Action Potentials for October
1: New results from a manuscript creating a digital twin of the fruit fly Drosophila. It’s called NeuroMechFly v2. Like many cutting-edge projects, this is a synthesis of many components, including a model of the body based on a micro-CT scan of one adult fly, as well as simulation modules for vision, olfactory, mechanosensory feedback, and motor output. Here’s their whole schema:
What they added to this project was another module from a separate project, called Flyvis, that created a connectome-constrained visual system model for Drosophila. It’s called “connectome-constrained” because it uses the cells, connections, and synaptic weights derived from a volume electron microscopy data set of the visual system of one fruit fly that was preserved and scanned.
The Flyvis model represents most of the cells as point neurons, to simplify the complex structure of real neurons. The cells have passive leaky linear non-spiking voltage dynamics, reflecting the nature of early visual system neurons in the fly. And chemical synapses are modeled with threshold-linear function to relate presynaptic voltage to synaptic release.
The authors measured the ability with which this model was able to perform a “fly following” task in the NeuroMechFly v2 simulation framework. The fly with the simulated body and visual system had to follow another “leading” fly model across a terrain.
The visual input from the following fly’s eyes was processed through the connectome-constrained model, which simulates the layered visual processing in the fly brain. Then, the activities of certain neurons that seem to be output cell types were used for object detection. The detected object’s position was used to modulate a turning signal for walking. The time step of the simulation was 2 ms.
They found that feedback-driven head stabilization was crucial for reliable fly following, especially over rugged terrain. Without head stabilization, the following fly struggled to track the leading fly consistently when navigating complex surfaces.
I can imagine how simulation models like this could become increasingly biophysically realistic and integrated with more detailed data about how the fly’s brain works over time. Could be useful for… robotics.
2: Another simulation study of the Drosophila brain, this time using the whole brain connectome. They attempted to answer a rarely asked question: why are such a high percentage of neurons in the brain excitatory, as opposed to inhibitory? The percentage of excitatory cells is similar across species and regions, at around 70-80%, although some regions like the hippocampus have an even higher percentage at around 85-90%. Their answer is that this ratio helps the neural network to solve more complex functions, as long as the inhibitory neurons are highly connected.
3: Researchers have found that dopamine signaling in the nucleus accumbens creates a “memory module” for goal-directed navigation, allowing mice to rapidly form spatial memories of safe locations. By optogenetically activating dopamine neurons, they could artificially implant a place memory that guided escape behavior.
4: Comment of the month goes to Matt Gruner who pointed out this interesting paper about developing AI-powered virtual cells on last month’s post and asked: “Is a standard model of cell biology on the horizon?”
5: New study measures the effects of preserving human brain tissue in formalin for 1-20 years on neural antigens. Some antigens are found to decrease in staining intensity over time, while the staining levels for the astrocyte marker GFAP actually increase. Although cellular morphology was not a major focus of the paper, they reported being able to observe typical characteristics of neurons, astrocytes, and microglia even in long-preserved samples, suggesting that there were no dramatic changes in cellular morphology visible by light microscopy over the preservation period.
6: In a world’s first, CAR-T from donor cells is reported to successfully treat 3 individuals for autoimmune disease in China. Using donor cells could enable mass production and dramatically scale the procedure, vs having to manufacture them from the patient’s own cells.
7: One problem with animal longevity studies is that if the control group is short-lived, it can exaggerate the effect of longevity interventions due to statistical effects and other factors. A new study proposes a "900-day rule" for mouse studies, meaning that interventions should only be considered highly effective if the control group median lifespans approach 900 days or if the treated groups significantly exceed this threshold. Sounds kind of like the 5-second rule, but actually completely different.
8: This January a randomized trial of depression treatment in Karnataka, India was published in an economics journal. In this study, participants were randomly assigned to one of four groups: pharmacotherapy alone, “livelihoods assistance” alone, a combination of both, or a control group receiving neither intervention.
Participants who received pharmacotherapy underwent eight months of treatment, primarily with SSRIs, which are available at low cost in India. Pharmacotherapy alone reduced depression severity by 0.14 standard deviations compared to the control group, but the effects dissipated after the treatment ended.
