There were two epiphanies that led Steve Truong to the cutting edge of medicine. One took place during his career in finance; the other happened during a trip to a bowling alley.
About eight years ago, Truong stumbled upon an idea that excited him: You could take tools like risk management and asset diversification and use them to help cure diseases, including cancer. He read a few white papers that made that case and began to do some modeling of his own. Putting the theory to the test in a spreadsheet, he came away a believer.
“Clinical trials fail at a very high rate,” Truong explains. Although more than 90 percent of drugs end up failing during clinical trials, the strategy was viable if you could hit on a few big winners. “ If you put enough of them together in a portfolio, and a couple of them progress through [to FDA approval], they’re going to pay for the rest.”
But the problem that Truong kept running into was a matter of human expertise. Evaluating a company’s portfolio of drugs, as a precursor to investment, was a time-consuming task that was fit for only a small group of highly-skilled experts. There was an information bottleneck. But if he could clear that hurdle, it could change lives.
The second epiphany arrived in 2019, when Truong enrolled in Wharton’s executive MBA program, having written his application essay on the same concept that’d been gnawing at him for years. Just weeks into the program, he met some of his new classmates for a networking event at a bar inside a bowling alley. That night, Truong discovered someone else in the program who’d been trying to solve the issue that had stymied him. That person was David Latshaw, a former technology executive at Johnson & Johnson.
“He said, ‘I’ve had this idea about using AI to predict the outcome of clinical trials,’” Truong recalls.
Latshaw was convinced the medical community could use advanced algorithms to better anticipate which drugs were dead-ends and which were truly worth the expensive, trial-and-error-laden process of attempting to bring them to market. Instead of a 1 in 10 success rate, what if the industry could even double that? “There are billions of dollars that are wasted and lives that are impacted,” Truong says.
“There are other AI companies here, and if we can capture this moment, we can capture the cell and gene therapy discoveries as well.” — Steve Truong, BioPhy
What came of those conversations was BioPhy, a company which Latshaw and Truong co-founded with a third member of the Wharton program, Dr. Daniel Sciubba, the chair of neurosurgery at Northwell Health, which is the largest healthcare provider in the state of New York. “How do we make the system better? How do we help people with these horrible diseases?” says Truong. “That was our mission.”
Four years in, BioPhy has amassed a high-profile roster of backers. Chelsea Clinton’s venture capital firm Metrodora Ventures was part of a $4.5-million round of seed funding that the company raised last year. Jeff Marrazzo, the former CEO of Philly gene-therapy giant Spark Therapeutics, is another backer. Fast Company recognized BioPhy’s AI technology in its 2024 list of World Changing Ideas. And EveryCure, the nonprofit led by Citizen of the Year Dr. David Fajgenbaum, has formed a close partnership.
In addition to improving patient outcomes and lowering both the cost and time of bringing drugs to market, BioPhy’s founders are trying to open minds to what AI can do for the greater good — and how it can help Philly innovators realize the potential of their discoveries.
AI for drug research
MIT professor Marvin Minsky, a pioneering theorist in the field of AI, made the following prediction during an interview with TIME in 1970: “From three to eight years, we will have a machine with the general intelligence of an average human being.”
We’re still waiting on that. However, in the decades that followed Minsky’s prediction, the field of AI made dramatic strides by shifting away from the goal of general intelligence – creating computers that think like humans – and refocusing around specialized tasks. AI systems mastered chess, for example, in the 1990s, and later, skills like genetic sequencing. And as those narrow goals were steadily conquered over the last 50 years, computing power simultaneously increased, allowing for more and more complex challenges to be taken on by AI.
Some industries have been slow to adopt AI, but the pharmaceutical industry is not one of them. Increasingly, drug researchers are leaning on AI algorithms to speed up the discovery of new treatments. One recent analysis suggested that AI is not only significantly faster at identifying new treatments for diseases, but also more effective at doing so. While human-discovered drugs have a 5-to-10 percent likelihood of achieving final approval, the ones found by AI so far have had a 9-to-18 percent success rate, according to the study.
There are now thousands of AI-discovered drugs in development — including a molecule that could potentially kill E. coli — but identifying them doesn’t guarantee that they’ll work. “That’s what gets the hype and where most of the capital is being deployed,” says Latshaw. “But what nobody is thinking about is what happens after discovery.”
To go from a lab to the shelves of a doctor’s office, a drug faces a gauntlet of challenges that includes multiple levels of clinical trials, regulators, quality-control testing, and the manufacturing supply chain. That end-to-end process, if successful, costs more than $1 billion on average per drug. “Most people think the molecule shows up, you have a few conversations, and then a few years later, it gets approved,” says Latshaw. “But it’s usually a 7 to 10 year process.”
