Researchers Use NSF Convergence Accelerator To Shorten Drug Discovery Timeline
A $960,000, nine-month National Science Foundation (NSF) Convergence Accelerator grant has been awarded to Penn State researchers to explore faster and more cost-efficient methods of discovering pharmaceuticals using quantum artificial intelligence.
The researchers will use a branch of artificial intelligence known as quantum generative adversarial networks (GANs) to explore chemical compounds on Noisy Intermediate Scale Quantum (NISQ) computers. GANs use two networks — a generator and a discriminator — in order to model data. The generator creates synthetic (artificially created) data and the discriminator flags if the data is real or synthetic. The generator is iteratively trained using the discriminator feedback. By the end of the training, the artificially created data is so realistic that it is classified as real by the discriminator. This method allows the research to use the GANs to generate and assess potential pharmaceutical drug candidates on the powerful NISQ computers.
“Existing drug discovery pipelines that rely on known compounds, such as products of plants and bacteria, take five to 10 years and cost billions of dollars,” said Swaroop Ghosh, principal investigator and Joseph R. and Janice M. Monkowski Career Development Associate Professor of Electrical Engineering and Computer Science and Engineering. “The next-generation discovery of chemical compounds will come from the unexplored region of the chemical space. Since the number of pharmacologically relevant molecules could be on the order of 10 to the 60th power, deep generative models such as GAN have been proposed to exploit patterns in massive datasets and distill salient features that characterize the molecules.”
However, GANs can be computationally expensive and inefficient in sampling, according to Ghosh. Because of that, this research will explore quantum GAN, which can perform sampling at a faster rate from the unexplored regions of the chemical space at less computing power.
“This research will converge researchers from multiple disciplines such as engineering, science, medicine and education in quantum-enabled drug discovery to make a tangible impact on our society,” said Ghosh. “Computational approaches to accelerate drug discovery is of tremendous economic and societal relevance. If successful, this project will address two of the Grand Challenges in Engineering listed by National Academy of Engineering, namely, engineering better medicine and engineering the tools for scientific discovery. The importance of these two challenges is amplified under the current global crisis due to COVID-19. We are excited about the prospects of quantum artificial intelligence in resolving these important challenges.”
The co-PIs and senior personnel, all of Penn State, are: Nikolay Dokholyan, G Thomas Passananti Professor in the College of Medicine; Amy Voss Farris, assistant professor of education; Sean Hallgren, professor of computer science and engineering; Mahmut Kandemir, professor of computer science and engineering; Morteza Kayyalha, assistant professor of electrical engineering; Mehrdad Mahdavi, Dorothy Quiggle Career Development Assistant Professor; Nitin Samarth, professor of physics; and Bhuvan Urgaonkar, associate professor of computer science and engineering.
A key aspect of the convergence research is the partnership with industry, academia, national laboratories and local government. The team will partner with companies in the quantum computing industry, such as IBM, Rigetti and D-Wave Systems, as well as companies engaged in big data computing, such as Microsoft, Nvidia and Raytheon. National labs, such as Oak Ridge and Sandia; academic institutes, such as Georgia Tech, Rutgers, Duke, University of California Santa Barbara, Nanyang Technological University in Singapore and Korea University; and commercialization partners, such as Ben Franklin Technology Partners and Life Sciences Greenhouse of Central Pennsylvania, will also be engaged to meet the end-user needs and make an economic impact with the research outcomes.
This grant is phase one funding within the convergence accelerator model. At the end of this nine-month phase, selected teams will proceed to phase two, with potential funding of up to $5 million for 24 months.
The initiation of this research theme was facilitated in part by a Seed Grant award from the Institute for Computational and Data Sciences.
Source: The Pennsylvania State University