Nanomotion technology in combination with machine learning: a new approach for rapid antibiotic susceptibility test for Mycobacterium tuberculosis
A Vocat(1,2) A Sturm(2) G Jozwiakb(2) G Jozwiakb(2) G Cathomen(2) M Swiatkowski(2) R Buga(2) G Wielgoszewski(2) D Cichocka(2) G Greub(1,3) O Opota(1)
1:Institute of Microbiology, Lausanne University Hospital and University of Lausanne, Lausanne, 1011, Switzerland; 2:Resistell AG, Muttenz, 4132, Switzerland; 3:Service of Infectious Diseases, Lausanne University Hospital and University of Lausanne, Lausanne, 1011, Switzerland
The antibiotic susceptibility of Mycobacterium tuberculosis (MTB) is assessed using both phenotypic and molecular methods. Phenotypic culture-based assays, are hindered by MTB's slow growth. Nanomotion technology is a growth-independent approach that records vibrations of bacteria attached to microcantilevers. Changes in the extent of vibrations upon antibiotic exposure can distinguish susceptible from resistant bacteria.
We developed a nanomotion-based protocol for AST in Mycobacterium tuberculosis MTB, which was applied to predict strain susceptibility to isoniazid (INH) and rifampicin (RIF) using leave-one-out cross-validation (LOOCV) and machine learning techniques.
The MTB-nanomotion protocol takes 21 hours including preparation of the cell suspension, optimized bacterial attachment to functionalized cantilever and nanomotion recording upon antibiotic exposure. We applied this protocol combined LOOCV and machine learning to MTB isolates (n=40) and were able to discriminate between susceptible and resistance for INH and RIF with a maximum sensitivity of 97.4% and 100% respectively and the maximum specificity for both antibiotics was 100%; when considering each nanomotion recording independently. Triplicate experiments improved sensitivity and specificity to 100% for both antibiotics.
Nanomotion technology has the potential to reduce the time required for MTB antibiotic susceptibility testing (AST) to just a few hours, compared to the days or weeks required for current phenotypic methods. This technology could also be expanded to test other anti-TB drugs, including newer drugs like bedaquiline, providing more effective guidance for TB treatment.