There are tens, perhaps hundreds of billions of compounds which can be utilised by the pharmaceutical industry in the development of new drugs. An exhaustive exploration of more than a handful experimentally is impossible, modern computational techniques are however now beginning to lead the way. Quantum Chemical calculations run on High Performance Computers (HPCs) allow researchers to quickly and efficiently determine the chemical properties of compounds and accurately predict how they will interact with viruses and bacteria. This allows them to significantly narrow down the list of molecules to investigate experimentally, and hence develop better drugs, more quickly. Given the vast quantity of potential therapeutic molecules however, an exhaustive exploration of all molecules in a timely manner, even on the worlds most powerful computers is challenging.
Researchers at the Argonne National Laboratory (ANL), University of Chicago have been investigating the use of state-of-the-art Machine Learning Techniques and Artificial Intelligence (AI) with common methods from Computational Chemistry to optimise the rate at which drugs discoveries are made, applying it most recently to enable the search for COVID-19 therapies.
In the first of our COVID-19 centric Webinars, Austin Clyde, a Ph.D student at ANL, presents the work he is currently carrying out. He provides an thorough, interesting and accessible overview of the use of Molecular Dynamics Simulations from Computational Chemistry in the assessment of potential therapeutic molecules, and continues to show how he is refining existing methods through the use of AI. It is shown through the work of Austin and others that drug discoveries can be made up to 200 times more quickly on existing hardware through the use of common Machine Learning Techniques to discard molecules which are unlikely to be suitable, and Deep Learning Techniques such as Residual Learning for Image Recognition.
The Webinar is available via the embedded YouTube video below.