Links (48)
2 June 2021 | 12:00 am

Policy entrepreneurship at the White House

Mach 12 airplanes

Mach 17 airplanes

A... unique interview? with Marc Andreessen?

Recent interview with Peter Thiel

Gene therapy for elevated LDL cholesterol

I wrote a while back that "everything sleeps". Scientists looked at a new organism that has no brain (but has neurons), the hydra. It also sleeps.

Can you tell fake from true paper abstracts? (Via Scott Alexander)

Why is the Moscow metro so nice given that Soviet buildings are more on the brutalist end (also via Scott)

Curtis Yarvin on AI risk skepticism (Coincidentally, also my argument: diminishing returns to intelligence + inherent unpredictability of the world; though afaik I've never written about this.)

Reservoir computing

Ada Nguyen on not overthinking

UFO skepticism

From a social scientist that quit Academia: "most social psych research does not matter"

Looking like Covid was a manmade pandemic

Why do we produce LDL, I ask Twitter. Should we aim to have 0 LDL? I get some answers.

An article I wrote for Works in Progress: We don't know how to fix science (but we should try anyway). Read the whole issue here.

One more case of highly credentialed academic shitposting

Going up mountains might be bad for your brain

A bunch of reprogramming papers coming up this week: reprogramming muscles, better understanding reprogramming (including various OSKM combinations as well as non-Yamanaka reprogramming), and reprogramming for lifespan

Partial Vision restoration in a blind patient (retinitis pigmentosa) using an ontogenetic therapy. Check out the videos in the supplementary information as well.

Researchers train a model on citation metrics to predict top papers (and they infer, biotechnologies), and predict what will be the technologies of the future in bio. Some criticism.

Persuading the body to regenerate its limbs, one more Michael Levin piece

Sugar as the dark matter of biology, feat. glycoRNA

Wokeness is just applying the law in the US

Against against SATs

Critique of the Hallmarks of Aging

Martin Borch Jensen's longevity apprenticeship

Calling bullshit, a course


How to solve aging
8 May 2021 | 12:00 am

Answers to this question can be found online in the forms of shorter and longer term roadmaps. One could infer an answer from SENS or get a list of ideas from Celine Halioua's recent take on the matter (On what the field needs rather than how to deal with the problem itself). This is just my view.

An important question for anyone wanting to answer "How to solve aging" is "If aging is solved, have we then made a given individual, in normal conditions, immortal?" Or "Have we thus made it not immortal but negligibly senescent instead?". Or in other words, once aging is dealt with, how much work is left to do for medicine? Ultimately that's the goal, lest we want the aging field to end up like nanotechnology, that once aspired to the great lofty goal of atomically precise manufacturing but ended up small-minded due to PR concerns with the more visionary and ambitious goals of the early days.

An additional point to have in mind when thinking about the question is the the geroscience hypothesis, this is the idea that a small number of targets or pathways can have an impact on multiple diseases at once. This is not the claim that all disease can be treated in this way, just the more modest claim that instead of the usual Pharma approach of one disease : one target, we can find one target for multiple diseases.

My own answer to "what is aging", and the broader reading I've done leads me to believe that under a plausible (but not the only possible) definition the answer is no, once aging (in a narrow sense!) is solved we have not achieved either immortality or negligible senescence.

In one plausible definition of aging, and coherent with the geroscience hypothesis, once we identify those targets that can be hit to affect multiple diseases, and modulate them in the right way, then we have cured aging; we would have taken this highly connected set of processes that leads to multiple diseases and ameliorated them at once, lengthening lifespan and healthspan.

But we would be leaving out those nodes of the network that are not as tightly connected to the rest and thus require particular attention. My go-to example is the gradual narrowing and stiffening of the arteries leading to cardiovascular disease. This may have multiple causes (Calcification of the arteries, systemic or local senescence, inflammation, and just circulating LDL getting stuck in the arteries), but the phenomenon itself is not a consequence of some process deteriorating as we age; aging may accelerate it but does not cause it in the first place. Even if we get rid of inflammation and senescence, by the nature of what cholesterol is and how arteries work, it will get stuck in there and lead to problems (if children lived forever, this process alone, holding everything else constant would lead to mortality). It is not a matter of increased LDL production or decreased uptake, although of course everything else will modulate this accumulation process. It could be that holding everything constant still leads to cardiovascular disease at age 200 rather than age 60; not saying here that there aren't huge gains to be exploited by targeting aging in the narrow sense.

