Pharmaceutical researchers have for years used computers to aid them in, say, modeling molecules or mining clinical records. This allows them to come up with candidates for potential drug therapies, often thousands of them at a time. The researchers then produce medicines and test them—first in petri dishes, then in animals, and, if the drugs continue to prove promising, in humans. This process can take a decade or longer and cost billions of dollars, and means that the road to a cure for cancer, diabetes, or Parkinson’s (or even better palliative treatments for these and other diseases), is a long one.
Palo Alto-based startup twoXAR aims to make that road a lot shorter, by turning the task of navigating through the forest of potential drug candidates over to an algorithm. And the company’s founders, whose background is in computer science instead of pharmaceutical research, say they are uniquely positioned to make it happen.
“Consider the auto industry,” says twoXAR cofounder and CEO Andrew Radin. “When you ask an auto engineer to make a car safer, he comes up with seatbelts and airbags and antilock brakes. But ask a software engineer, and he says, ‘Let’s replace the driver.’ Driverless cars did not originate at GM and Ford.”
“Computers can do a lot of things,” says twoXAR cofounder and chief business officer Andrew Radin. “But they usually augment what researchers are already doing. People working in drug discovery are life scientists, chemists, molecular biologists; they build molecular models, and then decide which ones to test, instead of letting computers crunch the data to make the decision about which candidates to investigate further.”
You may have noticed something peculiar about those last two paragraphs—besides their introduction to the general premise of the company. And no, it wasn’t a type. TwoXAR’s cofounders have the same name: Andrew Radin. In fact, twoXAR stands for “two times Andrew Radin.” And the two Andrews—CEO Andrew A. Radin and CBO Andrew M. Radin—connected with each other in a way that’s almost as serendipitous as some of the drug discovery processes they are trying to replace.
It all started because Andrew A, a decade ago, purchased the Internet domain AndrewRadin.com. In 2008, Andrew A, recalls, he got an email from then-stranger Andrew M looking to buy the domain. “I said no, go away, and he kept arguing that since I wasn’t using it, I should sell it. I ended up being pretty rude to get rid of him.”
At the time, Andrew A, who had BS and MS degrees in computer science from the Rochester Institute of Technology, had been living in Silicon Valley for some time, working at startups and starting companies of his own. Andrew M, with a BS in biochemistry and cell biology and a BA in economics from the University of California, San Diego, was working in investment banking and evaluating MBA programs.
Approximately one year later, a friend of Andrew M, as a prank, started an Andrew Radin fan club on Facebook; Andrew A was one of several other Andrew Radins from around the country who joined the group. So the two were now connected on social media, though Andrew A still thought of Andrew M as that creepy guy who tried to buy his URL.
In 2010, Andrew A traveled for a month in China as a tourist, and posted the occasional update on Facebook. A few months after he got back, Andrew M, who was moving to China for six months to study Chinese, Facebook messaged him and asked if he could call him to ask him for China travel advice. They talked for a few hours on the phone. “That’s when he evolved in my mind to ‘not a creep,’” Andrew A recalls.
They didn’t talk again until 2013. Andrew A was studying bioinformatics at Stanford; Andrew M was at grad school at MIT.
“An MIT class of mine was coming out to Silicon Valley on a study tour,” says Andrew M. “So I contacted him [Andrew A] and had him sit down with my class to speak about his startup experiences; that was the first time we met in person.”
That’s, essentially, how the two Andrews became friends. The story of how they came to start twoXAR was a different kind of convergence.
Andrew A had worked in software development at Nortel Networks and then America Online, but then dove into Silicon Valley startup life, serving as CTO or engineering lead at startups, as he recalls, in “all sorts of different industries: making phone calls, videogames, advertising solutions, geolocation,” from 2004 to 2013. None were unicorns—or anything close, though the geolocation company, Locationary, was acquired by Apple and the acquisition gave him enough cash to not work for a while.
By 2013, he was tired of focusing on technically interesting but not exactly earth-shaking developments, and wanted to do something meaningful. Initially, he’d turned to considering the problem of homelessness, but his wife encouraged him to do something, he said “where your skills would make a difference.” He eventually settled on biomedical informatics, that is, using computer science to solve medical problems. Realizing he needed to go back to school to learn more about that field, he enrolled in a graduate program at Stanford.
