Longevity biomarkers + 3000 $VITA review article bounty

It would be educational for the community to put together a discussion of the strengths/weaknesses of current biomarkers for longevity.

Furthermore, it would be great for VitaDAO to author a review article and publish using Ants-Review, for example, or via the biorxiv journal pipeline and other alternative publishing methods (including an NFT). This could showcase the power of web3 and DeSci.

Squad: To self-assemble here through discussions
Proposed bounty for the article: 3000 $VITA

H/t @bowtiedshrike


To keep the discussion rolling, here’s a potential outline/list of things that may be considered:

Introduction - need/challenge for longevity biomarkers

Physiologic biomarkers
grip strength
skin elasticity
bone density
VO2 max
organ function

Neurologic biomarkers

Laboratory biomarkers
lipid profiles
immune function

Genetic biomarkers



Hello! It’s a nice initiative! What’s a longevity biomarker, though? Is it a predictor of how long you’ll live that can be measured at any (even early) stage in your life like, say, at 30 years old, – for instance, genetic variants/SNPs that predispose one to certain diseases, telomere length – or rather it’s a proxy to evaluate vitality and monitor healthy aging? It’s challenging to extrapolate a child’s grip strength to the rest of his life and predict lifespan.
We could attract 30-50 people, each writing a section (up to 1000 words) on their domain of expertise. The task is too overwhelming for a few people to accomplish but would require very little work from each participant if broken down into smaller parts.
And what’s going to be the focus? We could survey an array of biomarkers depending on the people’s expertise and willingness to read on the topic, or focus on a smaller set of non-invasive/minimally invasive biomarkers like those that can be measured at your physician’s office – e.g. grip strength, gait speed, bloodwork, skin autofluorescence, body scans, BMI, glucose tolerance, etc.
I would like to join the effort!

1 Like

30,000-50,000 words is book territory, not review article territory. 8000 words is a long review. The longevity textbook comes after this review is published, and whatever other key next review is published.

This can be knocked out by 5 people or so. Fewer if you’re in the longevity field, but I don’t mind more people. If any of the doxxed early career scientists want to participate, it will be one more publication for them if all goes well. I like the idea of a mix of anons and non-anons who are part of a DAO publishing a paper.

There are 2 main needs for the review to address. First is practical educational knowledge: provide a foundation for DAO members to evaluate proposals’ use of longevity biomarkers-- some biomarkers are better than others.

Second main need is to build a foundation for measuring success on longevity projects. How do we know how well a longevity intervention worked? Company will go bankrupt waiting 20-70 years to find out. Once we can put numbers on success/failure, it is easier to convince people to spend money on the winners. It also lets us track how well longevity biomarkers work, so we can update when it turns out some of them are useless.

Main focus should be most accurate biomarkers, followed by most commonly used (good or bad).

Yes, 30,000 for a review paper is too many. For comparison, your post is 223 words. If we’re aiming at 5,000 words (not counting introduction and conclusion), that gives us 20 biomarkers (250 words/marker on average). It depends on how we structure the paper. Agree on everything else.

So I’m just thinking out loud. To move the conversation forward, we need to define what constitutes a good biomarker. Are there any? Surely, there are variables that can prognose if a person suffering from some sort of pathology/is old will or will not perish within the next 10 years. However, how applicable are they to relatively young healthy people? If I started taking an anti-aging drug at 40, some of my markers would improve but there’s no guarantee that I would live longer than if hadn’t taken the drug. So should we think of a biomarker in terms of age (as in aging clocks) – i.e. if my bloodwork or blood vessel compliance are of those of a younger individual, I’m expected to be healthier than my peers and hence live longer? Statistically speaking, yes, there would most likely be a correlation. Or should we think in terms of risk – if my markers deviate significantly from what is considered normal for my age group, I’m more likely to die earlier?

The magic of a table or two is we can fit extra biomarkers people want even if there isn’t space in the main text. Also, where strengths or weaknesses are shared, it will cut down on words.

