ASAPDiscovery-Polaris-OpenADMET Blind Challenge

[hmacdope]

Polaris Link

• https://polarishub.io/datasets/asap-discovery/antiviral-admet-2025-unblinded
• https://polarishub.io/datasets/asap-discovery/antiviral-ligand-poses-2025-unblinded
• https://polarishub.io/datasets/asap-discovery/antiviral-potency-2025-unblinded

README

The real-world drug discovery data in this package comes from the ASAPDiscovery consortium's pan-coronavirus Mpro (main protease) program. Primarily comprised of lead-optimisation data this data was used in the ASAP-Polaris-OpenADMET blind challenge. See the blog post (https://polarishub.io/blog/antiviral-competition) and results page for more details (https://polarishub.io/competitions/asap-discovery/antiviral-drug-discovery-2025)

A high level overview of the program and the resulting outputs can be found on the ASAPDiscovery website here. Data spans four main modalities and is split over two protein targets SARS-CoV-2 Mpro and MERS-CoV Mpro.

ADMET

ADMET data was obtained from assays conducted at CROs across 4 endpoints: Mouse liver microsomal stability, Human liver microsomal stability, Kinetic solubility and LogD. Training data was split timewise while minimising leakage.

Potency

Potency data (in-vitro IC50s) was obtained from an assay run at the Weizmann institute. Data was split timewise while minimzing leakage.

Ligand poses

Protein-ligand complex X-ray structures were obtained from the high throughput beamline at Diamond Light Source. Training data consisted of SARS-CoV-2 Mpro structures from the COVID-Moonshot, while test data was SARS-CoV-2 Mpro and MERS-CoV-Mpro structures from active lead optimization.

Dataset Source

https://asapdiscovery.org/outputs/molecules/#ASAP-COV-MPRO

Dataset Curation

Not applicable: The data was generated by the ASAP Discovery consortium.

Dataset Completeness

• I confirm that I filled out at least the readme, source and curation_reference fields for my Polaris dataset.

Dataset guideline

• I confirm that I am aware of the dataset guidelines and have complied with them to the fullest extent possible.

Anything else we should know?

No response

External reviewers?

@JenkeScheen

[cwognum]

@hmacdope Thank you for submitting!

Referring back to our Dataset 101 Guidelines, we are happy to certify this!

• On relevance: Data was generated as part of an actual drug discovery program.
• On the data consistency: The dataset clearly stems from a consistent source, since it was generated internally.
• On obvious errors / ambiguous data: Several errors were identified and fixed throughout the competition by a number of different participants. That further strengthens my confidence in the quality of this dataset!

Curious to see how people will use this data!

Thanks again for the submission!