The development of new drugs in medicine can take many years and cost billions of euros. An international team of scientists, including researchers from Freie Universität Berlin and Technische Universität Berlin (TU Berlin) have succeeded in developing a platform capable of examining billions of potential drugs in a short time. A paper on their work was published in the leading journal Nature. Inspired by the coronavirus pandemic, researchers have now begun to use the platform to search for potential drugs that could block the coronavirus proteins.
A proven principle of medical drug research is inhibiting disease-causing proteins in the body by introducing small molecules which bind firmly to the disease-causing protein at a specific location and thus manipulate or inhibit it. A principle comparable to a lock-and-key system. The molecules (keys) fit onto a specific bonding site (lock) of the disease-causing protein and disable it. The issue: There are potentially a huge number of more or less well-fitting keys (molecules) that could be the basis for a new medical drug and potentially all of them have to be tested. Computer simulations replacing cumbersome lab experiments have the potential to solve these problems.
“In a computer simulation, we can test a molecule’s ability to dock on a disease-causing protein. The process of doing this with many molecules is called virtual screening. This approach already exists. However, until now only a relatively small number of molecules could be routinely tested in such virtual screenings,” explains Christoph Gorgulla, a postdoctoral research fellow at Harvard University, who completed his doctoral studies at Freie Universität Berlin with a grant from the Einstein Center for Mathematics (ECMath) and whose work forms the foundation of the publication’s findings. “We looked for a way to virtually screen significantly more molecules than previously. We are not only interested in the classic differentiation of ‘fits’ or ‘doesn’t fit’ but also ‘fits better’ and ‘fits worse’. This docking strength is one of the most important properties of drugs.
A library of more than 1.4 billion molecules
To this aim, the team created one of the largest molecule libraries in the world which provides docking molecules by calculating the three-dimensional geometry of more than 1.4 billion molecules and recording it in the database. The data is open source and can be used free of charge by drug researchers around the whole world. In addition, the research team developed the "Virtual Flow" platform, which uses computer simulations to efficiently test not only millions but billions of molecules in parallel for their ability to bind to a specific protein.
“Virtual Flow runs on supercomputers with thousands of processors instead of a traditional computer. It can also be run on clouds,” say Dr. Konstantin Fackeldey, adjunct lecturer at the TU Berlin Institute of Mathematics, which cooperates with the Zuse Institute Berlin in the Math+ Cluster of Excellence and is part of the project team. “For instance, we tested Virtual Flow on Google’s Cloud Platform with up to 160,000 processors.” Virtual Flow even considers additional complexity. Proteins are flexible so that the specific three-dimensional docking site of the molecules on a protein can also change under certain conditions. This potential flexibility of the docking sites is taken into consideration in the mathematical simulation.
Google has also now provided the researchers with additional resources to use its cloud computing platform and thus search for possible candidates as drugs against the coronavirus. “Although virtual screenings can be very fast, we do not expect this approach to lead to an approved drug in the short term. However these screenings can significantly better lead the search for potential drugs. To be approved, a drug must still be tested in the lab and in clinical studies, which are relatively time-consuming,” emphasizes Konstantin Fackeldey.
An open-source drug discovery platform enables ultra-large virtual screens
Authors: Christoph Gorgulla, Andras Boeszoermenyi, Zi-Fu Wang, Patrick D. Fischer, Paul Coote, Krishna M. Padmanabha Das, Yehor S. Malets, Dmytro S. Radchenko, Yurii S. Moroz, David A. Scott, Konstantin Fackeldey, Moritz Hoffmann 4, Iryna Iavniuk, Gerhard Wagner, Haribabu Arthanari: https://www.nature.com/articles/s41586-020-2117-z