Biotech company Insilico Medicine announced it had successfully utilized generative AI in its preclinical drug discovery process, using its Gen AI drug design engine, powered by NVIDIA Tensor Core GPUs, to generate innovative molecular structures.

The use of Gen AI led to the development of the drug candidate for idiopathic pulmonary fibrosis, a rare respiratory disease, which is now entering Phase 2 clinical trials in the United States and China.

Insilico claims by employing its commercially available Pharma.AI platform, it was able to achieve results in a significantly shorter time–just 18 months– and at a fraction of the cost compared to traditional methods.

The company employed a comprehensive approach to bridge biology and chemistry in drug discovery, using Pharma.AI to identify a target molecule for drug compounds, generate novel drug candidates, assesse their binding capabilities with the target, and predict the outcomes of clinical trials.

The entire process to develop drug INS018_055, which would have taken years and cost over $400 million traditionally, was accomplished in just two and a half years at one-tenth of the cost, the company claims.

Additionally, the company’s Pharma. AI platform, equipped with various AI models, including PandaOmics, expedited the identification and prioritization of disease targets.

Alex Zhavoronko, founder and CEO of Insilico Medicine, explains AI is reshaping every area of healthcare and generative AI, and while it is making many headlines, is just a part of the revolution.

“Most of our systems now look like Lego,” he explains. “You have generative, classification, and predictive building blocks that work with the different data types and you can arrange these in many ways for different applications.”

He adds Generative AI models are not mature for pharma–neither for target discovery, nor for generative chemistry or even clinical trials.

“To be useful for pharma, these models need to be trained on specialized biological and chemistry data and then trained by expert human trainers,” Zhavoronkosays.  “To be accepted by the pharmaceutical industry, you need to show that you can get a drug in human clinical trials – that is the minimum benchmark. Doing this requires substantial expertise in the pharma industry. That’s what we are doing.”

Insilico’s success in reaching Phase 2 trials represents a significant milestone for AI-accelerated drug discovery.

With promising drug candidates in its pipeline and ongoing clinical trials, Insilico’s achievements are an indication of the effectiveness and potential of generative AI systems in revolutionizing the drug development process.

Dr. Harvey Castro, a physician and healthcare consultant, says the most striking aspect of this news is the speed and cost-effectiveness of drug discovery using generative AI.

“Insilico Medicine, with the help of NVIDIA’s technology, was able to reach the first phase of clinical trials in just two and a half years after starting the project, which is significantly faster than traditional methods,” he notes. “Furthermore, they accomplished this at one-tenth of the cost, which is remarkable.”

Castro says this demonstrates the potential of AI in accelerating the drug discovery process and bringing life-saving treatments to patients more quickly and affordably.

“Generative AI has the potential to revolutionize the entire drug discovery process,” he explains. “It can be used to identify potential drug targets, generate novel drug candidates, predict how well these candidates would bind with the target, and even forecast the outcome of clinical trials.”

As AI models continue to improve and more data becomes available, he says we can expect to see even more sophisticated and efficient drug design processes.

Matthew Sanchez, vice chairman at RAI Institute, says it’s exciting to see generative AI being used to accelerate time-to-market and help companies save on development costs when creating new drugs.

“These improvements can have a tremendous impact on the overall health and well-being of millions of people,” he says. “Similar to the example from Insilico and NVIDIA, we may soon be able to expect generative AI to be leveraged for all drug development, which will have major implications on the healthcare industry.”

Sanchez adds this raises the question of how government regulators will respond to this approach, and if they are preparing to add any additional criteria to clinical trial processes moving forward.

“In the near- to long-term future, GenAI will play a critical role in creating and testing orders of magnitude and more variations of molecules than any existing methods, leading to faster breakthroughs,” he says. “Not only that, but GenAI will also be pivotal in driving the creation of net-new drugs that may have never been discovered using existing methods.”

He says while both possibilities are exciting, they need to also be weighed against the scientific, medical and regulatory communities’ ability to understand and adapt to these methods.

Castro agrees that while the opportunities of Gen AI-aided drug development are immense, there are also risks.

“AI models are only as good as the data they are trained on, and there is a risk of bias or error if the data is not representative or accurate,” he says. “Furthermore, using AI in drug discovery raises ethical and regulatory questions that must be addressed.”

He points out regulatory bodies are still grappling with the implications of AI in healthcare, and while some progress has been made in developing guidelines for AI in healthcare, the rapid pace of technological advancement means that regulation often lags.

“It’s crucial for regulatory bodies to understand the technology and its potential implications and to work closely with researchers and companies to ensure that the use of AI in drug discovery is safe, effective, and ethical,” Castro says.