Project:

Extension of the Ersilia Model Hub, a FOSS platform featuring pretrained AI/ML models for biomedical

The Ersilia Model Hub is a marketplace for ready-to-use Artificial Intelligence / Machine Learning (AI/ML) models for drug discovery in infectious and neglected diseases. It features open source AI/ML models from the literature as well as AI/ML models developed by the Ersilia team, and currently contains over 150 models. The internship focuses on 1) expanding the collection of AI/ML models available, 2) contributing to general maintenance of the software and 3) extending the capacities of the platform to enhance its performance. The model contribution pipeline has been streamlined thanks to the support of a team of GitHub volunteers, read more about the process in the contributor’s documentation.

It focuses on three key areas to strengthen the Ersilia Model Hub:

1. Model Expansion
2. Maintaining the Platform
3. Enhanced Performance

Resources:

• https://ersilia.gitbook.io/ersilia-book
• https://github.com/ersilia-os/ersilia/issues/1029#issue-2175655359

https://github.com/ersilia-os/eos4tcc

eos4tcc

The BayeshERG model, designed for predicting hERG channel blockers, combines robustness, reliability, and interpretability.

• Predictive model to assess likelihood of small molecules blocking the hERG ion channel.
• It is used for human Ether-à-go-go-Related Gene(hERG) ion channel.
• It predicts whether a given molecule is likely to block the hERG channel, potentially affecting heart function.
• pretrained using a large dataset of 300,000 molecules as a transfer learning exercise.
• Among 12 compounds with predicted probabilities exceeding 0.95, one showed strong hERG inhibition (IC50 < 1 μM), and three exhibited moderate inhibition (1 μM < IC50 < 10 μM).