In the race to revolutionize human health, artificial intelligence (AI) tools are accelerating drug discovery and biomarker identification, yet the timeline for human clinical validation remains linear. At the Precision Medicine World Conference (PMWC), leading longevity and AI experts discussed the current roadblocks responsible for the drop in the pace of medical interventions compared to the exponentially-expedited discovery pipeline.
The Standardization Crisis
Sherry Zhang is the executive director of external strategy & partnerships at the Buck Institute.
Courtesy of Precision Medicine World Conference
Sherry Zhang, executive director of external strategy & partnerships, Buck Institute
First, we need to have a standardized definition of “what is healthspan?” After that, we can address how to measure it. Those are the two things that I see everyone we speak with—clinicians, policymakers, individuals, or biohackers—defines differently. To eventually have computable, accessible, and measurable outcomes, I think the definition of healthspan needs to be standardized first.
The Diversity and Scale Problem
Zeenia Framroze, chief executive officer, Alethios
Zeenia Framroze is the CEO of Alethios.
Courtesy of Precision Medicine World Conference
On my end, I would focus on very large—not even necessarily distributed—study designs that include a diverse population and capturing data that is multimodal. There are lots of different kinds of data that we can collect. We can ask much better questions, and we can do much better analysis now. But, people must be able to get into those trials, and we have to ask more questions and measure more than just biomarkers. That is actually representation because someone may answer something differently based on the context that they are in.
The Safety Trial Bottleneck
Nathan Price, researcher, Buck Institute
Nathan Price is an aging-research expert at the Buck Institute.
Courtesy of Precision Medicine World Conference
Our ability to mine literature, drive personalization, and drive depth is going up exponentially with AI tools, but there are many things that aren't expedited. Our ability to carry out safety trials, for instance, is not going up exponentially. We are making some progress with distributed trials and so forth, but there's a whole continuum of things that has to happen. As a group, we need to think about what it will take for us to really start to see breakthroughs in healthspan in a big way and recognize the barriers that need to be broken for that to happen.
The Causality Hurdle
Andrei Georgescu, co-founder and chief executive officer, Vivodyne
Andrei Georgescu is the co-founder and CEO of Vivodyne.
Courtesy of Precision Medicine World Conference
The frontier approach to identify potential targets in genetically evidenced medicine currently is to use a genome-wide association study (GWAS). These kinds of studies look at the correlation between the expression of certain genes and markers in patients and the association with disease in those patients. The really hard thing about that is we do not have causality. We don't know that A causes B; we just know that A and B kind of come together.
There are so many signaling pathways and feedback loop compensatory mechanisms that are active. So, there is so much association between a disease and all these biomarkers that seemingly drive it, but the causality of one specific marker solely being responsible for driving some phenotype is so difficult to find from this data when looking at post hoc analysis.
Outlook on AI in Precision Medicine
Most experts at PMWC agreed that the AI age will only deliver on its promise of increased healthspan if the industry can modernize the clinical pipeline to match its discovery engine. How quickly it happens for different fields remains to be seen.
Responses have been edited for length and clarity.
