A Universal Method for Fingerprinting Multiplexed Bacteria: Evolving Pruned Sensor Arrays via Machine Learning-Driven...

Area of Science:

• Biomedical Engineering
• Nanotechnology
• Clinical Diagnostics

Background:

• Array-based sensing offers potential for biological system analysis but faces challenges in multianalyte identification and diverse disease diagnostics.
• Developing universal strategies for simultaneous detection of various analytes and meeting clinical needs for multiclassified diseases remains difficult.

Purpose of the Study:

• To introduce a combinatorial method for constructing sensor arrays with dual bacterial targeting capabilities.
• To develop a rapid screening strategy using machine learning for optimal array generation.
• To demonstrate the diagnostic potential of these arrays for clinical infectious diseases.

Main Methods:

• Assembling two types of group-specific elements to create a library of 100 sensing units.
• Employing a three-step screening strategy optimized by machine learning algorithms.
• Utilizing nine multiclassification algorithms, including multilayer perceptron (MLP), for array optimization and performance evaluation.

Main Results:

• Rapid generation of a library of 100 sensing units with dual bacterial targeting capabilities.
• Identification of optimal five-element arrays for diverse clinical infectious models through machine learning-optimized screening.
• Successful quantitative detection and identification of bacterial strains at disparate mixing ratios.
• Achieved 100% accuracy in diagnosing clinical urinary tract infections (UTIs) and 99.4% accuracy in clinical sepsis detection using the optimized MLP model.

Conclusions:

• The combinatorial library construction and screening process provides a standard approach for generating powerful sensor elements.
• The developed mini-sensor arrays demonstrate high recognition and discriminative capabilities for clinical diagnostics.
• This method offers insights into creating effective sensor arrays for complex biological and clinical challenges.