Atomwise, a San Francisco startup that’s using artificial intelligence to hunt for compounds that could make effective new drugs, announced today that it raised a $45 million Series A round.
The small company, founded in 2012, mines the accumulated data about biochemical interactions to search for molecules with the properties needed to block faulty processes in the body that lead to disease. That digital matchmaking, using supercomputers, augments a longstanding method of new drug discovery—the automated, high-throughput physical screening of thousands of compounds for evidence that they could disable troublesome biomolecules in human tissues.
Atomwise has been partnering up with big pharmaceutical firms, biotechnology companies, and university research labs in an effort to speed up the discovery of new drug candidates for neurodegenerative diseases, cancer, and other disorders. In June, the startup also announced a collaboration with Monsanto to find compounds that might protect crops against pest infestations and diseases.
The funding was led by Monsanto Growth Ventures, the venture capital arm of Monsanto, along with DCVC (Data Collective), and B Capital Group. Joining in the round were previous investors Y Combinator, Khosla Ventures, and DFJ, as well as new investors Baidu Ventures, Tencent, and Dolby Family Ventures. Atomwise has now raised a total of $51 million since its inception.
The big influx of capital creates a long runway for Atomwise to scale up and aim for “many years of robust growth,” says company co-founder and CEO Abraham Heifets. The money will be used to expand Atomwise’s staff of 17—which includes 15 PhDs. The staff may double to keep up with demand as Atomwise seeks out new corporate partners and university research collaborators, he says.
In April, Atomwise started a program to motivate academic scientists to explore the drug-hunting potential of its technology. Researchers can apply to the company’s Artificial Intelligence Molecular Screen (AIMS) awards program by identifying the disorder they hope to treat, and the disease-causing biomolecules they want to defeat with a drug. Successful applicants will receive 72 compounds that Atomwise predicts are most likely to work as that drug. The scientists can then test those compounds in their labs.
Heifets says Atomwise hopes to collaborate with 100 such researchers in 2018, not only to assist their work on a wide variety of diseases, but also as a way of training Atomwise’s AI-powered AtomNet drug research technology so that it improves with experience.
The new staffers Atomwise plans to hire will add to the depth of knowledge needed to analyze highly complex biological systems with advanced computational tools such as deep learning algorithms, Heifets says. Atomwise’s new investors will also bring in further scientific and corporate expertise, he says. The startup is gaining three new board members: Matt Ocko, managing Partner of DCVC, Kiersten Stead, partner at Monsanto Growth Ventures, and Gavin Teo, partner at B Capital Group.
Heifets declined to say what valuation had been placed on the company under the terms of the Series A financing round. Atomwise isn’t disclosing whether it has brought in any revenue to date.
Details of most of Atomwise’s partnerships are confidential, so Heifets wouldn’t say which projects have yielded the most encouraging signs that a drug candidate surfaced by the company’s algorithms might end up as a new treatment. He could say that one compound Atomwise identified as a potential treatment for multiple sclerosis has been licensed by a U.K. pharmaceutical company, and others are being tested in animals. Under Atomwise’s partnership agreements, the young company may eventually share in the financial returns when new drugs or other products go to market, he says.
Heifets says Atomwise can not only flag promising compounds from among the 600 million that have already been synthesized and made available for physical screening, but can also evaluate many millions of virtual compounds that could be designed or imagined, based on the chemical characteristics needed for a particular drug to be effective and safe in the treatment of a specific illness.
If Atomwise can predict with greater than 99 percent accuracy that such a theoretical compound would work as a drug, that would justify the expense of synthesizing and testing it, Heifets says. Atomwise, which screens more than 10 million compounds a day, claims that its success rate at identifying potential therapeutic drugs is as much as 10,000 times higher than that of physical high-throughput screening.
Illustration courtesy of Atomwise