BioAge is one of a growing number of companies using machine learning methods to reduce cost and speed up discovery of drug targets, development of small molecule drugs, or peptides, or other aspects of traditional medical development than have been painfully costly, inefficient, and slow. Faster, cheaper processes in medical development are a benefit to humanity, and right now the novelty of this sort of work gives it a high profile in the entrepreneurial and investment communities. Faster and cheaper isn't a substitute for choosing the right strategy for the development of therapies to treat aging, however.
After all, every therapy for aging prior to the development of senolytics to clear senescent cells was fairly marginal in its benefits. Many of the therapies presently under development in the longevity industry alongside the targeted destruction of senescent cells will also be marginal, because, unlike senolytics, they fail to meaningfully reverse a cause of aging. Producing marginal therapies at an accelerated rate is not a success story. Better infrastructure will only efficiently help the end goal of greatly extended healthy life span when coupled to a program of research and development that aims to repair the molecular damage that lies at the root of aging.
I was wondering if you could briefly explain how BioAge verifies whether the aging targets they identify are valid?
BioAge begins with human data. We find human cohorts that have banked blood samples from decades ago, coupled with electronic health records that have followed those people ever since, in some cases, until their deaths. We send these blood samples for deep omics profiling: proteomics, metabolomics, transcriptomics, stuff like that. From that, we can find what's in the blood, for instance, the transcription profiles of the blood cells and soluble protein metabolites, which is correlated with age-related diseases and mortality. That's only part of the picture, of course, because that doesn't tell you what's causal, it only tells you what's correlational. So, from there, we adopt a systems biology approach where we connect the results to whatever datasets we can find out in the world or among those we generate ourselves, which gives us a few extra clues. However, ultimately, the only way we can really verify if a target is valid is by testing it experimentally, and so that's what we do. After we pull everything together data-wise, that only gives us so much. We really need to test these targets in animal models as well as cell models, but we prefer to test in vivo.
Since research budgets are limited, what is your view on how these budgets should be allocated across these differing priorities, i.e. should the identification of biomarkers for more diseases or the development of interventions take precedence if we need to choose, and why?
That's how BioAge started: the whole point was to generate biomarkers. At the same time, these so-called biomarkers are often themselves druggable targets. Part of the evolution of BioAge as a company is that first we find these biomarkers, and then we turn them into drugs. In terms of how society should allocate resources (biomarker research versus the development of interventions), we personally don't have to consider that as much, seeing as we're doing both in-house. I can't really say what anyone else should do, but I think we found something that works for us.
Aubrey de Grey's idea is that if we develop a therapy for one subtype of causative damage of aging, it will be much easier to extend that therapy to similar types of damage. Since BioAge is working on using computational approaches to find the molecular pathways that drive aging, I was wondering if you are using a similar type of clustering approach to facilitate faster intervention development?
To some extent, because I love Aubrey's integrated approach. For me, personally, his 'seven deadly things' talk was kind of my introduction into the field back in 2004/2005. But BioAge, at least initially, takes an opposite approach in the sense that we don't cluster things. We look for mortality as our first differentiating factor. Any target that we look at as something that we might want to pursue must be associated with mortality, and mortality is really as broad as it gets. That being said, once we've screened for mortality, we then examine what specific disease indications would make the most sense. I can't really get into detail about what those are. But I like the way we look broadly at the data first, in a kind of "hypothesis-free" sense, with an open mind, letting the data speak for itself.