Sonification and Machine Learning Contribute to a Novel Assay of Bacterial Swimming Dynamics

Abstract:

Bacterial chemotaxis has the potential to enhance the efficiency of bioremediation, transporting bacteria preferentially to their substrates of choice for degradation. Qualifying and quantifying chemotaxis, however, can be particularly challenging due to the chaotic motion of individuals within a population, and often require complicated experimental setups over extended periods of time. Sonification, which is the process of converting data into sound, provides an innovative approach to the difficulties involved in extracting useful information from the chaotic motion to yield a real-time screening assay for chemotaxis, and has the added benefit of being more accessible to sight-impaired individuals. We sonified video data of Escherichia coli (E. coli) in solution taken by an Olympus Wide Field microscope under high phase contrast using a Max 8 software patch. The software patch takes each individual video frame as an input, and then applies a Fourier transform to map the pixel values to Fourier data bins that represent increasing frequency in the y direction. The resulting matrices of data are then played back as sound. We found that for E. coli under a gradient of the chemoattractant α-methylaspartate, the sound density decreases and is clearly distinct from that of E. coli not exposed to a chemoattractant gradient. We hypothesized that under phase contrast, tumbling bacteria appear as bright spots that are captured in our sonification algorithm, leading to a higher sound density - with chemotaxis suppressing tumbling, this would explain the difference. To confirm this hypothesis, we sonified videos of normal run-and-tumble E. coli, smooth swimming E. coli mutants, and tumble-only E. coli mutants. As a precursor to a human validation study, we took the sonified data and used it to train a neural network, using a supervised model to assess how different chemotactic and non-chemotactic bacterial samples sound from each other. This tool, which can be used with live video feed, provides a novel technique for researchers to sense chemotaxis or other motility changes in real time.