I’d define a biomarker as a measurable phenotype that correlates positively or negatively with a specific biological function or failure thereof. So both terms of risk factors and health factors. I would define a “good” biomarkers as having the following characteristics: 1) Robust dynamic range of predictive power, 2) Strong predictive power, 3) Strong correlation with the biological function, 4) Easy to measure in humans and 5) Low cost to measure.

A robust dynamic range means there is a big difference with minimal overlap between a positive result and a negative result. Consider the difference between two populations, on a scale of 1-200. In the first scale, the populations have an average of 38 and 40, respectively. With a giant population size, there could be a correlation with longevity there. But it sucks, as there’s likely tons of overlap. Better to have the average be 30 vs 100. A second aspect of dynamic range is the age range over which the biomarker is useful.

A strong predictive power means the biomarker works well. To work well, the biomarker is hard to manipulate without also altering the phenotype (ie longevity). For example, grip strength. If grip strength is a proxy for general sarcopenia, training only grip strength might improve the biomarker without improving lifespan. Also, predictive power includes the odds ratio. For example, if people with the biomarker are 1.2x more likely to die sooner, that’s not very strong. If they’re 10x more likely, way more useful. Instead of odds-ratios, some might use age-adjusted years, which I think Savva calls QALYs.

However, a strong correlation is also needed. If that 10x difference has a p value (likelihood of being false on 0-1 scale) of 0.06 vs the 1.2x having a p value of 0.00001, that means the 10x might be less reliable. Similarly, if the correlation coefficient is lower, it may not be as useful.

Easy to measure means biomarkers should be assays that are quantifiable, have minimal operator-bias, and can be performed with minimal training. Things like blood draws and standard assays, VO2max, and grip strength vs MRIs, cognition tests, spinal taps.

Cost includes dollar amount per test, specialized equipment needed for the test, time to perform, and time to analyze. If a blood test performs equally to an MRI, the blood test is superior.

I think for this review, discussion of interventions should be avoided when possible. Instead, the focus is on ‘how well and under what circumstances do these biomarkers correlate with longevity, and what are the weaknesses of these biomarkers?’

1 Like

I would be happy to join the squad and contribute
I am currently deeply researching digital biomarkers and aging biomarkers in general for my new role.


So we’ve started an airtable to put together a list of biomarkers. I also suggest we create a collaboration group with one of the citation manager services to keep track of our references. It’s going to save us a lot of time down the line. I personally prefer Zotero – it has integration with Google Docs, very convenient.

Mendeley is fine too, but it doesn’t work with Google Docs.

1 Like

I use Endnote, but Zotero should work ok. Thanks for setting up the airtable.

Existing panels should be considered as well for biomarkers.

For stage 1, let’s list biomarkers and panels.
Stage 2, choose the ones we want to talk about
Then stage 3 talk about them.

Let’s aim for Friday 15 July to get Stage 1 done.

1 Like

I propose getting rid of cost here since technology is changing and biomarkers should be evaluated regardless of commercial applicability. If an absurdly rich person wants to change their longevity marker set, we should be as inclusive as possible of anything with high correlation. A good example here would be telomere length which would have been quickly excluded from such a list 15 years ago if cost was a determinate. This can be aspirational, right?

1 Like

Can you link to the airtable or other tool(s) here?

Well, cost is an important parameter. No aging bimorker would ever be used at any reasonable scale with costs of 200-400 USD per test. I would suggest adding an estimated cost which is possible to achieve within reasonable time frame (3-5 years) if tech is scaled, advanced or other factors. Sort of first principle assessment of cost (harder) or trying to find historical cost trend in other biomarkers and project estimate from there (easier).

