After doing deals with big pharma – Eli Lilly and Co., Merck & Co. Inc., Abbvie Inc. and Bayer AG – Atomwise Inc. announced three partnerships last week with smaller preclinical companies to use its artificial intelligence (AI) platform to develop drugs for their targets.
The deals with larger companies typically come with up-front, milestone and royalty payments. "The way we work with small pharma companies is usually different. We're setting up joint ventures where we'll both succeed together as the science progresses," Abraham Hpeifets, CEO of Atomwise, told BioWorld.
Even with its recent emphasis on smaller companies, Atomwise hasn't sworn off large deals. It also announced a fourth deal last week with Hansoh Pharmaceutical Group Co. Ltd. Atomwise will help Shanghai-based Hansoh develop drugs for 11 targets that will bring in technology access fees, option exercise fees, royalties and income based on sublicensing or sale of assets that could reach $1.5 billion.
The company's Atomnet platform uses convolutional neural networks, the same AI technology used in speech or image recognition. Instead of a two-dimensional image of colors, Atomwise creates a three-dimensional grid of the atoms — carbon, oxygen, nitrogen, etc. — that define the protein structure, which it uses to predict small molecules that can bind to the protein. Multiple rounds of predictions validated by biochemical assays can dial in the drug discovery from hit discovery to lead optimization to a clinical candidate.
Heifets credits chemists for the boom in AI-driven drug development. "Over the last decade, there's been these massive successes by the synthetic chemists," Heifets noted. "The number of molecules you can get access to today is something like 11 billion molecules, so that is like 2,000 times big pharma corporate collections. And most of those molecules have never existed on this planet, but you can order them up for off-the-shelf costs and get them in six weeks."
Cleveland-based Oncostatyx LLC, which is coming out of stealth mode, is based on research on hexamethylene bisacetamide induced protein 1 (HEXIM1) by co-founder Monica Montano, a professor at Case Western Reserve University. HEXIM1 is a tumor suppressor with reduced expression in triple-negative breast cancer and other solid tumors. While looking for compounds that induce HEXIM1, Montano and colleagues discovered lysine-specific demethylase 5B (KDM5B), which when inhibited, reactivates expression of HEXIM1.
Oncostatyx has lead compounds that inhibit KDM5B at nanomolar concentrations in cancer cells but was looking to improve the pharmaceutical properties before moving into the clinic, so it contacted Atomwise, which led to the formation of the joint venture.
"It's a game changer for small startups. As a biotech in the middle of the country, we don't get access to a lot of VC money," Matt Lawes, CEO and co-founder of Oncostatyx, told BioWorld. "You don't need to raise millions and millions and millions of dollars to hire 20, 30, 40 chemistry staff."
Sengine Precision Medicine Inc., of Seattle, is also working with Atomwise through a joint venture to develop inhibitors to targets that are essential for the growth of mutant cancer cells. Sengine, which was founded out of work at the Fred Hutchinson Cancer Research Center, uses patient-derived organoids to look for targets that create synthetic lethality, where combining two inhibitors can kill the tumor cells.
"We have a really rich portfolio of targets," Carla Grandori, founder and CEO of Sengine, told BioWorld. "But in order to develop drugs efficiently against our dozens of targets, we needed a partnership that would allow us to go through these targets in a much more efficient manner than the traditional empirical small-molecule testing in biochemical assays."
The companies will start with an undisclosed target that is synthetically lethal with KRAS, but Grandori said she sees the potential for the joint venture to develop a "true pipeline" of drug candidates.
By using AI, Grandori notes that the joint venture can screen through billions of potential inhibitors. "The chance of identifying a molecule that fits in the structure of our targets is much higher and much faster and, of course, less costly," she said.
And even screening millions of compounds the traditional way wouldn't be particularly feasible since patient-derived organoids are a limited material. "The plan is to arrive to candidates in the order of 100s, which is much more doable and less costly," Grandori noted.
In the third joint venture announced last week, Atomwise is partnering with Atropos Therapeutics Inc., of Belmont, Calif., to identify and develop compounds for undisclosed targets that modulate cellular senescence. By inducing senescence, the companies hope to stop growth of tumor cells.
Each of the joint ventures are set up to develop clinical candidates, but at that point the companies will need to make a joint decision about who will bring the candidates into the clinic.
"We can plan whatever we want, but the outcome is going to be determined by the science we have in hand," Atomwise' Heifets said of future plans for the joint ventures. "We're going to have to see when the time comes."
"I think both sides would be looking for licensing partners," Oncostatyx's Lawes opined. "Obviously, Oncostatyx would like to be positioned to take things forward into a clinical trial, IND, etc. That would require us to go through another round or two of raising [capital]."
Sengine's Grandori said she also thinks licensing the drugs out or co-developing them with a larger company would be the best approach. "We have strong roots at the Fred Hutchinson Cancer Center and connections with many oncologists, so we would like to still be involved in the clinical development, especially if we could couple it with testing organoids from patients to ensure success of the trial."