Rapid delineation of Acinetobacter baumannii dose responses to zosurabalpin based on machine learning-assisted interpretation of bacterial nanomotions.

UNLABELLED: Carbapenem-resistant Acinetobacter baumannii (CRAB) represents a major clinical threat, underscoring the urgent need for both new antibiotics and rapid antimicrobial susceptibility testing (AST) methods that can inform treatment. Here, we describe a susceptibility assessment that combines nanomotion technology with machine learning (ML) to rapidly classify responses to zosurabalpin (ZAB), the first-in-class antibiotic targeting the LptB₂FGC complex of the lipopolysaccharide transport pathway in A. baumannii. Nanomotion recordings from five A. baumannii strains spanning a wide range of minimum inhibitory concentrations (MICs) were measured at multiple ZAB concentrations. A logistic regression ML model was trained using a labeling strategy that categorized recordings as susceptible or non-susceptible based on strain MICs obtained by reference methods and antibiotic concentrations measured. This model, built on four quantile features, achieved >90% accuracy compared with susceptibility responses. These results highlight four key advances: (i) nanomotion-based AST can deliver reliable phenotypic classification for new antibiotics before clinical breakpoints are available, (ii) accurate results can be obtained within only 2 h, (iii) the method extends nanomotion's applicability to high-priority pathogens, such as CRAB, and (iv) the nanomotion response can reflect early cellular changes that precede a measurable decline in viability, suggesting that susceptibility may be detected before killing becomes apparent by conventional methods. Together, this work establishes nanomotion-ML AST as a promising diagnostic platform to support the evaluation and future integration of novel drugs into clinical practice, advancing next-generation rapid AST workflows against critical multidrug-resistant pathogens.

IMPORTANCE

This study demonstrates a rapid, nanomotion-based approach with potential to inform antibiotic susceptibility assessment for Acinetobacter baumannii infections. By combining nanomotion technology with quantitative machine-learning-assisted signal analysis, susceptibility to the novel antibiotic zosurabalpin, which is currently in clinical development, could be determined within 2 h. The assay reliably distinguished susceptible from non-susceptible phenotypes across strains with different MICs and in both standard and horse serum-supplemented media. Notably, the nanomotion signature reflected early cellular responses that preceded a measurable decline in viability by conventional methods. Together, these features provide rapid phenotypic insights that may support future development of timely, targeted therapy decisions. The findings highlight nanomotion-based antimicrobial susceptibility testing as a promising tool to accelerate diagnostic readiness for new antibiotics and improve treatment management in infections caused by multidrug-resistant A. baumannii.