Meanwhile, Andrew M, 13 years younger, had always had the startup bug. “My goal,” he says, “was always to start a biotech company. I was doing genetic engineering when I was in high school and chose UC San Diego because there were a lot of biotech startups that came out of it; I majored in biochemistry and economics.” Andrew M’s first job out of college was working for a bank in tech investing; then he enrolled in MIT’s graduate business school where he focused on entrepreneurship.
So, in 2014, the two Andrews, had one more thing in common: they were both in graduate school.
Over at Stanford, Andrew A, taking BMI 217, was assigned to come up with an idea for a medical problem that could be solved by computer science. He came up with a methodology to use data science to find new drugs. Essentially, the idea involved an algorithm for merging and mining a number of disparate biomedical databases to make predictions about what drugs were most likely to be successful in testing as particular disease therapies. The databases used included, but weren’t limited to, clinical record searches, molecular similarity models, and gene expression information, among others. (Clinical record searches examine individual medical records of people taking a drug for one disease to see if they have unusually low rates of any other diseases or symptoms that indicate efficacy; molecular similarity models look at drugs already in use and try to find other substances with similar structures, gene expression looks at certain gene signatures and matches them with molecules known to regulate them).
“It was essentially a tool for comparing relationships between huge and disparate data sets to distill signal from noise,” he says. The technology can combine any number of public and proprietary data sets to rank the probability of a particular drug candidate being effective in treating a particular disease.
Initially, he planned to simply write a paper about his methodology. Then he had lunch with V. Paul Lee, former president of Electronic Arts, who had invested in some of Andrew A’s startups “Paul gave me the ‘You’re an idiot’ look and said, ‘If you write a paper nothing will happen; if you want something to happen you need to write a patent application and start a company.’”
So, Andrew A says, he continued for months grinding away at his dilemma—paper or company? Chatting on the phone with Andrew M, he reviewed his dilemma, and asked him if he thought a company could be built around the technology. As Andrew A recalls, “He told me to come out to Boston and whiteboard it with some MIT professors.” He did, in June of 2014.
The professors helped the two Andrews see that there was indeed a business opportunity. They made the final decision to start twoXAR while sailing in one of MIT’s Tech Dinghy’s on the Charles River and incorporated the company in June of that year. In July, Andrew M came out to Silicon Valley, moving in—temporarily—with Andrew A and his wife.
Today, twoXAR has ten employees, $4.3 million in venture funding, and a sweet deal on office space in Palo Alto. They are hiring, and by the end of the year expect to add a few key team members in data science and bioinformatics.
The company has been working with research institutions and biopharmaceutical companies to test their prediction algorithm on known drugs, that is, running sets of data though their algorithms and seeing if the algorithm includes existing drugs in its set of drugs that it determines are promising candidates. So far, they’ve evaluated over 40 diseases in this way, including cancer, autoimmune, and neurologic disorders. Based on those tests, they came up with four areas for their initial work, which the company has yet to publicly identify. In those areas, they have made drug candidate predictions and have begun testing those compounds in animals. The two Andrews, at this point, do not plan to sell the software to pharmaceutical companies; they believe it would take too long to change an entrenched culture that remains skeptical about letting computers lead decision-making. Instead, they are working to build their own pipeline of drug candidates, typically working with partners at research organizations.
They say several partnerships have been signed with academic and commercial organizations but only one is public. That one, announced last month, has twoXAR working with the Department of Dermatology at Stanford’s School of Medicine to identify drug candidates targeting lymphatic malformation, epidermolysis bullosa simplex (EBS), and other rare disorders.
Today, there is “a tremendous opportunity for finding more precise medicines for diseases, common and rare, that have inadequate or non-existent treatments,” says Andrew A. “Drug discovery to date has been the reverse of Moore’s Law; it is becoming logarithmically more difficult to find a new drugs and the costs keep going up.”
TwoXAR aims to reverse that trend.
And that domain name? The two now have a joint website at AndrewRadin.com.