Insilico Medicine Sees Potential Quantum Gains in Using Quantum Generative Adversarial Networks in Generative Chemistry
- Insilico Medicine uses quantum computing and generative AI to explore key candidate discoveries in the drug development process.
- The team added that they demonstrated the potential advantages of quantum generative adversarial networks in generative chemistry.
- Critical Quote: “Applying quantum computing in drug discovery has the potential to help reduce research and development time and costs.” — Min-Hsiu Hsieh, PhD, Director of the Research Center for Quantum Computing, Hon Hai Technology Group
PRESS RELEASE — Insilico Medicine, a clinical-stage generative intelligence (AI)-based drug discovery company, today announced that it is combining two of the fast-growing technologies, quantum computing and generative AI to explore the discovery of key candidates in the drug development process and successfully demonstrated the potential advantages of quantum generative adversarial networks in generative chemistry.
The study was published in May 13th at the American Chemical Society Journal of Chemical Information and Modelingleading journal in computational modeling, led by Insilico Taiwan center and UAE center focused on pioneering and building breakthrough methods and machines with fast-evolving technologies including generative AI and quantum computing to accelerate drug discovery and development, supported by University of Toronto Director of the Acceleration Consortium Alan Aspuru-Guzik And Dear Hi (Foxconn) Research Institute.
“We are pleased to reach this milestone in collaboration with Insilico Medicine. Quantum computing can be used to solve complex computational problems. The application of quantum computing in drug discovery has the potential to help reduce research and development time and costs,” said Min-Hsiu HsiehPhD, Director of the Research Center for Quantum Computing, Hon Hai Technology Group (Foxconn®)
Generative Adversarial Networks (GANs) are one of the most successful generative models in drug discovery and design which have shown remarkable results for generating data that mimics the distribution of data in various tasks. The classic GAN model consists of a generator and a discriminator. The generator takes random noise as input and tries to mimic the distribution of the data, and the discriminator tries to distinguish between fake and real samples. The GAN is trained until the discriminator cannot distinguish the resulting data from the actual data.
In this paper, researchers have explored quantum gains in small molecule drug discovery by substituting each part of MolGAN, the implicit GAN for small molecule graphics, with step-by-step variational quantum circuits (VQC) including noise generators, patch method generators and quantum discriminators. , and compare its performance and with the classic pair.
This study shows not only that trained quantum GANs can generate molecules like the training set by using VQC as a noise generator, but that quantum generators outperform classical GANs in the drug properties of the resulting compounds and goal-directed benchmarks. In addition, this study demonstrates a GAN quantum differentiator with only dozens of learnable parameters can generate valid molecules and outperforms classical pairs with tens of thousands of parameters in terms of the resulting molecular properties and KL divergence scores.
“Quantum computing is being recognized as the next technological breakthrough that will have a major impact on all of society, and the pharmaceutical industry is believed to be one of the first batch of industries to benefit from this advancement. This paper demonstrates Insilico’s first footprint in quantum computing with AI in molecular generation which underscores our vision in the field,” said Jimmy Yen-Chu LinPhD, GM from Insilico Medicine Taiwan, and corresponding author of the paper.
These promising results will further support Insilico’s UEA team to integrate the hybrid Quantum GAN model into Chemistry42, Insilico’s small molecule generation machine for more efficient and accurate results in AI-driven drug discovery and development processes. As one of the pioneers to utilize GANs in Insilico’s de novo molecular design published the first paper in this field in 2016 and the company has submitted 11 preclinical candidates supported by Pharma.AI’s end-to-end platform based on a generative AI model since 2021, three of which have entered clinical trials.
“To our knowledge, this is the first time in the industry to systematically replace every component of a GAN with VCQ and successfully generate the molecule. I believe this is also the first small step in our journey,” he said Alex Zhavoronkov, founder and CEO of Insilico Medicine. “We are committed to accelerating effective, high-quality therapy to patients to extend healthy productive lives for all people on this planet with the support of the latest technology. The Insilico UEA Center is currently working on groundbreaking experiments with real quantum computers for chemistry and looks forward to sharing Insilico best practices with industry and academia.”