Open-source software from OpenPandemics - COVID-19 helps the researchers

The OPN team is developing open-source tools that help improve the predictions run in WCG and benefit researchers worldwide.


OpenPandemics - COVID-19 was created to help accelerate the search for new candidates for the design of COVID-19 antiviral drugs. The project goal is to identify small molecules that bind to SARS-CoV-2 proteins by means of computational simulation. The most promising molecules are chosen to be tested in laboratories of the collaborators of the research team. Molecules with confirmed experimental activity will constitute the starting point for the development of potential drugs.

Currently, there are two small molecules that have been tested in humans for treating COVID-19 showing good results: molupiravir and ritonavir. The insurgence of drug resistance to existent antivirals, as is has been happening with HIV, can be addressed with multiple, diverse antivirals. For that, the availability of multiple alternatives for the design of different antivirals will be crucial for overcoming resistance and address the challenges that new SARS-CoV2 variants may present.

Open-source tools

The research team is actively building open-source software to help other scientists performing computational simulations to support drug discovery projects. For example, new features such as flexible residues, modifiable pair potentials, and a contact analysis that have been added in the latest two releases of the AutoDock-GPU docking engine (v1.4 and v1.5) were developed to specifically address the needs arising during a large scale project such as OpenPandemics - COVID-19, but will benefit the whole community of researchers using these tools.

Efficiently splitting the work between CPU and GPU

The difficulty of the docking search and with it the corresponding number of score evaluations necessary to produce good binding modes increases exponentially with the number of rotatable bonds in each molecule.
This increase in the number of score evaluations is less steep for AutoDock-GPU (AD-GPU) than it is for AutoDock4 (AD4) due to a better search algorithm (Adadelta) implemented in AD-GPU. Both AD-GPU and AD4 easily dock simple molecules with few rotatable bonds. However, only AD-GPU is capable of docking complex molecules efficiently. So in order to improve overall WCG OPN performance, starting with packages OPNG_0087175 and OPN1_0063952, molecules with six or fewer rotatable bonds are docked by AutoDock4, while molecules with seven or more rotatable bonds are docked by AutoDock-GPU. As a result, overall throughput as well as prediction quality are improved.

Lab testing

Almost 300 compounds have been ordered as of this update, with another round of selection under way, with the goal to expand this number. Testing continues on these compounds in collaborating labs.
For some compounds, the further validation of initial promising results is ongoing, and the team is cautiously excited about the data obtained, and is working on including the structural information gathered thus far to inform the analysis of the new docking results obtained from volunteers.