RaMoA: Raman Microspectroscopy and Deep Learning for the Classification of Antimicrobial Mechanism of Action.

Area of Science:

• Microbiology
• Spectroscopy
• Artificial Intelligence

Background:

• Understanding antimicrobial Mechanism of Action (MoA) is crucial for developing new drugs.
• Raman microspectroscopy offers detailed chemical insights into bacterial responses.
• Deep learning excels at analyzing complex spectral data.

Purpose of the Study:

• To develop and validate a technology (RaMoA) for classifying antimicrobial MoA using Raman spectroscopy and deep learning.
• To assess the novelty of antimicrobial MoA for potential drug candidates.

Main Methods:

• Utilized Raman microspectroscopy to capture spectral signatures of *Escherichia coli* treated with various antibiotics.
• Employed a 1D Convolutional Neural Network for spectral classification of MoA.
• Applied an autoencoder architecture to evaluate the novelty of MoA for candidate antibiotics.

Main Results:

• Achieved 96% accuracy in classifying MoA using individual Raman spectra.
• Reached 100% accuracy in MoA assignment when aggregating spectral predictions.
• Successfully identified colistin as a novel MoA, distinguishing it from known cell wall and protein synthesis inhibitors.

Conclusions:

• The RaMoA technology accurately classifies antimicrobial MoA and predicts novelty.
• This approach can accelerate the discovery and development of new antimicrobial agents.