It takes a while to determine the medication susceptibility profile of a bacterial illness. Researchers at the Nara Institute of Science and Technology and its collaborators have just released studies on a technology that, by greatly speeding up this currently laborious process, may be able to save lives.
According to the US CDC, antibiotic-resistant diseases kill more than a million people annually across the globe. Quickly determining a suitable medication to which the infective bacteria are amenable is essential to controlling resistant diseases. Senior author Yaxiaer Yalikun claims that “susceptibility data are frequently required much more quickly than conventional tests can provide them.” We created a technology that can suit this demand in order to address this.
The core of the group’s work is impedance cytometry, which examines the dielectric characteristics of individual cells with high throughput—more than a thousand cells per minute. Because the physical response of a bacterium to an antibiotic and its electrical readout coincide, it is simple to determine if an antibiotic kills the bacteria. Before calibrating the impedance of the two particles in a sample using conventional impedance cytometry, technical specialists must do substantial post-processing on the test (antibiotic-treated) and reference (untreated) particles. The team was adamant about getting through this significant obstacle.
In a work that appears in ACS Sensors, the team develops a novel impedance cytometry technique that simultaneously analyses the test and reference particles in various channels, resulting in readily analysed separate datasets.
The cytometry’s nanoscale sensitivity allowed it to identify even the tiniest physical changes in bacterial cells. In a similar study that was printed in Sensors and Actuators B, the team developed a machine learning method to analyse the impedance cytometry data. Because the novel cytometry method separates the test and reference datasets, the reference dataset may be automatically labelled as the “learning” dataset and used by the machine learning tool to learn the features of an untreated bacterium.
By comparing live data with cells that have received antibiotic treatment, the technology can identify whether the bacteria are susceptible to the drug and may even determine what proportion of bacterial cells are resistant in a population with mixed resistance.
Another senior author on the team, Yoichiroh Hosokawa, notes that despite the fact that their study had a misidentification error rate of less than 10%, they were still able to distinguish between susceptible and resistant cells two hours after antibiotic treatment.