Action Potentials for February
1: A connectome census of the mouse visual cortex:
This data set allows the identification of several structural features of neurons, like synapse size or dendrite radius:
One reason this is important: the more that intrinsic neuronal mechanisms are important in memory formation or storage, the more they will need to be modeled to reproduce memory behavior in artificial neural networks. But this is not what was found in this study.
Still, while it does not seem that neurons become more excitable when forming place cell properties, other electrophysiologic changes such as alterations in homeostatic plasticity could still be occurring.
4: Study showcasing problems with the words "dominant" and "recessive" in clinical genetics. Even in diseases traditionally thought of as "dominant", heterozygous states often have biological effects too. Thinking of "semi-dominant" inheritance is usually more precise:
5: Study finds that in the short term, a large amount of midbrain dopamine increases the variability of responses, while in the longer term, it increases the likelihood of actions being performed again.
6: Nice review of engineering T cells for cancer. Making the T cells "off the shelf" seems like the most valuable thing to me, because it would allow more rapid testing of different approaches. Not sure how feasible it would be, though.
7: Study finds that local delivery of T cells was effective in clearing residual cancer cells following incomplete surgery.
8: News to me: the black plague of 1347-48, followed by subsequent waves, caused severe inflation.
9: Reinforcement learning with legal standards: using legally precise language to communicate with LLMs. Seems like a great idea to me as the kind of thing that would be valuable in slow take-off worlds.
10: Forum with people discussing how they are using rapamycin (generic sirolimus) to try to slow their aging. Sirolimus is FDA-approved for a few conditions including preventing organ transplant rejection. I recently learned that a good number of people are taking this off-label for aging. There’s good evidence from animal studies that this will work. Another reason that the future is already here, it's just not very evenly distributed.
11: What are the useful parts of Jungian psychology for a strict materialist?
12: For the first time, partial reprogramming is shown to extend the lifespan of aged (2.3 years) mice. Specifically, they found a 109% extension in median remaining life — control mice had 8.9 weeks of life compared to 18.5 weeks for the mice reprogrammed using OSK Yamanaka factors. The treated mice also had improved health during this time — it actually seems pretty hard to dissociate healthspan extension from lifespan extension in studies like this. The team is Rejuvenate Bio, in San Diego.
13: Study finds that parents play more with children that have higher polygenic scores for educational attainment. This is within-family, so genotype is considered randomly assigned. If true, I’m still not sure how the causality would work. Do such children get their parents to play more with them, or do parents want to play with them more, or is it something else?
14: Infusion of the protein GDF11 in aged mice improves memory and alleviates depression-like symptoms.
In a relatively young cohort (average age of 26 years), serum GDF11 levels also seem to correlate with whether someone is currently in a depressive episode. They're suggesting it could be used as a biomarker of depression. My guess is that this is confounded and/or won't replicate, but big if true. 🤔
16: Narrow healthcare networks (i.e. with few practitioners) save insurance companies money, which is why the companies do it even though it's awful for everyone else. It seems like this mostly saves them money because it helps companies to select for healthier patients, who don't care as much about the quality being worse. In other words, worse is better for health insurance companies.
17: Preservation of a fish brain from 319 million years ago. This seems to have come about due to the breakdown of fatty acids in the brain in a low-oxygen environment, which led to the formation of calcium phosphate. The authors note that the conservation of neural tissue in fossil fishes is likely more frequent than is typically believed, with erroneous assumptions of non-preservation causing people to overlook potentially valuable information about brain structure. 🧐
18: Recipes and chemical ingredients for ancient Egyptian mummification. A wide range of materials was used including resins from multiple types of trees, beeswax, and plant oils.
19: Increasing social isolation is a public health crisis. The authors note that loneliness is a leading factor in determining health, comparable to that of smoking (15 cigarettes/day) and heavy drinking (6 drinks/day).
20: Max More (former CEO of Alcor) is leaving Alcor and joining Biostasis Technologies.
21: Cryonics organization Tomorrow Bio is planning to expand services to several US cities in 2023-2024: New York, Portland, Southern Florida, and San Francisco.
22: Oligodendrocytes prune axons containing α-synuclein aggregates, suggesting that this mechanism might be the cause of glial cytoplasmic inclusions in Multiple System Atrophy.
23: Better than human performance on a science question-answering task using a multimodal chain of thought reasoning (as opposed to language-only reasoning). Cute tagline for the project: "Imagine learning a textbook without figures or tables."
24: A trial of 5-minute daily cyclic sighing finds it is effective in reducing physiologic stress.
25: Miyake events for improved carbon dating — "Miyake events’ allow researchers to put a precise pin in the year in which wooden artifacts were created, by detecting a specific distinctive spike in carbon-14 levels and then counting the rings that formed since then".
26: Recent synthesis of the evidence for antidepressants. There is good evidence of clinically meaningful efficacy, but several outstanding questions remain, including the effects of long-term maintenance.
27: Obviously, this has been a big month for AI. I put together a transcript of one of Sam Altman's interviews touching on AI safety. Here’s Peter McCluskey on AI alignment progress. And using out-of-distribution tokens to detect mesa-optimizers, which seems like it could be generalized in a powerful way.
28: Damage to the epigenome seems sufficient to accelerate aging.
29: A comprehensive review of chronic traumatic encephalopathy (CTE). In my opinion, the evidence for a causal relationship linking repetitive head impacts (RHI) to CTE and subsequent (delayed) clinical symptoms seems extremely strong.
30: Heritability of disease from a cohort of 6.3 million people in Denmark. Mental disorders are among the most heritable, behind endocrine disorders (such as type 2 diabetes), and congenital disorders (which are often defined by a genetic cause):
31: A large exome sequencing study finds that rare coding variants explain much less phenotypic variation (1.3%) than common variants. Most of this is due to ultrarare loss-of-function variants, with allele frequency < 1E−5. The same is true for schizophrenia and bipolar disorder risk, where rare coding variants explain around 2% of the risk. Common genetic variants are much more important for prediction. However, because rare variants are enriched in the same cell types as common variants, rare variants are likely very helpful for elucidating disease mechanisms.
32: Gene therapy cures metachromatic leukodystrophy if the genetic condition is identified soon enough. IMO, just another reason we should be doing WGS on any newborn whose parents consent.
33: It's still unclear if using psychedelics to treat grief is helpful.
34: The adage that “science advances one funeral at a time” seems to both not be true (people, and older people, change their minds all the time!) and be a harmful meme because it is a common argument used to argue against life extension research. Please stop saying this unless you actually have good evidence.
35: A study of 25% calorie restriction (i.e. intaking 25% fewer calories than the person normally would intake) for 2 years seems to slow the rate of human aging as measured by an epigenetic clock, resulting in 2-3% slower aging. In their discussion, the authors suggest that this corresponds to a decreased mortality risk of as much as 10–15%, similar to the effect of a smoking cessation intervention. Participants had a BMI of 22 - 27.9 at baseline and the treatment also led to improvements in cardiometabolic health. However, it did not lead to significant changes in total biological age estimates measured by DNA methylation clocks, which the authors suggest may have been due to insufficient power.
36: Social media seems to be especially harmful for the mental health of teen girls.
37: A downside of inadequate sedation after cardiothoracic surgery.