Research Highlight

Models to identify drug candidates for COVID-19

doi:10.1038/nindia.2021.74 Published online 19 May 2021

The models could decrease the number of experiments required to identify potential drug molecules for COVID-19.

© DrugCentral

A suite of machine learning models can help identify compounds that could potentially inhibit the growth of the novel coronavirus inside host cells1.

The models, collectively named REDIAL-2020, help identify the compounds by studying their activities against several viral proteins, an international research team reports.

The models, the researchers say, may ultimately accelerate the identification of novel drug candidates for COVID-19, saving time and cost for drug discovery.

The scientists, including a researcher from the Birla Institute of Technology and Science in Rajasthan, India, developed the models by using experimental data on the proteins and enzymes related to virus infection, including viral entry, viral replication and their interactions with various host proteins.

Using the models, the researchers identified 27 compounds that show anti-SARS-CoV-2 activities. Of these compounds, the models correctly predicted the roles of 16 compounds. The models also allowed them to detect six compounds that could inhibit a specific viral enzyme that helps the virus to replicate inside the host cells.

Accessible from any web browser, REDIAL-2020 is available as a web application through the DrugCentral web portal, they write.  

The researchers say that it could prioritise desired steps before planning experiments, particularly for screening large molecular libraries. This, they say, could eventually decrease the number of experiments required to identify potential drug molecules for COVID-19.


References

1. Govinda, K. C.  et al. A machine learning platform to estimate anti-SARS-CoV-2 activities. npg.  Nat. Mach. Intell. (2021) Doi: 10.1038/s42256-021-00335-w