San Francisco-based startup Atomwise has built an artificial intelligence system that they hope will help them generate new prescription drugs.
The system, called AtomNet, learns the interactions between molecules by looking at LOTS of pairings of 3D models of molecules and simulated drugs and figures out which are most likely to interact. As COO and cofounder Alexander Levy explained to Quartz: "You can take an interaction between a drug and huge biological system and you can decompose that to smaller and smaller interactive groups. If you study enough historical examples of molecules … and we've studied tens of millions of those, you can then make predictions that are extremely accurate yet also extremely fast."
It's kind of like a password-cracking tool: Humans could certainly type in all of the combinations for possible passwords and then try them out, but a machine is much, much faster. Atomwise doesn't find the password, it just narrows the possibilities way down. The company has already had some success with drugs for multiple sclerosis and Ebola; an MS drug has been licensed to an unknown pharmacology company in the UK and a paper on the Ebola drug is being submitted to a peer-reviewed journal.
The company is inviting scientists from universities and research institutes in the US and Canada to tell them which diseases they're trying to come up with treatments for and let AtomNet try its hand. Researchers have until June 12 to submit their proposals and Atomwise will notify up to 100 labs in September if they've been selected, in which case they'd receive, for free, 72 compounds that the program says have the highest probability of interacting with their target disease. Then the labs can test their algorithmically selected compounds to see what works and make sure it's safe.
Atomwise isn't the first group with this idea. Other companies are also looking into machine learning for drugs: Johnson & Johnson subsidiary Janssen is working with startup BenevolentAI to evaluate the potential of new drugs in development, while others still are working on specific drugs including new glaucoma medications and pancreatic cancer treatments.
It's unclear if this process would be cheaper than the way drugs are currently developed, but the high cost of drug discovery is partly why they created the program; it stands to reason that shortening the testing window could only help on the cost front.
Read This Next: The High Price of Insulin Is Literally Killing People