PPubMed21 Jun 2024Cited 8×

Ongoing Implementation and Prospective Validation of Artificial Intelligence/Machine Learning Tools at an African Drug Discovery Center

Jason Hlozek, Kelly Chibale, John G. Woodland

Abstract

and Prospective Validation of Artificial Intelligence/Machine Learning Tools at an African Drug Discovery Center Authors: ['Jason Hlozek' 'Kelly Chibale' 'John G. Woodland'] Artificial intelligence (AI) and machine learning (ML) are anticipated to accelerate drug discovery programs. Following our development of an end-to-end virtual screening cascade at the University of Cape Town (UCT) Holistic Drug Discovery and Development (H3D) Center, we report the ongoing implementation of open-source AI/ML tools for use in resource-constrained settings.

A figure from Ongoing Implementation and Prospective Validation of Artificial Intelligence/Machine Learning Tools at an African Drug Discovery Center — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

The H3D Center in South Africa implemented the open‑source ZairaChem AI/ML pipeline, achieving AUROC 0.77‑0.89 in prospective malaria/TB screens and creating new high‑performing target models.

What’s new

i.Developed and deployed the open‑source ZairaChem pipeline for automated AI/ML model training at H3D.
ii.Prospectively validated whole‑cell pathogen activity and solubility models in 2023 with AUROC 0.77‑0.89.
iii.Created new high‑performing models for PvPI4Kβ, PfPKG, β‑hematin, and hERG with AUROC >0.89.
iv.Integrated models into a GUI via StarDrop to lower the barrier for medicinal chemists.
v.Established a sustainable capacity‑building program for AI/ML adoption in an African drug‑discovery center.

How they did it

The team collaborated with the Ersilia Open Source Initiative to train AI/ML models on in‑house assay data, retraining biannually and prospectively validating predictions on new compounds; models were incorporated into a virtual screening cascade and accessed through a GUI plug‑in for StarDrop.

What they found

Whole‑cell pathogen activity and aqueous solubility models achieved AUROC between 0.77 and 0.89 in 2023 prospective tests.
New target‑specific models (PvPI4Kβ, PfPKG, β‑hematin, hERG) showed AUROC scores above 0.89.
Models trained on a curated set of 35,000 compounds from ChEMBL and CO‑ADD demonstrated versatility for broader screening.

Caveats

✕Microsomal intrinsic clearance (CLint) models dropped performance in late 2023 due to under‑represented chemical series.
✕Technical complexity and black‑box nature of AI/ML can intimidate wet‑lab scientists.
✕Initial lack of a graphical user interface limited accessibility for non‑computational users.
✕Model performance depends on the similarity of new chemical space to training data, requiring continual retraining.

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Ongoing Implementation and Prospective Validation of Artificial Intelligence/Machine Learning Tools at an African Drug Discovery Center

The paper, in five minutes

Ask a question about this paper

Similar papers in all four archives

Similar papers in all four archives

The paper, in five minutes