In this week’s newsletter

✅ What we know about Open AI’s new model. ✅ A new longevity PAC. ✅ DeepMind is building a virtual cell - and Hassabis is very interested in longevity. ✅ ”Little people trying to grow inside your liver.” ✅ My ambulance to the future. 

🤙🏼 Want to connect? Add me on LinkedIn. 🙏🏼 Not subscribed to the LEVITY podcast on Youtube yet? Do it here. 🎧 More of a listener? The LEVITY podcast is also available on Spotify, Apple Podcasts and other places.

Open AI gets the headlines but I want to Shift your attention to another startup using AI to tackle aging

I've often said that the biggest breakthrough in longevity science this decade isn't a specific drug or genetic discovery - it's artificial intelligence. Aging is biology, and AI is rapidly becoming the ultimate decoder of life's language.

We've already seen AI-driven biology lead to a Nobel Prize, thanks to Alphafold's protein-structure predictions, and countless other initiatives are underway across research labs and biotech companies, each contributing a piece to the puzzle.

Need AI-developed drugs or predictive modeling for clinical trial success? Insilico has that covered. Want to simulate 500 million years of evolution? That’s the mission of Evolutionary Scale. Looking for a large language model designed to create better gene-editing tools? Enter Profluent.

Meanwhile, Nvidia is making a major push into biology with its molecular design framework, and CEO Jensen Huang has declared: "For the very first time in human history, biology has the opportunity to be engineering, not science."

It was only a matter of time before someone applied this to longevity. But when that someone is OpenAI - the most influential AI company in the world - it carries a weight that few others can match. Their involvement means mainstream attention, and longevity science could certainly use more of that.

That said, with the limited information available so far, it’s hard to gauge just how groundbreaking OpenAI’s efforts in this space truly are. It very well might be, but based on what’s been shared, I can’t say I’m particularly blown away - yet.

In fact, I’m (so far) more impressed by what Daniel Ives and his company, Shift Biosciences, are working on - something Daniel shared with me last fall, which I covered here.

And I mention Shift not just because they’re also at the intersection of AI and longevity, but because there’s a deeper connection between their work and the OpenAI news.

Some context might be in order.

Let’s start with Sam Altman, OpenAI’s CEO, who in 2022 invested $180 million* in the San Francisco-based startup Retro Bio. According to an MIT Technology Review article, Altman’s interest in longevity sparked after reading a paper on the rejuvenating effects of plasma dilution in mice.

* Just as I was about to send this newsletter the Financial Times reported that Retro is set ”to raise $1bn for project to extend human life”.

He reached out to entrepreneur Joe Betts-LaCroix, now Retro Bio’s founder and CEO, who pitched an even better idea: cellular reprogramming.*

Retro Bio’s strategy centers on three pillars: cell replacement (via full reprogramming ex vivo), partial reprogramming**, and enhancing autophagy (cellular cleanup). Their goal? “Adding 10 years to healthy human lifespan.” A worthy intermediate step on the path to solving aging.

* According to this article, Retro is pursuing plasma dilution inspired by pioneers Irina and Michael Conboy. However, as of now, there’s no mention of this approach on Retro’s website.

** Alright, we’re diving into the science now. If you’re curious about partial and full reprogramming - or want to learn more about the Yamanaka factors and pluripotency mentioned below - I’ve got a handy guide for you here. That’s a Premium article though. But you can also get very well educated on the topic by checking out our LEVITY episode with Yuri Deigin, here.

My first thought when I heard of the OpenAI and Retro Bio collaboration was: ”Oh, they’re working on partial reprogramming”. But that’s not the case.

Instead, they’re using a custom-designed large language model - dubbed GPT-4b - to improve the efficiency of reverting cells to pluripotency (full reprogramming).

While the Yamanaka factors - a groundbreaking and Noble prize winning discovery allowing specialized cells to revert to stem cells - are remarkable, they’re notoriously inefficient.

This is where GPT-4b comes in. Trained on protein sequences and interaction data, the model can propose modifications to proteins that would be nearly impossible to discover through traditional methods.

The results have been encouraging. The AI-enhanced versions of the Yamanaka factors (dubbed ”RetroSOX” and ”RetroKLF”: two of the four Yamanaka factors are called Sox2 and Klf4) are showing 50 times greater efficiency in cell reprogramming experiments. This improvement is visible under the microscope, with significantly more cells successfully transforming when treated with the optimized factors.

Take a look:

Impressive indeed, but as Daniel Ives at Shift Bioscience pointed out, calling this a ”longevity model” may be premature. What OpenAI and Retro have developed is more accurately a narrower “pluripotency model” - one that has proven useful in enhancing the efficiency of stem cell creation outside the body for use in cell replacement therapies.

It’s also unclear how much Retro’s approach will actually impact longevity. Their current focus appears to be on harvesting a patient's cells, fully reprogramming them ex vivo with these optimized factors, and then reinfusing them into the patient. Their initial targets? Blood stem cells (HSCs) and brain immune cells (microglia).

However, Yuri Deigin, CEO of partial reprogramming startup YouthBio Therapeutics, suspects there might be more at play. He tells me that demonstrating improved Yamanaka factors could just be a proof of concept for Retro’s AI model. The real goal, he suggests, might be to use this platform to discover novel rejuvenation factors that are less potent than Yamanaka factors at inducing full reprogramming - a development Retro may prefer to keep under wraps for now. I need to underline that this is purely speculation.

Why would a weaker factor be preferable? Because if you aim to rejuvenate cells and tissues inside the body (in vivo), you absolutely do not want those cells reverting all the way back to pluripotency. You don’t want specialized cells losing their identity and turning into confused stem cells - that’s how you get organ failure and, ultimately, death. What you do want is for your cells to function as they did when you were 25: healthy, resilient, and fully operational.

