Insilico Medicine has demonstrated a breakthrough in Artificial Intelligence ("AI") and drug discovery -- linking together generative chemistry and biology for the first time -- to discover a novel preclinical candidate addressing idiopathic pulmonary fibrosis ("IPF") and be validated with multiple human cell and animal model experiments.
Often found implicated in a wide range of diseases and multiple organs such as lung, liver and kidney, IPF addresses a very broad medical need that affects hundreds of thousands of individuals worldwide.
The pinnacle of the deep learning revolution can be pinged to 2014 when deep learning systems accelerated by NVIDIA GPUs and AI software started outperforming humans in image recognition and generative adversarial networks were invented. It is also the year when we started the company. In 2016 we demonstrated that a deep learning system can identify a novel biological target from omics data with experimental validation. From 2017 and 2019 we consistently demonstrated that generative AI can invent and design novel molecules that work in human cells and in animals.
But the big question remained --- can AI design a novel molecule for a novel target that has no known inhibitors and has not been validated in a disease? And now we have successfully linked both biology and chemistry and nominated the preclinical candidate for a novel target, with the intention of taking it into human clinical trials, which is orders of magnitude more complex and more risky problem to solve.
To my knowledge this is the first case where AI identified a novel target and designed a preclinical candidate for a very broad disease indication. It is a major milestone for us as our ultimate moonshot is to go after senescence and we need to have many enabling AI technologies that help us understand and manipulate human biology in other chronic diseases."
Alex Zhavoronkov, PhD, Founder and CEO of Insilico Medicine
The company also announced the formation of a team of over 20 expert drug hunters and drug developers in Shanghai led by Dr. Feng Ren, former senior VP of biology and chemistry at Medicilon, and former head of chemistry at GSK, who joined Insilico in February as Chief Science Officer.
This team will be responsible for taking the AI-discovered drugs into human clinical trials and creating a broad portfolio of preclinical assets.
Moreover, since Insilico will have the financial resources to go after a limited number of novel targets, the company made its target discovery and generative chemistry systems available to the pharmaceutical and biotechnology companies via its Pharma.AI software suite.
PandaOmics target discovery system is available as software as a service and Chemistry42 small molecule generative chemistry platform is available for on-premise deployment since September 2020. So far, the most advanced pharmaceutical companies have started deployments, and some deployed Chemistry42.
PandaOmics is now used by multiple academic institutions and pharmaceutical companies specializing in novel target discovery.
The preclinical candidate is a first-in-class novel small molecule inhibitor of a novel biological target with an unprecedented mechanism of action (MOA). It demonstrated experimentally great in vitro and in vivo efficacy in preclinical studies for idiopathic pulmonary fibrosis and a good safety profile in the 14-day repeated mouse dose range-finding study.
To nominate a preclinical candidate, Insilico Medicine started with a set of 20 completely novel targets discovered by AI for fibrosis and narrowed down the target to specifically address IPF. Subsequently, Insilico generated a set of novel compounds to selectively inhibit the novel target.
The molecules had to be selective, bioavailable, metabolically stable, capable of oral administration, safe, and have many other properties of a good drug. The company also predicted the high-probability of success of the phase 2 clinical trial outcome in IPF.
The molecules were first generated using Insilico's Chemistry42 system, powered by NVIDIA V100 Tensor Core GPUs, that adopts a Structure-based Drug Design (SBDD) generative chemistry approach, tested in the human cell and animal models.
Subsequently, the molecules were re-designed using the Ligand-based Drug Design (LBDD) to optimize for additional properties and then tested in human cells and animal models. After a review by a large team of internal and external veteran drug developers specializing in fibrosis, a preclinical candidate was nominated, and IND-enabling experiments started.
From target hypothesis to preclinical candidate selection, Insilico was able to complete target identification, molecule generation, and validation through traditional laboratory experiments in less than 18 months and at a total cost of approximately $1,800,000 for IPF and $800,000 for other fibrotic disorders, with less than 80 small molecules synthesized and tested.
Insilico's AI-discovered and validated preclinical candidate for IPF achieves several multiple industry firsts in the fields of biotechnology and drug discovery:
- By proving the effectiveness of AI-imagined novel drug targets and molecules through successful validation with multiple human cell and animal model experiments, today's breakthrough marks the industry's first-ever scientific validation of AI for R&D leading up to preclinical studies.
- Linking chemistry and biology for the first time during drug discovery: Historically the steps involved in designing new drugs and validating their effectiveness through preclinical and clinical studies have been disparate parts of the drug discovery process.
- Insilico has achieved a new record speed and lowest cost for preclinical candidate selection -- enabling the dramatic acceleration of preclinical development while drastically decreasing drug development costs by millions of dollars.
"We're rewriting the playbook for drug discovery by being the first mover and leader in creating the first and only AI-powered integrated system for drug discovery," said Dr. Zhavoronkov. "By creating the first universal system linking all of the areas of drug development from target identification, small molecule design, and soon clinical trial outcomes prediction, Insilico's AI platform will be capable of supporting every step of pharmaceutical R&D."
Industry commentary and additional information
"AI is transforming the healthcare industry, enabling breakthroughs that can improve millions of lives. Using NVIDIA's AI platform, Insilico Medicine, a premier NVIDIA Inception member, has done what we only dreamed of a few years ago: applying AI to dramatically accelerate drug discovery. And it is doing so at a time when it's never been more critically important to bring the power of AI to every industry to solve our greatest challenges." -- Jensen Huang, CEO and founder of NVIDIA
"One of the most difficult steps and biggest mysteries in drug discovery is related to target validation, specifically identifying the targets that have a strong impact in a clinical setting. Insilico Medicine has managed to tackle one of the biggest mysteries in drug discovery through its AI endeavors." -- Dr. Tudor Oprea, Professor and Chief of the Translational Informatics Division at the University of New Mexico and experienced drug-hunter with 25 years of industrial and academic experience in drug discovery
"Speed is everything in drug development. At least 90 percent of the costs associated with getting a drug approved for human use are in late-stage clinical trials. With its AI-powered universal system for drug discovery, Insilico is enabling researchers to figure out how to fail faster much earlier during the many phases of the drug discovery process leading up to clinical trials before it gets too late." -- Dr. Charles Cantor, Professor Emeritus at Boston University, member of the Science Advisory Board at Insilico Medicine, co-Founder of Sequenom Inc., and co-Founder of Retrotope Inc.
"This achievement of Insilico Medicine is another piece of evidence that AI is a powerful tool for drug discovery. By using AI in as many steps of the process as possible, AI can significantly reduce the time and cost to developing effective therapies." -- Dr. Alán Aspuru-Guzik, Professor of Chemistry and Computer Science at the University of Toronto, and co-founder of AI companies Kebotix and Zapata Computing
Machine learning early warning system reduces non-palliative deaths in general medicine unit