Stage 1

ULTIMATE LIST OF LEAD COMPOUNDS BY SCREENING A BILLION MOLECULES AGAINST  COVID-19 | 33 DAYS

The Moonshot: Test molecules against SARS-CoV-2 at a scale never seen before by computational screening of enormous libraries against high-resolution protein structures using supercomputing and potentially machine learning. Use cross-correlation of results within and among teams to determine a “high potential” compound list of unprecedented quality: a fast track to clinical testing. The overall time to find active compounds reliably will be massively shorter when compared to that of current methods.

The goal : In this stage we aim to improve the accuracy of simulations by adopting an approach that has successfully been applied by climate scientists. That is, different simulation approaches are compared to one another and in this way errors in assumptions are averaged out. The remaining “ultimate list” of compounds for each protein target can then quickly be screened experimentally and will be made available publicly.

​

Details:

• 250.000 Euros for the top-ranked team or consortium that uses three independent computational methods to select from a library of > 1 billion compounds and provide a consensus list[1] of the top 10.000 lead compounds (ranked by affinity) for at least 3 Covid-19 targets[2]that are involved in viral entry or replication. Computational methods may include, but are not limited to, molecular dynamics, deep learning, deep docking, virtual flow, etc., and should consider ADME-Tox[3] qualities where possible. The screened compounds must include known FDA-approved compounds[4] and can optionally consider biologicals. The aim is to find good binders with < 100 nM affinity. If needed, teams will be asked to provide additional data or clarifications.

• Strong evidence of reproducibility, such as publicly-available source code and/or submission to a peer-reviewed journal is desirable. Rapid open-access sharing without embargo periods, for example via preprint servers, is requested. Available experimental data4,[5] should be taken into consideration.

• Teams need to show that the relevance and quality of protein targets are the highest possible and need to use the largest compound libraries available (by merging existing libraries). See our resources page.

• An “ultimate list” will be compiled for each protein target and will consist of (up to) 10% of cross-correlated compounds across teams (depending on overlap of the lists), and 90% of top-ranked compounds from each team’s list. To ensure fair competition, an equal number of compounds from each team will be used. This ultimate list will be made available commercially as a physical library by our partners (for example as a DMSO stock), taking into account synthetic feasibility. The winning team has the highest score of compounds on the ultimate list(s), where score = (number of compounds)*(1/Kd); threshold = 1e7 M.

• The dissociation constants of compounds on each ultimate list will be measured experimentally by our partners. Only compounds that can be reasonably made and targets that can be readily obtained will be considered. The program manager and the scientific committee will judge whether a team/consortium has satisfied the conditions for the prize to be awarded. If needed, teams will be asked to provide additional data or clarifications.

[1] Consensus list: a list of compounds that come out of each of the three different methods (i.e., cross-correlated). The format to be provided is *.csv using the template provided on http://covid19.jedi.group.

[2] Make sure to use high-quality protein targets. See for example: http://insidecorona.net ; http://covid.molssi.org . Consider glycosylation: https://doi.org/10.1016/j.bbagen.2019.05.012

[3] https://en.wikipedia.org/wiki/ADME

[4] Including at least the SWEETLEAD library (https://simtk.org/projects/sweetlead) , and expanded with as many (FDA-approved) molecules as possible.

[5] For example: https://www.diamond.ac.uk/covid-19/for-scientists/Main-protease-structure-and-XChem.html