The conventional drug discovery process is very inefficient, long (takes approx. 8-10 years from research, development to market), complicated and expensive. Less than 12% of all the drugs entering into clinical trials end up in pharmacies, costing billions of dollars. Researchers estimate that approximately 15-20% of the cost of a new drug goes to the discovery phase absorbing hundreds of millions of dollars and 3-4 years of work. Faster computational power plays an integral role in expediting the overall process and bringing down the effort from years to months and reducing the cost significantly.
In the field of drug discovery, the need for computational power becomes absolutely critical as it can tackle several challenges- from accelerating the time to market of required drugs, managing the overwhelmingly large amount of data being generated through studies, mapping the right drug with the respective diseases to continually evolving with the changing patterns of health needs. These challenges must be dealt with the utilization of the right compute resources, obtaining simulations for scientific methods representing the underlying hypotheses, and automation of models that result in the right predictions and accurate solutions, thereby intensifying the need for supercomputers.
With the advent of massively parallel processing, the performance of supercomputers and related technologies has improved drastically resulting in significant advancement in the field of computational science, especially in pharmaceuticals. Its immense potential to perform a wide range of computationally intensive tasks and offer deeper insights into highly complex unexplored systems has enabled Supercomputers (combined with AI algorithms) to bring breakthroughs to the pharma industry.
The process of drug discovery employs several scientific domains such as molecular dynamics, chemo-informatics, quantitative relationship modelling, and computational analytics in order to obtain simulations for petabytes of molecular data through implementations of respective solvers on high-performance computing systems. It is also based on trial-error techniques which test a library of millions of molecules and further narrow them down to a few thousand that reflect some initial binding to a target protein. The best candidates are then refined as a drug that can be tested in animals or cell cultures after which a few (maybe 10) finally qualify for human clinical trials. The entire process requires handling of complex input and huge pharmaceutical big data sets at scale and analysing them through traditional systems & usual lab methods are not possible. Supercomputers, with thousands of processors working together in parallel to analyse huge amounts of data and large computation, process it much faster than a basic computer system. It can quickly calculate the binding energies between various small molecules and proteins, thus paving the way for a more rational approach to the discovery of efficient drugs with fewer side effects and at the same time making the entire process cheaper and less time-consuming.
Supercomputers or High performance computing systems, hence, prove to be effective in discovering relevant drugs from a list of existing drugs or finding a novel drug altogether during a sudden outbreak of an epidemic. With Computer-Aided Drug Discovery and Design (CDDD), researchers can make in silico improvements as computer simulations trigger the possibility of early drug discovery by manifold. Since the beginning, CDDD has made significant contributions in accelerating the initial stage of pharmaceutical research and facilitating the development of new drugs. Today, CDDD has evolved as an effective computational tool that helps in storing, managing, analysing, and modelling of compounds at almost every stage of a drug discovery project – from lead discovery, optimization, validation and to even preclinical trials.
Here are a few examples of how supercomputers are contributing to various aspects of drug discovery:
Laboratory for Computational Molecular Design at the RIKEN Center for Biosystems Dynamics Research(BDR) hosted a new supercomputer of 1.3 PF, dubbed MDGRAPE-4A, which is dedicated to conducting molecular dynamics simulations for drug discovery. It models approx. 100,000 atoms that make up the proteins and water molecules, can carry out simulation of approximately 1.1 microseconds in a day which allows to model the interaction between drugs and protein in the body.
Exascale Compound Activity Prediction (ExCAPE) is developing an exascale supercomputer to quickly discover the right medicine using machine learning & AI.
The European High-performance Computing (EuroHPC) was established in November 2018 with the support of the EU budget to deploy world-class supercomputing infrastructure in EU for the development of personalised and precision medicine to make treatments more effective
In the US, a consortium of US government and technology companies like IBM, Alphabet, Amazon etc. is providing free access to its supercomputing resource which is helping researchers team to aim and quickly find the drug to combat COVID-19.
Researchers at Argonne National Laboratory, in the U.S., is harnessing AI with physics-based drug docking and molecular dynamics simulations, and utilizing some of the most powerful supercomputers of the world to reach their ultimate goal of finding the most promising molecule that binds strongly to the protein.