Action Potentials for December
Note: I haven't done deep dives on all of these links. Correction, criticism, beratement, &c, is welcome.
1: The size of synapses between two neurons tends to be correlated. A new study finds that in L2/3 pyramidal cells of the mouse primary visual cortex, there is actually a bimodal distribution of sizes, modeled well by the sum of a binary variable and an analog variable drawn from a log-normal distribution. The binary variable of synapse size is correlated among dual connections, which seems to be related to a synaptic plasticity rule like Hebbian learning. On the other hand, the analog variable is not correlated among dual connections, which means its size might just be due to random fluctuations. If true, this suggests that measuring precise synapse size might not be as important to capture functional information, as long as you can classify the synapse into one of the two major “buckets” of sizes.
2: Mapping out climate tribes — not only answers this question but also has some really interesting thoughts on doomerism in general. Highly recommended. The author, Nadia Asparouhova, is the niece of FM-2030, who is famous in the transhumanist community and is cryopreserved at Alcor.
3: A review of the current evidence base for psilocybin, LSD, and ayahuasca for depression. They conclude that given the methodologic limitations of current RCTs, the evidence for the efficacy and safety of serotonergic psychedelics for depression is currently low.
4: In this population-level study, menopausal hormone replacement therapy (HRT) seems to increase the risk of dementia, with approximately 1 additional case of dementia per 250 women receiving HRT, and a dose-dependence trend (p <0.0001). This is consistent with previous RCT evidence. In the discussion, the authors note that this finding may be due to an increased ischemic burden, as HRT also leads to a higher risk of stroke.
5: Scott Aaronson on a reform approach to AI alignment. In comment #62, he describes an approach for AI alignment in slow takeoff worlds.
6: Great article explaining why mental illness is medically real when using the same standards applied to other chronic medical illnesses. The author is right to call out former NIMH Director Thomas Insel, who should read this article.
7: An argument that AI will not take over the world — ?at least in the medium term —because it will be good at short-term tasks but will not have as much of an advantage in long-term planning, where it is harder to optimize and show competence compared to others.
8: Dendrite-like neural network computing, rather than point neuron/synapse-like neural computing, seems to present some advantages.
9: Very nice large connectomics study of the cerebellar cortex in mice.
10: A study finds that APOE4 causes oligodendrocytes to aggregate cholesterol, leading to dysfunctional myelination. Also suggests that treating this with cyclodextrin may help to stabilize cognitive function. A sample size of 32 does not seem very well-powered for their human postmortem brain findings.
11: New genetics study for Alzheimer’s disease, finds that rare variants in ATP8B4 and ABCA1 are associated with Alzheimer's disease risk. Variants in ADAM10 also narrowly missed significance. This corroborates my priors in implicating APP processing in the disorder, as distinct from amyloid plaques per se.
12: Study finds that oligodendrocyte precursor cells ingest axons in the mouse neocortex. Nice supplementary video to prove their point:
13: Higher social status in female macaques is associated with a younger relative transcriptional age of brain cells.
14: A decrease in gene-gene RNA correlation is a robust finding in aging cells, associated with increased DNA mutational load.
15: On choosing good quests for one’s life (HT: my brother):
16: Diplomacy AI win is not as impressive as it seems at first glance, in my opinion. The AI is winning mostly by knowing the correct moves at any given time point and by not playing the metagame that humans do, instead defecting in a one-off version of what is usually an iterated prisoner’s dilemma. It’s not like the AI is actually engaging substantially in strategy, deception, etc.
17: Increases in the CSF levels of microglia markers are associated with slower rates of tau accumulation in the brain and slower cognitive decline. This suggests that activated microglia might strengthen brain resilience to Alzheimer's disease pathology.
18: One of the many reasons we shouldn't expect anti-aging treatments to come about anytime soon — the business of clinical trials is ridiculously inefficient.
19: New imaging method for fixed tissue uses photocrosslinking of DNA barcodes onto complementary DNAs in situ, allowing in situ sequencing of full transcriptomes. Here they used it on mouse retina tissue:
20: A socio-cultural summary of the current status of substance use in elite and recreational sports.
21: Does Parkinson's disease start first in the enteric nervous system and spread to the brain or start first in the brain? This study presents evidence in favor of the latter, specifically finding that α-synuclein pathology is found early in the olfactory bulb.
22: Real median personal income in the US since 1988 has increased by about $10,000, with local peaks in 2008 and 2019.
23: A systematic review of intermittent fasting on mental health outcomes shows that it seems to be helpful for depression (SMD: 0.41; 95% CI: 0.05 to 0.76; I2=45%; n = 4), with no significant effect on anxiety or mood state.
