An antimicrobial drug recommender system using MALDI-TOF MS and dual-branch neural networks

Area of Science:

• Clinical Microbiology
• Computational Biology
• Mass Spectrometry

Background:

• Antimicrobial drugs are crucial for patient outcomes and preventing resistance.
• Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is standard for species identification.
• Extracting antimicrobial resistance (AMR) data from MALDI-TOF MS spectra offers significant potential.

Purpose of the Study:

• To develop machine learning models for predicting AMR profiles across various species and drugs.
• To create a drug recommender system for infectious diseases based on MALDI-TOF MS data.
• To enhance the clinical utility of MALDI-TOF MS in diagnostics.

Main Methods:

• Development of novel dual-branch machine learning models.
• Utilizing MALDI-TOF MS spectral data for AMR profiling.
• Model training and validation for clinical microbiology applications.

Main Results:

• The dual-branch method significantly outperformed previous approaches in AMR prediction.
• Models demonstrated efficient fine-tuning capabilities for different clinical laboratory datasets.
• Successful prediction of AMR profiles, enabling drug recommender system functionality.

Conclusions:

• MALDI-TOF MS can be leveraged beyond species identification to predict AMR profiles.
• Machine learning models offer a promising solution for improving antimicrobial treatment strategies.
• MALDI-TOF-based AMR recommender systems can enhance clinical diagnostics and combat antimicrobial resistance.