A new key to unlocking drug discovery
Be it aspirin for headache, or statin for cholesterol, or amoxicillin as an antibiotic, there are small molecules that we refer to as drugs that can offer therapeutic remedy. These drugs achieve their effectiveness by interacting with micromachines of living organism and viruses namely proteins and nucleic acids. There are trillions of potential chemical compounds and some of these harbor the potential to fight disease. Infectious disease specialists are working to find the specific compound (key) that can bind to the right spot (lock) on a disease-enabling protein to block its activity. The key question: which of the small molecule among the trillion is the right one for a specific disease?
Given the range of possible molecule to protein combinations, finding the right small molecule that is able to bind strongly to a certain target site and inhibit its function is a time-intensive and challenging feat. Enter VirtualFlow, a new open-source software that performs screens, essentially matchmaking between molecules and proteins. The matchmaking is dictated by the principle of physics by which one can calculate the ability of a given small molecule to engage a protein.
Harvard Medical School researchers developed the VirtualFlow platform that tests compounds through computer simulations. Using Amazon Web Services (AWS) and an AWS Cloud Credit for Research grant, the researchers ran VirtualFlow in the cloud. “AWS provided us with research credits to demonstrate that VirtualFlow is able to run in the cloud, for which we are grateful. AWS provides all underlying components which are required to run VirtualFlow on AWS, such as the SLURM cluster (via AWS ParallelCluster) or fast cluster file systems such as Lustre. Virtual screenings can drastically decrease the time and resources required in the drug discovery process, and the vast computational resources on AWS can further speed up the virtual screening procedures,” said Christoph Gorgulla, a research fellow at Harvard University.
The platform runs docking programs that bring together three-dimensional models of 1.4 billion compounds with proteins of interest, one at a time. Billions of computations are processed to identify the cases where the compounds and proteins connect. The instances with the strongest binding affinity—when the virtual key (small molecule) fits the lock (protein)—are surfaced and then tested in a lab experiment.
VirtualFlow currently draws from the largest existing library of prepared compounds, but can screen larger libraries should they become available in the future. Built to be compatible with cloud-based platforms, it’s possible to run these simulations faster and more cost-effectively than ever before. The team’s goal is to democratize in silico screening—which is conducted via computer simulations—to accelerate drug discovery.
In response to the COVID-19 pandemic, the researchers have been using VirtualFlow to screen for compounds that could potentially target coronavirus proteins. Though the team acknowledges it’s more likely that repurposing existing drugs would deliver the most rapid solution if successful, the team believes it would be beneficial to explore all possible options in the long term to find the specific molecules that could protect against the virus.
VirtualFlow’s user-friendly functionalities were designed to meet the dynamic needs of the biomedical research community. “Computational drug discovery will play a significant role in the drug discovery pipeline. With conglomeration of factors from different disciplines of science, including the ability to determine high-resolution structures of proteins, the power of rapid chemical synthesis, increased accuracy of computational docking programs, and access to unprecedented computational power especially in cloud-based platform, computational drug discovery is poised to be a dominant force in drug discovery. Imagine you can buy a million lottery tickets instead of ten, that is the power of VirtualFlow to identify potent molecules that can disable a protein in a disease state. We hope that many scientists around the world who don’t have access to large computing clusters could use our open-source platform and leverage AWS to identify promising hits to cure several diseases,” said Haribabu Arthanari, assistant professor of biological chemistry and molecular pharmacology at Harvard Medical School and Dana Farber Cancer Institute.