(Page 4 of 4)
speech analysis. Cortexyme, a Bay Area company developing an antibacterial drug to treat Alzheimer’s disease, is using Winterlight’s speech analysis as one measure of its drug’s effectiveness in a Phase 2 study.
And there are many others developing so-called biomarkers—indicators of underlying disease or risk of disease—based on other behaviors, such as changes in walking patterns, eye physiology and movement, or phone usage. Harper argues that voice is the best indicator because it’s the most complicated human activity, using nearly the entire brain and complex muscle coordination.
Tech teams are also building AI systems to analyze written responses to questionnaires and weed out potential placebo responders—more “gate guarding.” NetraMark in Toronto, for example, says it has identified unusual questions from a clinical questionnaire, or scale, on patient attitudes: “Do you believe in medicine?” or “Do you like your doctor?”
The pharma company that supplied the data set didn’t think those questions were important, NetraMark CEO Joe Geraci says. “But it ended up being fundamental,” along with other questions about energy levels and sleep patterns, Geraci says.
Are pharma companies buying it? Geraci says NetraMark has attracted the interest of companies looking to understand failed trials. The next step is to convince those starting new trials to incorporate NetraMark’s 30 extra questions into their patient interviews.
What’s In Your Circuits?
Others are going farther, using AI to rethink traditional psychiatric diagnoses—and, they hope, to tamp down placebo response. The idea is to recategorize people with mental disorders by examining not just their behaviors but their underlying brain circuitry and genomic profiles. The push to explore this recategorization is known rather drily as research domain criteria, or RDOC.
BlackThorn Therapeutics in San Francisco is taking the RDOC framework and pushing it into drug development. Using AI, the firm believes it can identify more precise tranches of patients—not just those with depression, say, but those with depression characterized by anxiety and an inability to feel pleasure (anhedonia)—then design a drug that changes their brain chemistry. It has worked on a study, for example, to map the brain circuitry common to schizophrenia and obsessive-compulsive disorder that regulates motivation and cognition. The results were fed into its own computational system that aims to match these circuitry profiles—what the firm calls “neurotypes”—with drug targets.
That’s just the first part of the challenge. Once Blackthorn thinks it has locked onto a target and has created an experimental drug, it must identify people who might be eligible without using costly brain scans. That means using a set of questions or tasks, honed through AI, that reveal the underlying brain biology, says CEO Bill Martin.
“If we can understand the relationship between behaviors and how those behaviors are represented at the brain level without putting you in a scanner, we can understand that you don’t just have schizophrenia, you have psychosis with cognitive impairment, or psychosis with anxiety—things people don’t think go together,” Martin says. “If I think you have bipolar and give you a bipolar questionnaire it’ll only reinforce what I already thought. But what if I asked you three or four types of questions across these categories and gave you a task on a phone or iPad?”
For now, BlackThorn is starting with a single therapy for mood disorders, which it has tested in a small safety trial in healthy volunteers. “But there are a lot of drug targets in those brain circuits,” Martin says. Like cancer or HIV, new brain maps might ultimately suggest combination therapies to cover more than one target.
To truly be precise, new mental health treatments must escape the trap of human subjectivity, says Dan Karlin, a psychiatrist and entrepreneur who spent five years at Pfizer integrating informatics and clinical practice. Current traditional diagnosis of, say, major depressive disorder (MDD) is “false precision,” he says, because “many pathophysiologies end up in the MDD bucket.”
But a diagnosis, say, of depression with anhedonia is no better because our definition of anhedonia is also imprecise, Karlin says. It remains to be seen if the biological precision that’s changing cancer care— genomics and other markers—can make its way to mental health. It’s certainly what BlackThorn is shooting for.
Outside of clinical trials, the power of positive expectations in healthcare can be a boon—what Karlin tells his psychiatry trainees the placebo effect is “the best drug we have. It’s what a lot of great doctors do anyway.” That is, develop a good rapport with patients, give them confidence in their care, have honest discussions about risks and benefits. Some people believe a placebo itself, taken knowingly by the patient, might help some neurological conditions.
But in the world of drug development, the placebo effect remains a bane, not a boon. Even Kim Witczak, wary of the drumbeat to have more drugs approved, agrees that “we need to rethink” the development process: “People’s lives can be impacted greatly—for good or for bad.”