Daphne Koller combines a variety of titles and competencies — a specialist in biology, computer science and education, the founder of the educational unicorn Coursera.
Now, her startup Insitro, using artificial intelligence technology, is trying to change the approach to finding drugs.
If Insitro, using its revolutionary technique, finds a cure for a dangerous liver disease, it will receive $ 1 billion from Gilead Sciences
Few scientists ask for a joint photo, but with Daphne Koller this happens all the time.
“This happens at almost any event where people from IT come,” says Koller. “I’m a little embarrassed.” Not that I think I deserve it. ”
Selfie requests are just one of the hallmarks of Koller’s stellar status. She came to him after twenty years at the intersection of computer science, biology and education.
Along the way, she was awarded many titles: at the age of 18 she received a master’s degree from the Hebrew University of Jerusalem, at 26 she became a teacher at Stanford University and worked on machine learning, almost ten years later she won the MacArthur “grant for geniuses” to conduct research that combined artificial intelligence and genomics. And finally, she co-founded Coursera, a platform that was valued at $ 1 billion, and one of the first to allow people from all over the world to take university courses for free.
The next project of the 51-year-old scientist is Insitro, a company from San Francisco, which is going to find new drugs by analyzing data arrays. If the company succeeds in this, the approach to finding drugs for the treatment of diseases will change dramatically.
Biologists who work in laboratories usually select several specific proteins as targets for drugs. If these attempts fail, data scientists put forward new assumptions. Insitro, by contrast, wants to collect a much larger amount of data before biologists go hunting. The company will take advantage of bioengineering advances (such as Crispr genome editing technology) and software that allows computers to see what people don’t notice.
“Now machine learning is delivering incredible results if you have enough data. We finally have the opportunity to create large volumes of biological data, ”says Koller.
Insitro Computing Experts and Biologists are working together to develop laboratory experiments that would provide huge, specialized data sets. Machine learning models then find patterns to offer new tests and possible cures. Automated dosing devices reduce the impact of the human factor. All of this, according to Koller, will help Insitro conduct experiments «in weeks, not years.»
According to her, artificial intelligence, combined with biology, its long-standing area of interest, was the «best choice» for investors. For six months, Koller raised $ 100 million from ARCH Ventures, Andreessen Horowitz, Foresite Capital, GV, Alphabet Venture Fund and Third Rock. Later, Jeff Bezos joined in with investors. In April, Koller made a deal with Gilead Sciences, under which Insitro will receive $ 15 million now and $ 1 billion later if the company helps find a cure for the deadly form of non-alcoholic fatty liver disease. This disease in the near future may become the main cause of liver transplantation.
“Few people understand both aspects of the project,” says Mani Subramanian, who leads clinical research on liver diseases at Gilead. “Both in biology and in deep learning.”
Gilead’s future benefits to Insitro depend on whether the company can identify the five proteins that will be targeted by the drugs and whether the release of these proteins will lead to the development of certified drugs to treat liver disease. Deferred payments, which include the share of profits from successful drugs, allowed Insitro to get into the first Forbes AI 50 rating, which includes the most promising companies working with artificial intelligence.
More than 20 other startups are striving to fulfill the dream of finding new drugs faster and cheaper with the help of artificial intelligence. These included Notable Labs with $ 55 million in venture capital and Verge Genomics with $ 36 million. Novartis announced a five-year AI collaboration with Microsoft, while Merck and GSK are partnering with startups.
Artificial intelligence does not simplify the laws of biology. “I don’t think the platform is capable of magic,” Koller admits.
Before Insitro begins to make a profit, it will take several hundred thousand laboratory tests. Koller has enough energy. Moving around the Insitro office, she gave her chair to one of the 53 employees because she herself had never used it — she moves from a room called Macrophage (white blood cells) to Elastic Net’s room