Clinical trial of the first drug designed with artificial intelligence could begin this year, Google DeepMind’s CEO Demis Hassabis said Tuesday.
Speaking at a panel discussion at the World Economic Forum, Hassabis said that he expected to have “some AI-designed drugs in clinical trials by the end of the year… That’s the plan.”
Isomorphic Labs had been trying to speed up the creation of medicines using machine-learning since 2021. He said.
AI is a hot topic, even though Hassabis, his colleague John Jumper, and their work AlphaFold – a deep-learning system that predicts protein structures – earned them a Nobel Prize.
Pharmaceutical firms are interested in AI as it can save them a lot of time and money. According to an article published by the Journal Nature Medicine in a recent issue, it can take up to 15 years to create a new drug. It also costs approximately $2.6 billion.
As fewer than 10% of clinical trials where humans consume the drug are successful, many drugs never get approved for use. Pharma companies will benefit from anything that can reduce costs, accelerate development, or increase the success rate.
According to researchers, machine learning models have many ways that they can improve and speed up parts of the drug discovery processes. Hassabis believes that huge savings in both time and cost are possible.
This optimism needs to be moderated because it is difficult to find high-quality data for training due to privacy regulations and data-sharing policies. Hassabis is confident that these challenges are not insurmountable. He said.
This could be done through collaboration with clinical research organisations or by using synthetic data. He said AlphaFold2 did this extensively. Synthetic data can pose problems, as we have discussed in the past . Hassabis stated.
“You’ve got to be very careful if you’re using synthetic data, that it’s actually correctly representing the distribution and you’re not somehow training on your own errors,” Hassabis said. He said. Hassabis’s work on the application of machine-learning to drug discovery is not unique. Nvidia is also enthusiastic about AI-augmented drug discoveries, perhaps because they will have more reasons to purchase their hardware.
In the northern autumn of last year, Nvidia opened sourcedthe BioNeMo frameworks for GPU-accelerated machine intelligence. These frameworks are used in drug development and molecular designing. The company has also taken measures to repackage models like DeepMind’s AlphaFold2 or MIT’s DiffDock2.0 as microservices in order to make them more easily consumable.
Nvidia also partners major pharma companies including Danish pharmaceutical giant Novo Nordisk to bring new research system online. Denmark’s Gefion Supercomputer, which uses machine learning in biological sciences and new treatment development, is an example of this. (r)
