Large-scale ligand-based virtual screening for SARS-Cov-2 inhibitors using a deep neural network

Markus Hofmarcher, Andreas Mayr, Elisabeth Rumetshofer, Peter Ruch, Philipp Renz, Johannes Schimunek, Philipp Seidl, Andreu Vall, Michael Widrich, Sepp Hochreiter, Guenter Klambauer

ELLIS Unit @ LIT AI Lab & Institute for Machine Learning, Johannes Kepler University Linz, Austria

---

Due to the current severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic, there is an urgent need for novel therapies and drugs. We conducted a large-scale virtual screening for small molecules that are potential CoV-2 inhibitors. To this end, we utilized ChemAI, a deep neural network trained on more than 220M data points across 3.6M molecules from three public drug-discovery databases. With ChemAI, we screened and ranked one billion molecules from the ZINC database for favourable effects against CoV-2. We then reduced the result to the 30,000 top-ranked compounds, which are readily accessible and purchasable via the ZINC database.

Technical report:

Abstract+PDF (arxiv), Abstract+PDF (SSRN)

Training data set for machine learning or other methods:

Data set (if you use this data set, please do not forget to cite this work!)

Additional data sets for multi-task learning:

• LSC data set
• Tox21 data set

Predictions and ranking for DrugBank compounds:

Available on reasonable request for scientific purposes. Send request to: klambauer@ml.jku.at

List of 30,000 ranked compounds from ZINC

The list is available in csv format for automated processing and in Excel format for manual inspection.

Here are some examples of the top-ranked molecules:

ZINC000254565785	ZINC000726422572	ZINC000916265995	ZINC000916356873
ZINC000806591744	ZINC000178971373	ZINC000000155607	ZINC000016317677
ZINC000193073749	ZINC000769846795	ZINC000755523869	ZINC000763345954
ZINC000001448699	ZINC000016940508	ZINC000005486767	ZINC000005527649
ZINC000755497029	ZINC000746495682	ZINC000005719506	ZINC000002149503