On another, broader, view of aging as damage accumulation, we wouldn't care that we need a myriad of specific interventions to complement the main geroscience path to health. If arteries get stiffer and LDL builds up, we could theoretically fix that directly. This is hard to do and historically Pharma has focused not on this (damage removal; here cyclodextrins) but on slowing down the damage (with statins, or various forms of PCSK9 inhibition). But it's a necessity if one wants to comprehensively reverse all functional decline, or so it seems to me.

Could it be that aging broadly (all damage) and aging in the narrow sense (the tightly connected core processes that degrade but not all) are the same? That if we fix the latter we get a fix for the former for free? My intuition is that no, that we would still need targeted interventions, but it would be great to be wrong! The argument for this is the existence of classes of molecules (long lived proteins, but also oxidized cholesterol and others) that we may never have had the capacity to naturally remove (or what'e the same, that we can't remove in a sufficient amount), if so no amount of up/downregulation of anything will fix that (In one proposal to address this, we'd need to replace entire tissues or organs).

Then, if that process has some way to lead to mortality without a way to compensate for it, that alone will lead to mortality [what we ultimately want to push back against, along with functional decline in general]. This is not to say of course that treating aging in the narrow sense will do nothing: it can still push back many diseases at once, which is a huge win. But going beyond that will require a comprehensive approach that goes beyond this.

That comprehensive approach will have treating aging at its core, as well as scoping out what is left to do. It will involve the regular Pharma approach of listing all diseases and looking at why they happen. The geroscience optic is key here, usually if you're working on pancreatic cancer you look at the pancreas, if you are working on fixing dementia you look at the brain. But it could be that treating Alzheimer's requires fixing the immune system, rather than doing anything to the neurons. Importantly, and unlike most of regular drug development that targets a single disease cause, a cure for a condition will very likely involve combined interventions, especially interventions that do not slow down or speed up the rate of accumulation of some molecule; rather interventions that are directly removal or addition. For Alzheimer's we may want to target various amyloids to clear then, get the immune system back in shape, reduce the various sources of inflammation, and perhaps the cribriform plate as well. This is rarely ever tried! Let me repeat that: We do have on the table multiple ways to address a disease; we suspect or know that diseases are multi-factorial and yet we rarely test interventions together (except in oncology).

It is very plausible that if a combination cures a disease each part of the combination should ameliorate it by some amount. But if we never try the combinations we'll never know if the effects are synergistic. Four drugs to treat Alzheimer's may yield a boring 10% reduction in cognitive decline, but what if together they effectively stop or reverse the disease. We don't know. And there's no good scientific argument for why this is doomed to fail. There are, to be sure, incentive-based arguments for why doesn't happen more often:

  • Scientists want to understand how a disease works. Poking the system times four makes it difficult to see exactly what is causing what
  • Entrepreneurs are not regulatory masochists: You can help a lot of people and make a pile of money by dealing with the low hanging fruit first. What would a Phase I or II trial look like if you have 4 therapies going on? How many combinations of dosage would you have to test? Not that this is impossible: we see this sort of setup in oncology, but it may be because it's a very visible life-threatening condition and even if the treatment harms the patient they were not going to live that long anyway so maybe the FDA is more chill with oncologists than with everyone else, who knows.

Summing up, there are two things we should be doing more of to treat aging in the broad sense

  • 1 to N treatments (the geroscience approach) where one (1) drug can ameliorate multiple (N) diseases
  • N to 1 treatments (the combinatorial approach) where a number of interventions (N) is concurrently tested to treat a single (1) condition

Wishy-washily I like to think of gerontology as divided in various stages:

  • Aging 1.0 is the classic geroscience (1toN) approach via a set of usual suspects (mTOR, calorie restriction, IGF-1). This has been at the core of the field since its inception, they are the most robust interventions we know, and I have nothing against them, but it's important to note that there is geroscience beyond this. These will probably be the first interventions to hit the clinic in some form. (Off-label, some already have)
  • Aging 2.0 is a set of novel interventions that seem to have systemic effects but that are not that well characterized yet (As are reprogramming or the family of interventions adding/removing factors directly from blood). This is where we are now.
  • Aging 3.0 is the concurrent deployment of 1toN and Nto1 interventions together. This will benefit from a deep understanding of the individual interventions and diseases.

