Rhonda: What do you think the major advantages of using these epigenetic aging clocks are in terms of, you know, their use? I mean...

Dr. Levine: Yeah, so, people have been really interested in this idea of quantifying biological aging or estimating biological aging. Because again, chronological age is just an imperfect proxy of this process that we actually care about, that, you know, is why people get diseases when they get older, why people die when they get older. So, if you can actually quantify the process as best as possible, it's better than chronological age.

But there's been some disagreement on the types of data or the types of markers that you could use to do that. Some people think we should look at just, you know, clinical markers, and those are useful. But for me, the exciting thing about epigenetic clocks and measuring DNA methylation for this is that you can use the exact same clock across almost any tissue type and almost any cell type. So, I can compare aging in, again, skin to aging in brain using the exact same measure. And I don't think there's any other type of biomarker that you could do that with.

Rhonda: What about the differences in these clocks with respect to, you know, the original Horvath clock you hear about versus the one that you developed with your mentor Steve Horvath with respect to phenotypic aging, phenotypic age clock, GrimAge? Like what are some of the major differences in those clocks in terms of their predictive power?

Dr. Levine: Yeah. So, the original clocks were used. So, when you develop these clocks or train these clocks, we use machine learning, so, we're usually trying to predict something. So, you take all your methylation data and you say, "How can I predict whatever this is?" And the original clocks use chronological age. So, the idea is, "Can I look at methylation and predict someone's chronological age?" And so, again, that's kind of this idea of the pattern of someone who's this age.

But we know, again, chronological age is an imperfect proxy of this process we're actually trying to quantify. So, what the second-generation clocks did, the one that we published in 2018 was the first example, is we said, "Oh, can we come up with a better thing to try and kind of tune these measures to?" So, in that case, we used kind of normal lab tests that we combined into a measure that was predictive of mortality and then we trained a predictor of those lab tests.

And a similar thing was done with the GrimAge clock where they took these different proteins and they trained predictors of that and then trained the predictor of mortality. So, we think that something that captures mortality or health span or kind of physiological decline is going to be a better thing to tune these clocks to than just chronological age.

Rhonda: How do these epigenetic aging clocks, I think more specifically the second-generation ones, the PhenoAge one that you mentioned, or the GrimAge, compared to measuring biological age to like some of these clinical biomarkers, so, HbA1c, your cholesterol, you know, lung function, like the classical things that people are measuring to measure biological age? And, of course, I think a similar question would be also mortality risk as well, how do they compare in terms of their predictive power, you know, as a biomarker?

Dr. Levine: So, if you're measuring the epigenetic clock in blood, I would actually say that they're on par with not the individual clinical lab tests, so, you know, they're going to be more predictive than if I just look at HbA1c or just look at cholesterol, but we can also combine these clinical tests into a single kind of risk measure, which is what we do with PhenoAge. And then I would say they're actually on par with that. The advantage of epigenetic clocks is, again, you can get a different measure for different tissues or cells in your body. So, the blood tests, these clinical tests, you'll get one measure, one biological-age measure out of them, but for the epigenetic clocks we can measure your skin's age or, you know, if you had a biopsy, you can measure different organs' age. So, even though people usually just use them in blood, they have a lot more potential just to compare kind of how different organ systems are aging.

Rhonda: Kind of that was kind of my next question a bit is that, you know, as you know, people age at different rates but, even within an individual, their organs can be aging at different rates as well. I know there's been work from Dr. Mike Snyder, who we had on the podcast not too long ago, who I know you're familiar with his work showing that some people are metabolic agers. So, their liver and, you know, their kidneys, they may age quicker. Some people are cardiac agers where they seem to be more at risk for having heart problems. Or they could be immune agers, so, their immune system, they're more susceptible to pathogens and stuff with age. So, if you were to measure like a blood sample from a person using...you know, would the epigenetic clocks pick up that or is it measuring more of the system's level type of aging?

Dr. Levine: So, the current ones are not going to capture more of this multi-dimensionality in aging. And I'm a huge fan of Mike Snyder's work and actually it's kind of influencing some of the work that we're doing now. So, what we're actually trying to do now is to build clocks that are proxying aging in different organ systems. So, these aren't out yet but the idea is that, if you can build a clock that's going to, essentially, try and proxy what your brain aging is or your liver aging or your kidney aging, then you can have multiple kind of epigenetic-age estimates and really understand more of the profile of the person.

And so, yeah, it's kind of getting back to this idea of ageotypes. We both age at a different rate from each other but also in a different way. Right? So, I might diverge more down kind of this way and be more of a metabolic ager, or whatever it is, and someone else might be more of an immune ager. And understanding that is going to give people both, potentially, insight into what interventions might be the most helpful for them and also what they might be most at risk for.

Rhonda: Right. Mortality risk is a big one that you see with like GrimAge and even, I think, PhenoAge as well, like, how well do they predict mortality risk and even disease, you know, specific mortality, like your cancer mortality risk or your...and how do they compare to like a frailty risk or a frailty index measurement or something like that where, you know, you can also measure mortality risk?

Dr. Levine: Yeah. So, they're actually pretty powerful when it comes to mortality risk. And I would say right now GrimAge is the best in terms of predicting, what we would consider, all-cause mortality, so, basically any mortality risk all combined together. GrimAge is particularly good at cardiovascular risk mortality, which is why it does well at all-cause mortality because that's the biggest killer of people, at least in the United States. But I think, in terms of predicting more specific types of mortality that someone might be more or less at risk for, I think this is where you need more of these systems' measures. But they are pretty powerful at predicting kind of remaining life expectancy or all-cause mortality. Of course, they can't predict who's going to get hit by a bus, or whatever, but in terms of kind of population averages, are you more or less likely than someone else with the same chronological age to have early mortality, they're actually pretty good at that.