated for hospital situations.)

Mykrobe Predictor is designed for easy use by clinicians, and creates reports that break down an infection’s resistance and susceptibility drug by drug, along with the evidence for each call. The program can be run on a mobile device through a web browser, and the Iqbal lab has released the code for non-commercial use on GitHub.