Clinical-stage AI company Insilico Medicine has nominated a novel small molecule inhibitor known as ISM9274 as a preclinical candidate for cancer treatment.
The company used its PandaOmics AI platform to analyze genomic data from more than 90 tumor types and identified CDK12 as a promising target for multiple cancers including triple-negative breast cancer, lung cancer, and pancreatic cancer.
Next, it used its AI drug discovery engine Chemistry42 to design ISM9274 to selectively inhibit CDK12 and CDK13. Preclinical studies showed ISM9274 demonstrated potent antiproliferative activity across 60 cancer cell lines representing 13 tumor types. It also showed efficacy in animal models as both monotherapy and in combination with other therapies.
“Our target discovery philosophy is to find an optimal balance between commercial tractability, novelty and confidence,” said Alex Zhavoronkov, founder and CEO of Insilico Medicine.
The company’s AI platform was integral to the process with its PandaOmics AI target discovery engine and generative AI drug design engine Chemistry42 involved. Additionally, its AI-based clinical trial prediction tool inClinico assessed the molecule’s likelihood of successfully transitioning from phase 2 to phase 3 trials. The AI capabilities allowed the company to streamline and accelerate the typical drug discovery workflow. “These three components in our end-to-end Pharma.AI platform work together in a constant feedback loop,” added Zhavoronkov.
“This platform is very good at finding connections between biological processes and diseases that human scientists might miss, and its analysis identified a strong association between CDK12/13 and many cancers with high unmet need,” he continued.
Its lead asset for fibrosis is in phase 2 trials in the U.S. and China. Insilico recently licensed a USP1 inhibitor to Exelixis for an $80 million upfront payment.
The bulk of the company’s pipeline is focused on cancer.
In a recent interview, Zhavoronkov shared more about ISM9274, the role of generative AI in discovering it and Insilico Medicine’s use of AI and generative models to design novel drug compounds. He also provides an overview of the additional preclinical studies needed to submit an IND application to start clinical testing of ISM9274.
Could you detail the process of designing the ISM9274 compound using Insilico’s generative AI small molecule engine, Chemistry42?
Zhavoronkov: We look to identify the best possible molecule out of generative AI that is likely to make a great drug – something that satisfies all the rules of classical medicinal chemistry but at the same time exceeds human intelligence and basic computational tools.
Once we identified CDK12/13 as the target, Insilico’s generative AI drug design tool, Chemistry42, was used to create ISM9274, a novel small molecule optimized to be a highly selective and effective CDK12/13 inhibitor.
Unlike searching for suitable molecules in a molecular library, generative AI allows us to achieve multi parameter optimization and optimize for 30 different properties at once. We need to be sure that the molecule hits the right target, it hits only that target, it is efficacious in a disease, it is safe, it is metabolically stable, it can be taken as a pill, it does not induce any unintended side effects, and it can penetrate the tissue that we are targeting.
We synthesized a number of molecules that fit our criteria, and tested them. All the data we receive from our testing is fed back into our generative systems to continue to generate better molecules – ultimately selecting the best as the preclinical candidate. In these early stages of drug development, we make a plan and a forecast of the probability of Phase II to Phase III transition. Our goal is to ensure that we pass efficacy testing in phase 2 human clinical trials and we already have incorporated that information into the decision-making process.
What were the significant findings from the preclinical studies involving ISM9274, and how did these studies validate the choice of CDK12/13 as a target?
Zhavoronkov: In in vitro studies, ISM9274 has demonstrated potent antiproliferative activities in more than 60 cancer cell lines in 13 tumor categories and showed particular effectiveness in triple-negative breast cancer and pancreatic cancer. The small molecule inhibitor also showed robust in vivo efficacy in multiple xenograft models – or animal models with human tissue – for both monotherapy and combo-therapy, which can be a more effective treatment path for some cancers. The molecule also demonstrated excellent safety and favorable Absorption, Distribution, Metabolism, and Excretion (ADME) properties – a critical part of the drug discovery process that impacts its likelihood for success.
How did PandaOmics contribute to the identification of CDK12 as a high-potential target for multiple tumors?
Zhavoronkov: PandaOmics, our target discovery engine, uses more than 20 selected models and 10 trillion data points. In searching for the best target, we used PandaOmics to analyze and process data between tumor tissue and normal tissue in over 90 disease-specific databases, as well as text-based analysis from clinical trials, grants and publications. PandaOmics identified a strong association between CDK12/13 and many cancers, including triple-negative breast cancer (TNBC), lung cancer, colorectal cancer, and pancreatic cancer.
Could you share more about the IND-enabling studies and the planned IND application for ISM9274, and how AI is aiding in this process?
Zhavoronkov: Before a drug candidate can progress into clinical trials, it must receive FDA approval as an investigational new drug (IND), which provides assurance that it has met the necessary safety requirements to be tested in humans. The application requires animal pharmacology and toxicology studies, relevant manufacturing information to ensure the drug can be adequately produced, and detailed plans for future clinical trials. We are actively pursuing these steps now for our CDK12/13 inhibitor and plan to file our IND application in the second half of 2024. While AI is not accelerating this piece of the process, all of the data from our tests is fed back into our AI platform, further strengthening its predictive capabilities.
Filed Under: clinical trials, Data science, Drug Discovery, machine learning and AI, Oncology