The research was published May 13 in the Journal of the American Chemical Society. Journal of Chemical InformInsilico Medicine (“Insilico”), a clinical-stage generative artificial intelligence (AI)-driven drug discovery company, today announced that it has combined two rapidly evolving technologies, quantum computing and generative AI, to form a leading drug discovery company. It announced that it had searched for a candidate discovery and had successfully demonstrated it. Potential advantages of quantum adversarial generative networks in generative chemistry.ation and modeling, A leading journal in computational modeling, led by Insilico’s Taiwan and UAE centers, it uses rapidly evolving technologies such as generative AI and quantum computing to accelerate drug discovery and development. The focus is on pioneering and building breakthrough methods and engines. The research was supported by Dr. Alan Aspur Guzik, Director of the University of Toronto Acceleration Consortium, and scientists at Foxconn Research Institute.
This international collaboration was a very fun project. This meets drug discovery and sets the stage for the further development of AI. This is a global collaboration between Foxconn, Insilico, Zapata Computing and the University of Toronto. “
Alán Aspuru-Guzik, Director, Acceleration Consortium, Professor of Computer Science and Chemistry, University of Toronto
Generative Adversarial Networks (GANs) are one of the most successful generative models in drug discovery and design, showing remarkable results in generating data that mimic data distributions for various tasks. A classical GAN model consists of a generator and a discriminator. A generator takes random noise as input and tries to mimic the data distribution. Discriminators also try to distinguish between bogus and genuine samples. A GAN is trained until the discriminator can no longer distinguish between the generated data and the real data.
In this paper, researchers explore the advantages of quantum in small-molecule drug discovery by progressively replacing parts of MolGAN, an implicit GAN for small-molecule graphs, with variational quantum circuits (VQCs). bottom. We compare the performance of patch methods and quantum discriminators with their classical counterparts.
This study not only demonstrated that a trained quantum GAN can generate training set-like molecules by using VQC as a noise generator, We have demonstrated that it outperforms classical GANs in benchmarks. Moreover, in this work, we show that quantum discriminators in GANs with only a few tens of learnable parameters can generate effective molecules, and tens of thousands of parameters in terms of the properties and KL divergence scores of the generated molecules. It was shown to outperform conventional quantum discriminators.
Quantum computing has been identified as the next technological breakthrough of great impact, and the pharmaceutical industry is believed to be one of the first wave industries to benefit from its advancement. This paper demonstrates his Insilico’s first achievements in quantum computing with AI in molecular generation and underscores our vision in this field. “
Dr. Jimmy Yet-Chu Lin, Principal, Insilico Medicine Taiwan, Corresponding author of the paper
Based on these findings, Insilico scientists plan to integrate a hybrid quantum GAN model into Chemistry42, the company’s proprietary small molecule generation engine, to further accelerate and improve the AI-driven drug discovery and development process. doing.
Insilico was one of the first companies to use GANs in de novo molecular design, publishing the first paper in the field in 2016. The company has delivered 11 preclinical candidates through his GAN-based generative AI model, and the lead program is in phase I validation. clinical trials.
Alex Zaboronkov, Ph.D., Founder and CEO of Insilico Medicine, said: “We believe this is the first small step on our journey. We look forward to sharing it with the academic community.”
