AMR chemical collections

This page outlines our goal to create the reference database of purchasable compound collections with potential interest to the field of AMR

Background

We have plans to develop the Ersilia AMR Chemical Collections (E-AMR-CC) web app, initially focused on Klebsiella pneumoniae. E-AMR-CC will consist of a catalog of bespoke chemical libraries with predicted anti-Klebsiella activity. The resource will be extensible to other species in the future.

Current repositories

Below is a table of the current code repositories available at Ersilia that will be key to develop the E-AMR-CC resource.

Repository Name	Description	Relevance to E-AMR-CC
Ersilia Model Hub	Main CLI tool from Ersilia, offering easy-to-use functionality to fetch and serve AI/ML models for drug discovery.	The Ersilia Model Hub contains general-purpose models for drug discovery, which will be used to annotate the compound libraries.
ZairaChem	End-to-end modeling based on ensembles of descriptors and autoML tools.	For the supervised AI/ML models (phenotypic), this will be the main pipeline to build and distill models.
ChEMBL Tasks for Fine Tuning	Generic data collection tool from ChEMBL, with a focus on AI/ML-readiness.	This tool will be used to fine-tune TabPFN models, used especially in the modellability assessment.
ChEMBL Binary Tasks	Workflow to collect antimicrobial data from ChEMBL, including data preparation with LLMs, and modellability assesment.	The workflow has been tested on A. baumannii. It will now be used on Klebsiella pneumoniae.
A. baumannii Enamine REAL screening	Screening of the Enamine REAL 1B Lead-like library against 30 A. baumannii models.	This is an example repository demonstrating that billion-scale screening is feasible for anti-Klebsiella activity prediction models.
Mtb Targeted Protein Degradation	GC-ADDA4TB project aimed at discovering pan-engaging warheads against essential tRNA synthetases.	Structure-based approach containing robust detection of binding pockets in ensembles of structures. A similar strategy will be applied to Klebsiella targets.