The livelihoods assistance program consisted of two group meetings focused on work-related challenges, along with personalized support to help participants find employment or generate income. It alone had a smaller, statistically insignificant effect on reducing depression, lowering symptoms by only 0.08 standard deviations during the intervention.
In contrast, combining pharmacotherapy with livelihoods assistance had a greater impact, reducing depression severity by 0.26 standard deviations, with benefits lasting up to 26 months after the intervention. None of the interventions significantly improved work time or earnings.
It’s only one study without huge effect sizes so I wouldn’t put too much stock into the results. I found it interesting mostly because it reminds me of David Smail’s book “Power, Interest and Psychology”, which argues (inter alia) that most people have mental health problems because their lives are bad in various ways — i.e. external, material realities such as economic conditions, social status, or political power. And that directly addressing the ways in which their lives are bad can actually be more helpful than trying to help their mental health via therapy, because their distress might be a reasonable response to their social environment.
Although I disagreed with some of its macro political economics, overall I found it to have a pretty unique and interesting perspective. I thought of it because maybe the participants in this study found it encouraging that people were actually trying to help them with their economic problems alongside prescribing medication.
9: Study finds that longitudinal improvement in cognition in schizophrenia spectrum disorders is rare, with the exception of early onset psychosis.
10: A petition calling for the FDA to remove the Clozapine Risk Evaluation and Mitigation Strategy (REMS), discontinue mandatory blood testing for prescription refills, and eliminate prescriber and pharmacy certification requirements to improve access to the medication. The blood testing requirement for refills in particular is a huge issue. Brought to you by The Angry Moms. I signed the petition because I completely agree.
11: A study of 3 million working-age New Zealanders finds that among those who accessed public secondary health services, OCD was more commonly identified in women (by almost 1.5x), younger adults (more than 2x higher in people age 18-24 than 55-64), and those of European ethnicity. Overall, only 0.18% were identified with OCD, likely underestimating true prevalence, and probably many selection biases at play.
12: Sarah Constantin argues that psychotic delusions are more about intensity and preoccupation with certain themes, like mistrust or self-importance, as opposed to holding false beliefs. She points out that even true ideas can be experienced in a psychotic way, and that mental illness shapes the kinds of thoughts one is drawn to, but doesn’t automatically mean the person is wrong. This fits with Vaughan Bell and colleagues' work on this topic, showing that delusions tend to be highly socially themed and related to coalitional cognition, rather than just being generic “irrational beliefs.”
13: New study finds that water confined in nanoscale spaces has a higher glass transition temperature (-103°C to -73°C) than previously expected for bulk water (-137°C). This suggests that some amount of the water in biological systems, which is often confined within cells and between molecules, may vitrify at higher temperatures than previously thought.
14: Aldehyde groups: great in molecules used for structural preservation of tissue, not so great in medicinal chemistry because they are so reactive.
15: Interesting Reddit thread on what a covalent bond is.
16: Several Drosophila connectome papers finally drop at Nature and are covered by the NYT. Key quote: “Sebastian Seung, another leader of the project at Princeton, said the simulations reminded him of long-running speculations about how “mind uploading” could allow us to transfer our brains into computers. “Mind uploading has been a science fiction, but now mind uploading — for a fly, at least — is becoming mainstream science,” Dr. Seung said.”
Of course, a few of these papers have already been covered here on NN. When I can, I try my best to tell you about these things even before the Springer Nature industrial complex gives its blessing.
On a related note, Philip Shiu of the Shiu et al paper is now on the team at Eon Systems. They also have job openings, including one for someone who will help develop a pipeline in expansion microscopy.
17: Dario Amodei, CEO of Anthropic (need I say more?), on mind uploading: “One topic that often comes up in sci-fi depictions of AI, but that I intentionally haven’t discussed here, is “mind uploading”, the idea of capturing the pattern and dynamics of a human brain and instantiating them in software. This topic could be the subject of an essay all by itself, but suffice it to say that while I think uploading is almost certainly possible in principle, in practice it faces significant technological and societal challenges, even with powerful AI, that likely put it outside the 5-10 year window [after powerful AI is developed].” This could be thought of as an excellent example of an apophasis.
#13: Very interesting and apparently has nothing to do with pressure.
Also while I chuckled at "Springer Nature industrial complex", it's sadly a true description.