Last year there was a preprint by Sinclairs lab claiming ultra-cheap DNAm

At the end of 2020 I also asked my friend (CEO of DNA sequencing company) to give his estimate - he said 50$ / 500-600k CpGs is achievable with current tech if there is reasonable demand (200-300 tests a day)

1 Like

Hello! I sent you invites to the airtable and the Zotero collaboration group. We also reached 5 people max on my Airtable free plan. I wonder if we can migrate the table to the VitaDAO account. @longevion

I agree that costs will change over time, and faster for promising technologies/markers. Note that in 15 years, the review will be out of date based on the science, not just the cost.

There are two important reasons to consider cost in the review: 1) if two biomarkers perform similarly, but have a 10x difference in cost, it’s reasonable for funders to ask why the cheaper biomarker is not used. One goal of this review is to give VitaDAO members a reference when considering proposals.

  1. We need to be selective in this review. As @rpill points out, 250 words spent on each of 20 biomarkers fills out a review. There are way more biomarkers than that used, not to mention panels of biomarkers. We can summarize a lot of them with a figure, table or three. But. Seeing cost in the table would help the reader realize why we prioritized certain biomarkers over others for discussion.

I’ve been thinking about a unified statistical framework to evaluate biomarkers. Your feedback is appreciated.

p-value isn’t the best metric to compare for the analysis you want to do. p = 0.01 to p = 0.0001 is a 100 fold difference, but both p values indicate solid biomarkers-- they discriminate two populations. In contrast, going from p = 0.5 to p = 0.05 is a 10 fold difference, yet that the difference between a biomarker not working vs working. The p-value comparison would optimize for statistical abnormalities.

We may need to devise a VitaDAO index, which would account for all the parameters we think are important for a biomarker to make sense of all of them. Or we could use indices others have made. But it may be easier to cover them at a high level first.

Thank you for the feedback. Let me clarify my reasoning.

For instance, the difference between sample biomarker values measured at 30 and 40 years has a p-value (30 vs 40) = 0.01. The same biomarker value difference measured at 40 and 50 has a p-value (40 vs 50) = 0.0001. IPR = 0.01/0.0001 = 100. So, greater than 1 and the biomarker is good.

Let’s say in your second example p-value (30 vs 40) = 0.5, and p-value (40 vs 50) = 0.05. But I wouldn’t calculate IPR, because the former p-value is greater than the cut-off 0.05. This makes the biomarker bad. If it were p-value (30 vs 40) = 0.05, the IPR would be 1. That’s how I see it.
The second biomarker (IPR = 1) is okay, but not as good as the first one (IPR = 100).

Effect size generally correlates with p-value. And the use of non-parametric tests would make what I propose more robust.

If you want effect size, it’s better to use effect size instead. p-value only tells you if two populations are distinct from each other or not. Greater probability does not mean more distinct.

If the p value changes from 0.001 to 0.01 or 0.000001, it doesn’t matter. Populations remain distinct.

It may be best to start with what others use for ranking biomarkers and either say ‘we don’t like this method for X, Y and Z’, or just use their metric.

1 Like

I posted an anemic skeleton of the document in the old longevity thread in general discord chat (don’t have the phone number Google demands to create google account, lol). With 4 people here, I propose we break things down for the initial round of writing as follows:
@rpill – Biological Age clocks and Laboratory biomarkers
@strygah – Physiologic biomarkers
VictorB- Neurologic biomarkers
@Derek – Comorbidities
Shrike-- Genetic/Epigenetic biomarkers

Let’s aim to have these first targets completed by July 17. If you don’t want to contribute, please let me know and we’ll get the section reassigned. If July 17 is too aggressive timeline, let me know. If someone wants to participate and I didn’t tag you here, please let me know.

Second round will be biomarker panel section. Third round will be conclusions, problem statement, abstract, figures, finalize tables, and revisions of all content for coherency. Sort out author order at the end of the third round based on contributions. Fourth round will be revising/polishing/community feedback.

When it comes time for figures, https://bioicons.com/ is the free equivalent of BioRender. www.photopea.com is a free web version of Photoshop.

1 Like