While discoveries get the attention, BioPhy’s technology accelerates the other (less sexy) areas of development, including the path to regulatory compliance and clinical trials, which are the most expensive part of drug-making. Clinical trials are plagued by myriad issues: Researchers often choose an improper site — a location that doesn’t give them adequate access to a target population — which is only revealed once they begin. Another issue can be how researchers select the diseases their drugs should be tested in — say, running a clinical trial for a disease of a particular organ where the tissue lacks a specific receptor that’s required to interact with the treatment. Or, doctors simply fail to anticipate the side effects of a drug. BioPhy’s algorithms, which have been trained on data from tens of thousands of clinical trails — both successful and unsuccessful — are able to see around these corners, spotting mistakes before researchers have made them.
“What we’re doing today is enabling companies to be prepared for this future wave [of discoveries] that’s going to come through, where they have a significantly higher volume of molecules to deal with.” — David Latshaw, BioPhy
While not every negative outcome can be predicted in every trial, AI can do a more efficient job at anticipating them than people (even smarter-than-average humans with a MD). BioPhy’s patent-pending AI technology, according to its founders, can predict the outcome of a clinical trial before it finishes with an accuracy rate of 80 percent. It can also project the regulatory hurdles that researchers will face down the road.
For trials that are doomed to fail, BioPhy’s tools allow researchers to adjust their protocols and design ahead of time, saving time and money in the process. The company’s clients include two of the five largest pharmaceutical companies in the world, which are using the tech to better meet regulatory compliance. Other companies are deploying their AI in pre-clinical and clinical settings to predict the efficacy of their research portfolios.
“What we’re doing today is enabling companies to be prepared for this future wave [of discoveries] that’s going to come through, where they have a significantly higher volume of molecules to deal with,” says Latshaw.
But is there a downside to equipping drug-makers with much more certainty? In other words, by predicting would-be failures, is it possible that BioPhy — and like-minded solutions promising better efficacy in research and development — could result in a pharmaceutical industry that’s risk averse?
Latshaw believes that the opposite is true: Pharmaceutical companies, emboldened by cost-savings throughout their portfolio, will actually be incentivized to take on more intractable problems — not because of the likelihood of success, but because of the potential for widespread impact. “So as long as you’re considering all sides of the equation, as opposed to just the risk, I don’t see any challenges with us stifling innovation,” says Latshaw.
A boost to Philly as the center of medical research
The story of modern American medicine can’t be told without Philadelphia, which is home to the first hospital, the first medical school, and countless pioneering breakthroughs like the mRNA research behind the COVID-19 vaccine. But while our city is renowned for birthing medical discoveries and training innovators, we haven’t always been great at keeping homegrown talent or fostering an environment that allows for companies to scale up without leaving.
Right now, the city is facing the same struggle when it comes to cell and gene therapy, which started right here in Philly and has quickly grown into a mainstream part of the pharmaceutical industry.
Despite some indications that the business environment for these discoveries is improving locally — such as an expanding angel-investor network and Penn’s #1 ranking in “tech transfer” revenue — Philly is still lagging behind competitors in terms of capital. And the reality is that early-stage companies will continue to relocate to top-tier hubs like Boston, San Jose, and the Bay Area to be closer to those money centers.
But the emerging potential of AI like BioPhy’s technology might make geography slightly less essential. In January, a report by McKinsey Global Institute estimated that AI could generate $60 billion to $110 billion per year in economic value for the global pharmaceutical industry, largely by reducing the costs of R&D: “It can boost productivity by accelerating the process of identifying compounds for possible new drugs, speeding their development and approval.”
To be clear, the growth of AI won’t erase the venture-capital conundrum that exists within Philly’s medical ecosystem. But it’s making a positive difference. Audrey Greenberg, who is the co-founder of the Center for Breakthrough Medicines in King of Prussia — a contract development organization that supports biotech companies — has seen the impact of AI. “It really minimizes failure rates,” says Greenberg. “And then when it comes to cell and gene therapy development, there are ways you can use it at all stages … to discover new gene targets or design delivery vectors, or even predict patient responses.”
The business model of companies focused on cell and gene therapy remains difficult. However, each marginal improvement could make the difference between an early-stage company staying put and moving to where the money is. “Where cash is tight, you have to get it right. AI can only be our friend,” says Audrey Greenberg.
Philly could borrow strategies that have worked elsewhere to boost medical businesses, ranging from tax breaks to investment funds that are backed by public bonds. But another idea is voiced by Truong: “Just owning the Bio-AI space — that would be huge in drawing a critical mass here,” he says. “Philly politicians can recognize that there’s probably a moment, literally right now, to do this. It’s not just us. There are other AI companies here, and if we can capture this moment, we can capture the cell and gene therapy discoveries as well.”