And this is precisely why I find Shift Bioscience’s approach more exciting than what Retro is aiming for. Instead of refining Yamanaka factors, Shift is leveraging AI to move beyond them entirely. Their work has already led to new interventions - one of which involves just a single gene - capable of rejuvenating cells without triggering pluripotency.

And in a new development Daniel shared with me, their latest data shows that the human fibroblast cells they’ve been studying also maintains their identity ”whilst decoupling tumorgenicity”* even when continuously over-expressed. This represents a second key safety requirement.
* Using Yamanaka factors to rejuvenate cells and tissues may lead to the formation of an especially nasty type of tumor called teratoma - one potential consequence of having too many of the “confused cells” I mentioned above. This is, in fact, one of the reasons why Retro Bio is cautious in their pursuit of partial reprogramming therapies. Scroll down to find Joe Betts-LaCroix’s less technical - but more morbidly entertaining - description of teratomas.

So, what should we take away from OpenAI’s collaboration with Retro? One thing is certain: this is just the beginning. There’s no doubt in my mind that aging will eventually be solved, and I’m just as confident that artificial intelligence will accelerate that timeline - dramatically.

News from around the longevity and health space.

DeepMind is building a virtual cell

”It's an area I follow quite closely...I would be surprised if [120 years] is the limit”. These are the words from Nobel laureate Demis Hassabis who in a new podcast interview also says ”I think there’s nothing good about aging”.

Furthermore, Hassabis mentioned that Google DeepMind has initiated the Virtual Cell Project. While he has talked about this effort before, this is the first time I’ve heard him refer to it by this name.

To the best of my knowledge, neither DeepMind nor its spinoff, Isomorphic Labs, have detailed how they will build this virtual cell.

But researchers at the Chan Zuckerberg Initiative, Stanford, and KTH have outlined a plan for building a such a thing in a recent Cell article. Their approach relies on large-scale omics data (e.g., proteomics) as the foundational input. They then propose creating a comprehensive AI framework made up of interconnected ”foundation models” to capture dynamic biological processes at multiple scales, from molecules to cells and tissues.

Because living cells constantly adapt and face changing environments, these models must reflect that continual flux. One final step involves linking ”virtual instruments” -neural networks that can manipulate or measure the virtual cell - making it ready for in silico experiments. According to KTH researchers, accomplishing this is an enormous challenge, given the complexity and ever-shifting nature of real cells.

But if there’s one organization with the resources and talent to tackle this challenge, it’s Google DeepMind.

Antibodies for anti-aging

Last July I covered a Nature paper that had caused quite a stir. The study demonstrated that blocking IL-11 - an inflammation-promoting protein - using an antibody could extend mouse lifespan by up to 25% while simultaneously improving health in old age.

More recently, Ageless author Andrew Steele explored this concept in Wired, highlighting IL-11 inhibition alongside other examples to illustrate how antibodies could become powerful tools in the fight against aging.

Using quantum computers for drug design

Amidst all the news about AI it’s sometimes easy to forget that there’s also steady progress being made in building useful quantum computers. IBM is one of the leaders in this field. Here they teamed up with Insilico to generate a KRAS inhibitor*

* KRAS is a gene that regulates cell growth, but when mutated, it can cause uncontrolled cell division, leading to many forms of cancer.

A new longevity PAC takes action

Beyond scientific and technological breakthroughs, the longevity movement must also change public perception - convincing more people that aging is the leading cause of suffering and death, and therefore the most urgent challenge we face. Shifting mindsets, particularly among policymakers, is extremely important. That’s precisely the mission of LongeviPAC, a newly launched political action committee.

One of the driving forces behind LongeviPAC is Dylan V. Livingston, who also founded the Alliance for Longevity Initiatives (A4LI). In 2023, A4LI played a key role in launching the Congressional Longevity Science Caucus, a bipartisan effort to establish champions for longevity research in the U.S. House of Representatives. That same year, A4LI successfully proposed and helped pass SB422* in Montana, significantly expanding right-to-try access for patients in the state.

* SB422 expands the right-to-try law to enable any patient to access medications and treatments not yet given final approval for use by the FDA. You could view it as a ”longevity act” because it removes the previous terminal illness requirement, allowing individuals with chronic and age-related conditions to access experimental treatments.

Crypto, DeSci and longevity

Coinbase has made an investment in Próspera. This is my cue to tell you that you should take a look at our interview with Niklas Anzinger who, by the way, had this to say about the news:

Big promises

During the launch of Project Stargate - one of the biggest infrastructure projects of all time - Oracle CEO Larry Ellison said AI will be able to detect early stages of cancer and in no time develop personalized mRNA vaccine therapies.

Worth your time.

”These little people trying to grow inside your liver”

”So then you have these little people trying to grow inside your liver, which is awkward”. Now, that’s a novel way to describe teratomas. Yes, here’s a new interview with Joe Betts-LaCroix, whom you may recall from my text at the start of this newsletter. The interview is part of the first wave of content from Core Memory, a new ”sci-tech media company” started by journalist Ashlee Vance.

What I’ve been up to lately.

My ambulance to the future

Finally got my welcome box from my cryonics provider, Tomorrow Bio. Inside were a few bracelets and an armband (with instructions that amounts to, “If I die, contact Tomorrow right fucking now”), a membership card, emergency cards, and last will instructions.

Honestly, I can't think of a better use for my body once I'm dead. Compared to all the other ways of handling a dead person there’s a non-zero chance I take a long nap and wake up in the future.

Unexpected consequence of getting my Tomorrow membership: a strange feeling of comfort when realizing that there will be people doing all they can to revive me.

Hey, you’ve made it all the way here! Thank you so much for reading! 🫶🏼