Skip to main content
The European High Performance Computing Joint Undertaking (EuroHPC JU)

CAPHARD: Codon-Assisted Phage-Host Automatic inteRaction Discovery

50,000
Awarded resources (in node hours)
Leonardo Booster
System Partition
July 2024 - July 2025
Allocation Period

AI Technology: Generative Language Modeling 

 

According to the World Health Organization, the emergence of antibiotic-resistant bacteria is one of the top global public health and development threats of our time. A possible solution relies on using bacteriophages (or phages for short), viruses of bacteria that can kill their hosts in a very specific way. 

This extreme specificity is at the same time an asset and a challenge for phage therapy: it allows to design therapies with minimal side effects, but the phages used need to be selected in a time-efficient fashion based on the target pathogenic bacteria.

To achieve this personalized phage therapy concept, we will design an AI-powered platform that will be able to determine which phages of a phage bank are more likely to infect and kill a given target bacteria, using as only inputs their genomes. 

The computing power provided will allow us to train large language models (LLMs) on both phage and bacteria proteins, which will constitute the core of this AI platform. 

The project's ambitious goal requires a combination of skills (AI, biology, bioinformatics) and data (bacteria-phage interaction matrices) that is uniquely found in Phagos, and the LLMs developed within this project will pave the way towards AI-driven, effective personalized phage therapy.