Action Potentials for January
Note: I haven't done deep dives on all of these links. Correction, criticism, beratement, &c, is welcome.
1: A study on how distant memories are recalled. Recall seems to involve a subset of the original engram neurons, along with additional neurons that are strongly connected to those neurons.
2: Where is all that ancient DNA coming from? Mostly from tiny fragments of bone or fecal remains that are scattered throughout the sediment. The quantity detected is dramatically different in parts of the sediment, likely due to differences in biochemical microenvironments.
3: A randomized trial of chelation therapy in Parkinson's leads to worse outcomes. More patients required levodopa in the iron chelation group than in the control group (22% vs. 2.7%) and there was a worsening of disease severity. This is despite significantly reducing brain iron. In other words: a reason to be cautious about chelation therapy. Another take: just because there’s more of something in a disease state doesn’t mean that removing it is definitely going to be helpful, example bajillion and one.
4: Abnormal oligodendrocyte maturation in a Huntington's disease mouse model can be reversed by thiamine and biotin (T&B) supplementation. This supplementation also led to improved neuronal function in these mice.
5: Did octopuses and placental mammals evolve to become more intelligent via a mechanism related to the expansions of their miRNAs?
6: In an autopsy study (n = 27), the thickening of small arteries was not associated with myelin loss, but the abnormal accumulation of collagen in veins was. Sydenham said “man is as old as his arteries” — maybe he shouldn’t have slept on the veins.
7: Study uses optogenetics to enhance the mitochondrial membrane potential in C. elegans. Finds that this makes them live longer. In other words, this is causal evidence that mitochondrial dysfunction contributes to aging.
8: I think people don't always realize how slow the pace of medical innovation is. One example: here's a registered trial that is attempting to use a three-tiered system for prostate cancer screening. It uses PSA, a kallikrein panel, and an MRI prior to actual prostate biopsy.
It was registered in 2018. They're aiming to include 17,400 participants. This seems like an important study and I definitely don’t want to discount its value. But the expected completion date is not until 2032.
It seems like more than half of the people I know expect the world to be much, much different in 2032, mostly due to AI. At a minimum, isn't there a reasonable chance that we will have a better method for screening than PSA, kallikrein panel, and MRI by 2032 anyway? Like a better kind of non-invasive imaging that leverages AI? Or some new blood test? How should we think about possible future innovations in AI when designing and funding trials like this? One argument is that we should just be conservative and ignore the possibility of AI advances because things often happen slower than we expect, as self-driving cars have.
9: Over the past 20 years, average work hours by physician fathers have decreased by 11% and average work hours by physician mothers have increased by 3%.
The big story of course over this time period has been advanced practice professionals (e.g., nurse practitioners). Their collective hours have increased by 71% and this seems likely to accelerate in the coming decade. It may be that this is basically the only way for medicine to address Baumol’s cost disease. The second-order effects of this change are difficult to predict.
10: December 2022 interview with Ray Kurzweil. It seems to me that a lot of what Ray predicted years ago has been coming to fruition to a greater extent than what people expected. That said, I still strongly disagree with him on some things! For example, the likelihood of near-term anti-aging and rejuvenation treatments. Or the ability to infer someone’s internal personal identity based exclusively on external lifelogging data.
11: Human brain data set with 3d reconstructions. The underlying data is MRI alongside with 800 microscopy sections stained for cell bodies and nerve fibers. These came from two donated brains with postmortem intervals of less than 24 hours that were preserved via whole-body perfusion.
12: Adam Marblestone on how large language models will affect science.
13: Ketamine randomized trial (n = 156), published in Feb 2022, new to me. Ketamine had strong efficacy in decreasing suicidal ideations... for 3 days. The strongest effect was seen in people with a diagnosis of bipolar disorder. By week 6, the effect was no longer stronger than the placebo. There was no significant difference in suicide attempts. One person died of suicide in the 6-week trial, who was in the ketamine group.
14: In an observational study of people with a diagnosis of amphetamine or methamphetamine use disorder, prescription of lisdexamfetamine (Vyvsane) and methylphenidate (Ritalin) were associated with decreased all-cause mortality, while prescription of benzodiazepines were associated with increased all-cause mortality.
15: New randomized trial (n = 276) finds that an 8-week program of mindfulness-based stress reduction (requiring 45 minutes a day) was as effective as escitalopram (Lexapro) in decreasing anxiety as measured by a global impression of severity scale.
16: Why didn't we get the four-hour workday that was predicted in the early 20th century? It turns out we kind of did when you factor in how education has lengthened, how work has improved, and that retirement was invented.
17: Hepatitis B virus and the bacterium that causes leprosy were found in a present-day German burial site of people who lived in the 600s and 700s CE. Hepatitis B prevalence was quite high — 7% of adults and 26.6% of subadults — suggesting it was endemic. It is hard to adequately appreciate how much infectious disease has shaped human history.
18: Reporting by Michael Powell about the naming of the James Webb Telescope.
19: Binge drinking is associated with elevated measures of epigenetic age — for example, an increase of 1.38 years for any recent binge drinking (p < 0.001). Standard caveats of correlational studies apply.
20: Lecanemab — the new monoclonal antibody for Alzheimer's disease approved this month — seems to be associated with cerebral hemorrhage.
21: Review and computational analysis to assess current liter-scale organ vitrification approaches. The three main problems are ice formation, fractures due to thermal stress, and cryoprotectant toxicity. One key barrier: is the lack of volumetric cooling methods, limiting the maximum cooling rate.
22: Why we need to do metagenomic sequencing in septic patients, and why this isn't just a matter of money, but also the conservative culture of medicine that wants to stick to what it knows (ie blood cultures):
23: Why pre-publication peer review is the worst:
24: Why Paul Erlich was so wrong in his predictions about the problems with population growth. The key point here is that people have a tendency to adapt to and solve problematic exponential trends in ways that are hard to predict. Future cares, future cures:
25: Study presents evidence for a hierarchical neural representation of space in rats. Here’s one of their diagrams explaining their model:
Each colored point depicts a neuron. The 2D plane is used to represent abstract neural response properties. The degree of overlap between two disks, representing the response properties of two neurons, is indicative of the similarity between the neurons' response properties. Neurons with larger disks are assigned a higher hierarchical position, as quantified by their z-coordinate in 3D space, compared to neurons with smaller disks. This yields a tree-like structure, which can be viewed as a discrete mesh over the underlying hyperbolic geometry.
I find the use of a tree-like structure to be helpful to think about the connectivity of neurons in the brain and how the structure of their interactions with each other leads to function.