Original article

Mechanical motion dynamics of surface-adhered Escherichia coli exposed to ampicillin and polymyxin B: an exploratory computational analysis

---

Abstract

Aim: To explore whether motion features extracted from publicly available supplementary recordings distinguish the responses of surface-adhered Escherichia coli HCB136 to ampicillin and polymyxin B using Gaussian mixture models (GMMs) and Gaussian process regression (GPR).

Methods: This secondary computational analysis included all seven supplementary recordings published by Johnson et al. (2017): three polymyxin B and four ampicillin recordings of non-motile E. coli K-12 HCB136 with paralysed flagella adhered to quartz resonators. Motion features were extracted using optical-flow and centroid-tracking procedures and explored using GMMs, principal component analysis, K-means clustering, video-level linear trends, and GPR.

Results: Normalised motion decreased by 93–98% in the polymyxin B recordings and by 50–62% in the ampicillin recordings. Polymyxin B recordings showed lower and less dispersed GMM-derived motion features, whereas ampicillin recordings showed greater between-recording variability. Descriptive temporal profiles showed distinct patterns, but independent predictive validation was not possible.

Conclusion: The available recordings under the two antibiotic conditions showed different mechanical motion patterns in this secondary analysis. Because only seven publicly available recordings from one strain and one concentration of each antibiotic were analysed, the findings are exploratory and require validation against standard antimicrobial susceptibility endpoints.

Keywords: ampicillin, Escherichia coli, models, statistical, polymyxin B