Human sensor-inspired supervised machine learning of smartphone-based paper microfluidic analysis for bacterial...

Area of Science:

• Biomimetic sensing
• Machine learning applications in microbiology
• Environmental monitoring technologies

Background:

• Traditional bacteria identification methods (e.g., antibodies, nucleic acid sequences) are limited for unknown species in complex environmental samples.
• The biofilm-bacteria interface offers unique peptide interactions that can be leveraged for bacterial recognition.
• Existing methods may not be suitable for rapid, field-based environmental monitoring.

Purpose of the Study:

• To develop a novel, low-cost, and field-ready assay for bacteria identification.
• To investigate the use of supervised machine learning on bacteria-particle aggregation patterns for species classification.
• To enable rapid analysis of complex biological and environmental samples.

Main Methods:

• Peptides conjugated to polystyrene particles interact with bacterial species, inducing aggregation.
• Aggregation alters liquid properties (surface tension, viscosity), affecting flow velocity in paper microfluidic chips.
• Smartphone cameras capture flow velocity changes, generating unique bacterial fingerprints.
• Support vector machine (SVM) algorithm is employed for species classification.

Main Results:

• The developed method achieves 93.3% accuracy in identifying five bacterial species (E. coli, S. aureus, S. Typhimurium, E. faecium, P. aeruginosa).
• Flow rates are monitored in under 6 seconds, with a total sample-to-answer time of less than 10 minutes.
• The approach generates unique fingerprinting profiles for each bacterial species based on flow velocity data.

Conclusions:

• This biomimetic approach combined with machine learning offers a new paradigm for bacteria identification.
• The method is suitable for analyzing complex biological and environmental samples, even with unknown bacterial species.
• The low-cost, rapid, and field-ready nature of the assay facilitates environmental monitoring applications.