Towards Parallel Bacterial Motion Detection

Abstract

One way to tackle the rise in antibiotic resistance, is to develop new techniques for testing the susceptibility of cells to antibiotics. In this paper, a comparison is made between a novel bacterial susceptibility testing method and a modification on this method. Both methods rely on bacterial samples deposited on graphene cavities, where the bacteria will stick to the graphene by the addition of APTES. By making use of a 632.8 nm He-Ne laser, the sample is probed, and the cavities then serve as ultrasenstive sensors for determining bacterial nanomotion. The existing method (single spot readout) is based on focusing a laser on graphene drums, where the drums are read out one at a time by the use of a photodiode. The modified method (parallel readout) makes, by the addition of one lens, use of an expanded laser, and a CMOS camera. At 100 frames per second, four drums are read out simultaneously. This technique hypothetically makes very high throughput possible for antimicrobial testing.

Both methods rely on converting a signal based on the intensity of the incoming light to the membrane deflection in nanometer, and the bacterial motility is found by taking the variance. The comparison of the two methods is done by performing multiple experiments, in order to relate the quality of the signal by finding the standard deviation (noted as S) of the variance of the deflection σ_z² . 

From an analysis of S, statistical quantities describing the distances between probability distributions have been conceived, and a criterion is proposed to differentiate between the noise levels of the two techniques. One such quantity is the normalized distance between the signals of two types of experiments, the one being an experiment without bacteria as a reference and the other being an experiment with living bacteria.
In the case of hypermotile bacteria, parallel readout has an average variance of deflection of 5.95 nm², it is substantially higher than the method of single spot readout, having average 2.92 nm². The unitless metric D for the distance between two signals however shows that both methods score similarly in probing nonmotile (∆-MotAB) E. Coli, as for the parallel readout the measure has for ∆-MotAB a value 0.34 and for single spot readout it has the value 0.31. In the case of hypermotile (7740) E. Coli, parallel readout scores again better with a value of 2.35 versus 0.39, which is the value obtained for single spot readout.

Finally an outlook is given where interesting findings have been summarized. With the use of power spectral densities and heatmaps of either average intensity or variance of the signal, interesting phenomena are noted and are topics for future research.