These Wharton grads say they can use A.I. to predict the outcomes of drug trials. Chelsea Clinton is among their backers.
BioPhy says it can predict trial outcomes with 80% accuracy. Clinton's VC firm and others have invested $4.5 million.
A Philadelphia artificial intelligence startup says it can predict the outcomes of clinical trials with 80% accuracy, potentially helping drug companies to identify the most promising products and get them to market faster.
Chelsea Clinton is among the early backers of the company, called BioPhy, which launched its A.I. model this month. The startup announced it had secured $4.5 million in early-stage funding from several investment groups, including a venture capital firm that Clinton co-founded.
Artificial intelligence is fast gaining a foothold in health care, used in such diverse applications as interpreting scans, discovering new drugs, and managing electronic records.
BioPhy’s tool is designed to improve the efficiency of the long, costly process of drug development, chief executive officer Dave Latshaw II said. Drug companies can deploy the tool to evaluate their overall portfolio, and they also can use it to improve the designs of trials for individual drugs.
“If you’re looking at it proactively and saying we need to put a trial together for this particular drug, we can work with a company to optimize what that protocol would look like in order to optimize the probability of success,” he said.
The model’s 80% prediction rate came in a live test on more than 1,500 drug trials in the past 27 months, said Latshaw, a former Johnson & Johnson executive and a computational biomolecular and chemical engineer.
In addition to Clinton’s Metrodora Ventures, investors include New York-based Audere Capital, Miami-based TRCM Fund, and Jeff Marrazzo, the former chief executive officer of Philly-based gene-therapy firm Spark Therapeutics.
Metrodora describes itself as “a values-conscious venture capital firm backing early-stage businesses in health and learning.” Its other recent investments include a startup that equips seniors with communications technology and another that facilitates affordable housing to improve health outcomes.
Clinton’s questions
Latshaw said that when BioPhy met with Clinton and her colleagues to discuss a potential investment, the former First Daughter asked smart questions about how the A.I. technology fit her firm’s mission.
“She’s very engaged,” he said. “She’s very specific.”
As A.I.’s applications expand in health care, some of the tools are a work in progress. For example, one model predicted that 18% of hospitalized patients would develop sepsis, a severe inflammatory condition, yet most of them did not actually do so, University of Michigan researchers found in 2021.
Using A.I. to guide drug development — the concept that BioPhy is pursuing — has potential, said Naomi Scheinerman, an assistant professor of bioethics at Ohio State University College of Medicine. But she warned that such a tool could encourage companies to stick with tried-and-true categories of drugs at the expense of new approaches that might serve unmet needs.
“I would worry about it delivering only the positive outcomes, because that’s what the drug developer is looking for,” she said. “They’re looking for what’s going to be the big sale.”
Not a black box
Asked about the issue Scheinerman raised, Latshaw said the model is able to evaluate both novel and established classes of drugs, as it was “trained” on data from more than 450,000 completed trials from all three phases of testing.
In making its predictions, the model takes into account such factors as where the trial is to be conducted, the patient population, and whether the protocols are appropriate, Latshaw said. The results are delivered in a transparent fashion, so clients can see what data inputs led a particular drug trial to get a thumbs-up or, in a large fraction of cases, a thumbs-down.
“The system is not a black box,” he said. “It’s explainable.”
The model renders a yes-or-no prediction as to whether a trial will achieve its so-called primary endpoint — a principal goal that is established ahead of time, such as improved survival.
Latshaw co-founded BioPhy in late 2019 with Steven Truong and Dan Sciubba, a neurosurgeon. The three met as students in the Wharton School’s executive MBA program.
Based in One Liberty Place, BioPhy has 10 employees and plans to double that number by the end of 2024.