Alex Zhavoronkov, CEO of Insilico Medicine, a startup that generates potential drugs using artificial intelligence, was recently given a challenge by one of his pharma company partners. His team would see how quickly Insilico’s AI could identify new molecules that bind with a protein associated with tissue scarring. Then they’d put the molecules to the test, synthesizing a few of them in the lab to see if the AI was onto something, or only dreaming.
Why the rush? It now costs $2.6 billion, by one estimate, to get a new drug to market, and pipelines are only getting slower and more expensive. There’s hope—and hype—that AI could help chip away at that figure by reducing the time and labor before a drug starts clinical trials. The idea is that the same techniques used to generate realistic deepfakes and deftly play Go might be able to decipher the complex rules of drug design and generate molecules from scratch.
There are signs AI has potential. In December, Alphabet’s DeepMind debuted AlphaFold, an algorithm designed to predict protein folding—an important step for identifying potential disease targets. It beat the longstanding competition in the pharmaceutical industry, handily. Still, some experts remain skeptical of whether AI can dream up molecules that are both effective and truly practical.
On Monday, the AI drug explorers got some validation with the results of Zhavoronkov’s challenge, published in Nature Biotechnology. The team, along with collaborators at the University of Toronto, took 21 days to generate 30,000 designs for molecules targeting a protein involved in fibrosis. They synthesized six in the lab, of which four showed potential promise in initial tests. Two were then tested in cells, and the most promising one in mice. The team found their AI-generated molecule was both potent against the targeted protein and also displayed qualities that could be considered “drug-like.”
There’s merit in AI specialists doing that kind of real biology, says Mohammed AlQuraishi, a systems biologist at Harvard who wasn’t involved in the research. “The big new thing about this is actually testing these predictions,” he says. Lots of people are designing machine learning pipelines to produce virtual molecules, but relatively few have published research validating the work in the lab. Insilico’s work takes an extra step forward, AlQuraishi adds, in showing that its AI can be tailored to generate molecules that not only bind to a particular target, but behave well in cells and animals. That’s necessary for any potential drug candidate.
“That’s what pharma wants to see,” says Zhavoronkov. The favorable results in cells and mice were a pleasant surprise; he’d expected the AI-generated molecules would require more tweaks and rounds of computations before they found one with potential.
“It’s cool to see AI trained to think a little bit like how a medicinal chemist thinks,” says Adam Renslo, a professor of chemical biology at the University of California-San Francisco who also wasn’t involved in the research. Computational drug discovery has traditionally involved brute force methods of looking through millions of potential structures, with limited payoff. “This algorithm involves a creative process, not a data mining process,” he says.
The AI-generated molecules appear novel, Renslo says, and some could even be called creative in design. But the paper is best regarded as a proof-of-concept, he notes. The molecules aren’t a slam dunk, and would take perhaps a year of lab work to refine—meaning the AI wouldn’t save a big pharma company much time, if any. Plus, while the system was impressive at generating lots of candidate molecules, it was working in a niche where there’s plenty of data for the system to learn from. “It’s a suitable place to start, but it would be harder for the AI to solve a drug discovery problem where there isn’t data to start from,” Renslo says.
Still, it’s a starting point. Even if algorithms can’t yet outperform a team of chemists, AlQuraishi says, the research demonstrates how AI could quickly generate promising leads for human researchers to pursue.
Insilico’s method builds on two forms of AI: generative adversarial networks, or GANs, and reinforcement learning. It works by looking at past research and patents for molecules that are known to be effective against particular the particular drug target, as well as other structures. The idea is to prioritize novel, but logical structures, and those that could be synthesized in the lab. That’s similar to what a medicinal chemist might do in reading the literature and piecing together molecular components. From the 30,000 potential designs, the team selected 40 that represented a range of structures, of which six were concocted.
Real-world validation is important in a field dogged by a sense of unrealized hype. There are no AI-generated drugs close to market yet. Systems like AlphaFold, while flashy, show off advanced techniques that are exciting to researchers, but unlikely to yield results that can be quickly turned into new drugs. In April, the field saw a high-profile failure when IBM stopped selling its “Watson for Drug Discovery” system, which sought to scour medical literature and genetic data for overlooked cures. The product had reportedly underdelivered for its initial customers.
Investors are plowing forward all the same, with more than $1 billion poured into AI drug discovery startups last year, according to data compiled for Bloomberg. The pharmaceutical industry has started paying attention to the AI drug discovery upstarts, too. Insitro, hatched at Stanford, signed a deal in April with Gilead, potentially worth $1 billion including royalties, to develop molecules targeting liver disease. The same month, Exscientia, a UK startup, announced it had produced a molecule with potential to treat COPD, a lung condition, that’s headed to clinical trials with GlaxoSmithKline.
Insilico has its own drug company partnerships, but chose to begin identifying targets and molecules on its own in 2015. Zhavoronkov says the system described in the Nature paper, now more than a year old, has since been expanded to generate molecules for other, trickier drug targets with less available data. The company, which focuses on “age-related diseases,” is also using its AI to identify new potential drug targets, including those for cancers as well as fibrosis and NASH, a type of liver disease. If the molecules identified by Insilico’s AI can be validated in preclinical trials, they’ll seek a partner for clinical tests. But it’ll be a while still before we see those AI-generated drugs. Zhavoronkov says the company is likely still a few years away from clinical tests.
Source: wired.com
No comments:
Post a Comment