24: Large trans-ancestry study of tobacco and alcohol use. They identified four main ancestral clines: African, American (recently admixed ancestries primarily from the United States), East Asian, and European. The data set was approximately 21% non-European:
For tobacco and alcohol, a large majority of the variants associated with use showed consistent effect sizes across ancestry dimensions. However, polygenic risk scores developed in one ancestry did not perform well in people of other ancestries. Here, you can see the decreased out-of-sample predictive power of a polygenic score for tobacco use developed in people of European ancestry across other ancestries:
25: More data showing how the combinatorial expression of clustered protocadherins in neurons affects synaptic connectivity patterns and localization.
I have previously suggested that protocadherins, given their diversity and specificity to each cell, could be measured to help infer the original morphology information of cells if there is significant damage in a preserved brain.
26: One of the downsides of medical device implantation is that if the company making it goes out of business, the people who had the device implanted can be left with a non-functioning implant, which leads to all sorts of problems. This risk has to be considered when evaluating the cost-benefit ratio of an implanted device.
27: Genetic correlations of REM sleep behavior disorder (RBD). As expected, the strongest correlations are with AD, PD, and LBD. Also an expected association with excessive daytime sleepiness. It's interesting to see the weak but significant association with bipolar disorder, but I would like to see that replicate before reading into it.
28: Ever since they came out against supporting life extension research, I haven’t liked Elon Musk or Sam Bankman-Fried. This is one of my major litmus tests. Hindsight bias is, of course, 20/20.
29: Silent synapses are found to be present at filopodia, which are thin dendritic protrusions. They express NMDA receptors but not AMPA receptors:
As the authors note in the discussion, this is more evidence that the size of dendritic protrusions correlates with their associated synaptic states, including both synaptic strength and plasticity thresholds.
30: The industrial revolution commodified muscle strength. What happens when AI commodifies IQ? Here is a take that one's novelty quotient is going to be more valued.
My view is that while intelligence is commodified, high-quality judgment, and associated reputation, are going to be more valuable than ever. But I tend to think of intelligence, smarts, and judgment as fairly distinct.
31: Cross-disorder genetic correlation studies in psychiatry are difficult because so much of the variance is driven by schizophrenia, the psychiatric disorder with the strongest genetic signal.
32: New proposed FDA regulation for diagnostic tests by Scott Gottlieb and Mark McClellan. Switching to a "firm-based" regulatory approach so that test developers can market new tests and update their existing tests without undergoing the same pre-market review seems like a really good idea.
33: A nice study using agent-based modeling to study crime and corruption dynamics in a simulated society, with authors from the Universidad Nacional Autónoma de México. A key finding is that ongoing crime is dependent on the perception of high police officer corruption among citizens.
34: Nice reporting by Ioan Grillo on the confusing and tragic situation with asylum applicants in the United States. He also has an interesting article explaining why Mexico City has a surprisingly low murder rate relative to the rest of Mexico, lower than Dallas or Portland and approximately on par with Austin.
35: Study of haloperidol for delirium in ICU patients. 2.5 mg IV q8h and 2.5 mg as needed did not affect mortality and did not have significant adverse effects. Also, a chart review suggests trazodone may be effective as an alternative to antipsychotics for delirium in hospitalized patients.
36: In a new study, telehealth was just as good at detecting dementia as in-person assessments. This will allow clinical trials to be performed in a more decentralized manner, hopefully reducing costs and increasing access.
37: Crispr approach for Huntington’s disease with intrastriatal delivery reduces mutant HTT mRNA and protein levels, tested in mice, seems promising.
38: Small sample size GWASes with rich data are helpful for prioritizing genes within genome-wide significant loci, but do not substantively increase novel gene discovery because they are still not well-powered enough. There’s no substitute for sample size yet in genomics.
39: Largest study to date of self-reported sleep duration across the world. Sleep duration is shorter for men than women by 7.5 minutes. It drops in one’s 30s and 40s, which the authors report is due to increased career and childcare responsibilities, before then increasing again at age 53 on average:
Across the world, reported sleep duration is highest in Eastern Europe and lowest in East Asia.
40: Small, phase 1/2 safety trial of a gene therapy for Alzheimer's disease that uses a virus to induce ApoE2 expression in people who are ApoE4 homozygous. ApoE2 was found to be produced. Some CSF biomarkers of Alzheimer's disease improved. Seems like a very promising approach, although potentially difficult to scale since it involves an intrathecal injection. Even a difficult-to-scale but effective approach would be great news in Alzheimer’s, though.
41: Microglia are not required for myelin development, but they regulate the amount of myelin once it is developed, preventing hypermyelination and also demyelination during aging.
42: Covid RNA and protein can be found in the human brain, sometimes even months after the initial infection. However, in this cohort, it doesn't seem to be associated with specific disease-associated histopathologic changes such as inflammation, beyond non-specific vascular changes such as congestion and signs of hypoxia.
