I’m excited (and definitely a little nervous!) to share that I have started as Principal Investigator and faculty member at the Marcus Institute for Aging Research, a @harvardmed affliate! For the past few years, I’ve been obsessed with a pretty simple question: how do we understand and measure human aging well enough to actually do something about it? That question will be at the heart of what we build next. Our lab will focus on computational geroscience, bringing together AI/ML, multi-omics, and large-scale human data to understand the biology of aging and, ultimately, find ways to intervene on it. A big part of that will be building better biomarkers that can reliably measure aging and tell us whether an intervention is actually working. But I’m equally excited about what comes next: using computational tools to identify causal drivers of aging, discover new therapeutic targets, and help find and prioritize new therapies that could extend healthy lifespan. The bigger goal is to close the loop between measurement, discovery, and intervention, connecting computational work with experiments, human studies, and clinical trials. There is a lot to build, and an even longer list of things I still need to learn about running a lab. But that is probably what I’m most excited about. Building a team, learning from people across geroscience, medicine, computation, and industry, and tackling some interesting questions together. I’m incredibly grateful to everyone at Yale who helped me get to this point, and very excited for what comes next at Harvard ! PS I am definitely looking to hire some smart computational biologists and ML/AI folks so definitely reach out!

I think @MartinBJensen raises a number of really thoughtful points, and I actually agree with much of the underlying concern. In particular, I agree that responsiveness should not be interpreted as proof that an intervention slowed aging, and that ultimately we need much stronger physiological and clinical ground truths to establish whether changes in these biomarkers represent meaningful changes in human aging. I would just separate that question from what we were trying to establish in the paper. Our goal was not to demonstrate surrogate validity, but to systematically ask which biomarkers are responsive to human interventions in the first place. Until now, clocks have largely been evaluated on their prognostic validity. We see responsiveness as another necessary step on the path toward eventual surrogate validation. I would also be careful about interpreting responsiveness as evidence that the clocks are “unreliable.” Reliability, responsiveness, prognostic validity, and surrogate validity are distinct properties. In fact, our separate work specifically evaluates technical and biological reliability for this reason. I also agree that the x-axis in the commentary figure should not be interpreted as measured longevity. We did not make that figure, and the interventions clearly involve heterogeneous outcomes and populations. Where I think the discussion gets especially interesting is what constitutes the appropriate “ground truth” for human aging. Unlike worms or mice, we do not have a clean experimental measure of lifespan response available on practical timescales. Mortality, disease incidence, functional decline, physiology, and other outcomes each capture different pieces of the biology. That is exactly why I think we need to build the evidence incrementally: prognostic validity → responsiveness → reliability → ultimately surrogate validation against meaningful clinical outcomes. So I think Martin and we are ultimately arguing toward the same destination. The main difference is that our paper was designed to systematically establish one of the intermediate pieces, rather than claim that we have already reached the endpoint.