Explore novel drug candidates for COVID-19

To help researchers generate potential new drug candidates for COVID-19, we have applied our novel AI generative frameworks to three COVID-19 targets and have generated 3000 novel molecules. We are sharing these molecules under an Creative Commons here.

How it works

Save and export your findings for further evaluation

Save and export your findings for further evaluation

SMILES strings, attributes, and the intermolecular relationships of the top 10 AI-generated molecules, along with their closest neighbors in the pool of generated molecules and/or in PubChem, can be exported in a .csv file in order to facilitate further exploration and characterization of the AI-generated molecules for potential COVID-19 therapeutics.

Save and export your findings for further evaluation

See relationships among molecules

See relationships among molecules

One can select an individual AI-generated, top-ranked molecule and the tool maps its structural relationships to similar molecules with other generated molecules and molecules in PubChem. The five closest neighbors of the selected molecule, along with their structure and attributes, are displayed. The most common sub-structure in the selected neighborhood is also shown.

See relationships among molecules

View related molecules and nearest match in PubChem

View related molecules and nearest match in PubChem

The top 10 AI-generated candidates can also be displayed in an interactive visualization, along with other generated molecules from the same COVID-19 and PubChem databases. Hovering over each molecule displays its molecular structure and relevant properties. The visualization captures intermolecular similarity in the space defined by a set of selected attributes. Through different combinations of relevant attributes, the molecular landscape can be explored in a variety of configurations. Users can also import additional molecules of their own by using PubChem ID or SMILES strings and project those imported molecules in this molecular landscape.

View related molecules and nearest match in PubChem

See relationships among molecules

Select a biological target and filter generated molecules by important characteristics

Out of 1,000 AI-generated potential candidates for a specific COVID-19 target, the top 10 candidates are selected and ranked using a specific filtering criterion on an attribute of interest. The molecules are displayed along with a list of attributes. These top candidates can be further filtered and sorted with additional criteria to select the most promising molecules.

See relationships among molecules

Why this research matters

The traditional drug discovery pipeline is time and cost intensive. It can take up to 10 years and cost as much as $2.6 billion for a new drug to reach market. To deal with new viral outbreaks and epidemics, such as COVID-19, we need more rapid drug discovery processes. Generative AI models have shown promise for automating the discovery of molecules. However, there are many challenges in applying existing generative AI frameworks to accelerate the design of novel drug candidates. Current generative frameworks are not efficient in handling design tasks with multiple discovery constraints, have limited exploratory and expansion capabilities, and require expensive model retraining to learn beyond limited training data, both in terms of targets and ligands.

For the last two years, we have been developing robust generative frameworks that can overcome these challenges to create novel peptides, proteins, drug candidates, and materials. We have applied our methodology to generate drug-like molecule candidates for novel COVID-19 targets. Our hope is that by releasing these novel molecules, the research and drug design communities can accelerate the process of identifying promising new drug candidates for coronavirus and potential similar, new outbreaks.

We also hope this work will demonstrate our vision for the future of accelerated discovery, where AI researchers and pharmaceutical scientists work together to rapidly create next-generation therapeutics, aided by novel AI-powered tools.

If you are interested in adding or exploring other molecular features, or target proteins, or would like to work with us on further validating the generated molecules, please contact us via email or join the cogmol slack community to make suggestions and contributions, and to ask questions.

Log in with your credentials

Forgot your details?