Predictor

The Mykrobe predictor is designed for use by microbiologists and doctors, providing information needed in order to choose the best treatment. It analyses the whole genome of a bacterial sample, all within a couple of minutes, and predicts which drugs the infection is resistant to. No expertise is needed to run or interpret it, and it works on a standard desktop or laptop.

Staphylococcus aureus is a major cause of infections across the world. When resistant to the antibiotic methicillin, it is known as the “superbug” MRSA.

For S. aureus, standard procedure in a hospital is that a patient sample undergoes “blood culture” overnight, and a very rapid “Gram stain” test to establish that the infection is one of the family of species called staphylococci. If the DNA in this sample is then sequenced, the Mykrobe predictor takes that raw output data, which typically requires hours of analysis and interpretation, and within minutes identifies the species, checks for contamination, and predicts the resistance profile for 12 key antibiotics, including quinolones, macrolides/lincosamides, beta-lactams and tetracyclines. It will also test for the presence of Panton-Valentin Leukocidin toxin.

Download

Technical specs

Runs on Windows, Mac OS X and Linux, requires ~300Mb of RAM, takes between 40 seconds and 3 minutes to run.

Sequencing: Supports Illumina sequencing data as standard. Antibiotics supported: Beta-lactams (methicillin, penicillin), quinolones (ciprofloxacin), macrolides/lincosamides (erythromycin, clindamycin), tetracycline, aminoglycosides (gentamicin), glycopeptides (vancomycin), rifampicin, mupirocin, fusidic acid, trimethoprim.

Support for Oxford Nanopore Technology sequencing data is in development.

Predictive performance: Although not yet certified for clinical use, our tests on 992 samples show false negative rate of 0.3% (predicting susceptible when actually resistant) and false positive rate of 0.5% (predicting resistant when susceptible). See our paper for details.

Citation:

Bradley, Phelim, et al. “Rapid antibiotic-resistance predictions from genome sequence data for Staphylococcus aureus and Mycobacterium tuberculosis.”Nature communications 6 (2015).

Download the paper

Mycobacterium tuberculosis infects perhaps one third of the world population and causes the disease TB. In 2013 there were an estimated 450,000 cases of multi-drug resistant cases of TB, and there is an urgent need for fast and accurate diagnostics.

Gold-standard drug resistance tests are expensive and slow, taking as long as 6-17 weeks to get results. By comparison, if we use DNA-sequencing and then Mykrobe predictor analysis, this can reduce the time to results by 4-6 weeks.

Unusually for DNA-based diagnostics, the Mykrobe predictor detects rare “minor” populations of drug resistant bacteria, which may later spread, and our analyses show this gives the Mykrobe predictor greater ability to detect drug resistance for 2nd line drugs, a critical component of dealing with multi-drug resistant TB (MDR-TB).

Download

Technical specs

Runs on Windows, Mac OS X and Linux, requires ~300Mb of RAM, takes between 40 seconds and 3 minutes to run.

Samples: we assume samples undergo at least MGIT culture and flag positive (MGIT followed by Lowenstein-Jensen culture is also acceptable).

Sequencing: Supports Illumina sequencing data as standard.

Antibiotics supported: First line drugs (rifampicin, isoniazid, ethambutol) and second-line drugs (streptomycin, ciprofloxacin, ofloxacin, moxifloxacin, amikacin, kanamycin, capreomycin).

Support for Oxford Nanopore Technology sequencing data is in development.

Predictive performance: Currently Mykrobe is an in-silico equivalent of the HAIN-AID line probe assays (using almost exactly the same resistance panel of mutations), with the added ability to detect low frequency populations. Our tests on 1607 samples show false negative rate of 17.4% (predicting susceptible when actually resistant, comparable with the standard HAIN line probe assay) and false positive rate of 1.5% (predicting resistant when susceptible). See our paper for details including drug-specific error rates. As researchers find more resistance mutations, these can easily be added to Mykrobe, increasing sensitivity towards that of the gold-standard.

Scalability: unlike PCR-based tests,  the panel of resistance mutations used by Mykrobe predictor can trivially be extended by orders of magnitude without any problem.

Citation:

Bradley, Phelim, et al. “Rapid antibiotic-resistance predictions from genome sequence data for Staphylococcus aureus and Mycobacterium tuberculosis.”Nature communications 6 (2015).

Download the paper

Technical specs

Technical specs