In practice what does the road to get there look like? My bias is in the science (You could make the FDA go away and be able to yolo trials through but that wouldn't solve the core of the problem). I believe with Celine that the first aging drug could be approved this decade. But what happens next? The science is just not there yet. Here are some things (not necessarily all the things!) that I think the field needs:

  1. Funding for combined interventions. Take a single disease and roadmap an ideal way to target it in a combinatorial way. Multi-omic approaches may be helpful here to see exactly what one intervention does or does not then use that to see what else it should be combined with.
  2. Aging biomarkers. To treat aging in the narrow sense we could go to the FDA and say that drug D will decrease multimorbidity (a composite of various disease events). Great! But this trial will take a long time. Imagine having to wait 30 years to approve a statin. Instead, we need the LDL of aging, some biomarker that is related to multiple conditions that can be shown to causally influence them. Then, an intervention can use this as a surrogate endpoint and quickly validate the intervention. Clocks may be used to this end if properly validated.
  3. Better animal models. From younger B6 mice to diverse, aged mice. These are more expensive so would require partnerships and investments in upstream suppliers (CROs like JAX or Charles Rivers)
  4. Moving past the hallmarks of aging to a quantitative, causal, understanding of the drivers of aging. To a first order approximation, everything is affecting everything else in nasty loopy ways. But how much? We need to poke at the network of cause and effect that is aging to see how things are connected.
  5. Information collection. The aging field is complex! Biology is complex! I wrote the Longevity FAQ to give a primer to the field, but there are many, many, many more things out there. We would benefit from more synthesis work done by interdisciplinary experts. Imagine having a resource such that if you read it you are confident you don't need go chase information anywhere else because what's relevant has been already added there (This of course would need constant updates and the ability to express dissenting views as well as being able to choose the level of depth you want to go down to). But solving this problem would help point researchers, entrepreneurs, and policymakers in better directions.
  6. More people! Especially more researchers. But also people playing various forms of coordination roles across labs.
  7. More human tests. We need the aging biomarkers for this for robustness, but ultimately we need to take these interventions to humans. This can be done academically especially if the interventions have already been FDA approved, as with the Fahy work on thymic regeneration, the AMBAR study for albumin replacement or PEARL for rapamycin.
  8. Tooling. This is not just for aging but in general the history of science is the history of the struggle against the lack of good quality data. In some areas (as in some branches of physics), principled thinking may yield insight without having to measure anything (We've been confirming Einstein's hot takes since he proposed them, and he did not need fancy telescopes or Big Data to come up with General Relativity), but in biology it seems that the acquisition of better tools to perturb cells and organisms and measure their changes in detail is key. There is a growing list of -seqs out there as are delivery mechanisms for novel therapeutical modalities. But... wouldn't it be nice to have more?

Links (47)
4 May 2021 | 12:00 am

Want to own the Longevity FAQ as an NFT? You can bid for it here.

Why has nuclear power been a flop?

Construction costs around the world: How does the US compare?

Coming full circle, from endless complexity to simplicity and back again (On the history of cancer research)

What evidence do we need for biomarker qualification (at the FDA)?

How was it found that lithium works as a psychiatric drug?

Roundtable on Richard Hamming

Review of "The Kill Chain"

Why is everything liberal (as in leftist)? Because they care more than everyone else.

The slowdown in agricultural productivity growth and its causes

Is Venture Capital broken, producing unicorns that consistently lose money?

2021 batteries roadmap

ACT (SSC) invited book review on "Why Buddhism is True"

Prospectus on Prospera

Against alcohol

Once more the Emdrive is tested, once more it fails tests

The latest on the mammograms and cancer mortality debate

Michael Levin's xenobots and Michael Levin profiled. Levin is imho doing some of the most interesting work in the life sciences right now.

Crypto skepticism

Inside the NSF and Beyond

The problems of air pollution

Cloudflare vs patent trolls

Funding Risky Research. A good summary of the literature, though they forgot at least to cite some of the most recent work on using disagreement to choose what to fund.

Thread on the Soyuz rocket

Why is nuclear power so expensive? Article and threads

Against stoicism

When did SF started to engage in urbanist self-harm? 1978

SpaceX wins key NASA contract. Is this the end of SLS?

Ben Reinhardt's list of optimistic sci-fi

CRISPRoff, a new technique for epigenetic editing

Red Pen Reviews (a nutrition science book review site) has given "Eat, drink, and be healthy" their highest score ever. Nothing too weird here, grandmothers were right all along.

The case against vegetable oil (I don't have strong views here, other than being an olive oil supremacist)

New deep learning model to extract data out of e.g. papers

Physics is broken! (We'll see, but this doesn't look like the usual fluke)

Pseudoerasmus pushes back against the centrality of manufacturing for economic development

The automotive industry is now able to move faster from start to completion of a given project. This stands in contrast to the US military, which is now taking over two decades to finish a given program.

Drone delivery is already here, at least if you live in Christianburg, Virginia

Mitochondrial uncouplers in cancer therapy

US aggregate R&D is on an upward trend, being at a historic high



More News from this Feed See Full Web Site