Deep learning analysis of imaging and metabolomic data to accelerate antibiotic discovery against antimicrobial resistance

Introduction

Antimicrobial resistance (AMR) is one of the most pressing global health problems of our times. To counteract AMR, we urgently need new antibiotics, particularly with novel modes of action (MoA).

Limitations of Current Screening Pipelines

However, while typical antibiotic screening pipelines can identify compounds that impair bacterial growth, they are unable to predict drug targets and MoA. This limitation necessitates time-consuming target identification steps.

Proposed Solution

By synergizing our expertise in microbiology, genetics, advanced microscopy, metabolomics, medicinal chemistry, computational biology, and artificial intelligence (AI), we propose to create a new pipeline at the forefront of the antibiotic discovery field. This pipeline will be capable of informing simultaneously on the bioactivity and MoA of new antibiotic candidates.

Methodology

Working with seven pathogens, our improved acquisition strategies for both imaging-based high-content screening and metabolomics will generate a massive dataset of rich multidimensional phenotypes of:

1. Libraries of genetic mutants
2. Bacteria exposed to a range of perturbants

This will be achieved at an unprecedented scale.

Data Analysis

Deep learning analyses will then enable us to explore these massive datasets to correlate chemical-induced phenotypes to those from mutants. This process will link drugs to genes to elucidate the target/MoA of new drugs.

Exploration of Chemical Spaces

This innovative pipeline will enable us to explore unique chemical spaces, including:

• Complex natural product extracts (without the need for isolation of individual components)
• Novel synthetic compounds

Testing and Impact

Promising candidates with novel MoA will be tested against drug-resistant clinical isolates and against a future pandemic 'pathogen X'. This demonstrates our pipeline as an AI-powered solution for achieving higher productivity in antibiotic discovery.

Conclusion

AI4AMR will provide the community with a new pipeline to efficiently screen large compound libraries to identify novel antibiotics and define their MoA and target, helping directly to combat AMR.