Ideal Friction, T Cells Against Cancer, and Video Straight from the Brain: QWERTY Science Digest No. 385

Fundamental science rarely makes breakthroughs one at a time on schedule: more often, results from years of laboratory experiments across entirely different disciplines come together in a single week. This roundup brings together four studies from autumn 2026 that radically change established ideas about friction physics, cancer therapy, neurobiology, and decoding signals from the brain.

Almost perfect lubrication—with no lubricant

At school, they teach that the force of sliding friction does not depend on speed. This is Amontons-Coulomb's law. In practice, it is almost never followed precisely—real materials change their friction slightly as speed changes, and these deviations cause vibrations, wear, and disruptions in power transmission.

A team of scientists from the US and Japan used single crystals of muscovite, a layered mineral and a type of mica already used as a base for solid lubricants. They heated them from room temperature to 200 °C and measured how friction depended on speed. At 200 °C, the dependence disappeared completely. For the first time in experimental history, Amontons-Coulomb's law was observed in its ideal form.

Conventional theories of friction cannot explain this. Electron microscopy showed that at low temperatures, ripplocations—wave-like crystal dislocations—form in the mica layers. At 200 °C, they are absent. The hypothesis is that ripplocations are responsible for the dependence of friction on speed, and their absence produces ideal friction.

The next step is to conduct experiments inside an electron microscope, to observe defects appearing and disappearing in real time. If the pattern is confirmed in other layered materials, it could lead to a new class of solid lubricants and, possibly, a refined law of friction.

T-cell cancer therapy—now universal and cheaper

T-cell therapy works like this: lymphocytes are taken from a patient, modified to target their tumor, and returned to the patient. The problem is that this is a bespoke process. It has to be done separately for each patient, and it takes a long time and costs a lot. Donor cells are unsuitable because they attack foreign tissues.

The new approach is the drug Allo-EASO-T. The researchers used not mature T cells, but hematopoietic stem cells from donor umbilical cord blood. During maturation, they inserted a gene for the NY-ESO-1 receptor, a protein found in many solid tumors. The key is that if the target is set before maturation, almost all mature cells carry this receptor. They have almost none of their own “old” receptors that attack foreign tissues. The graft-versus-host reaction is sharply reduced.

They also preserved a backup mechanism: natural T-killer cell receptors that respond to stress in a cell. These are needed in case a tumor “drops” NY-ESO-1 and tries to evade the main attack. In laboratory tests, the backup helped destroy melanoma, ovarian cancer, and prostate cancer cells that no longer expressed NY-ESO-1.

In mouse models of ovarian cancer and melanoma, the modified cells accumulated primarily in the tumors, their numbers increased by about 100 times, and their activity lasted for several weeks. The control group, which received ordinary donor T cells from mature blood, controlled tumors less effectively; the cells spread to the liver and lungs and caused a graft-versus-host reaction.

According to the team's calculations, a small volume of umbilical cord blood can yield trillions of cells in ~6 weeks. That's around 1000 doses at ~$5000 each. If a validated receptor for another tumor antigen becomes available, it can be added to the platform without a major redesign.

Mice with human cerebral cortex

Neural organoids—three-dimensional clumps of human brain tissue—have been grown since the early 2010s. The problem is that in laboratory dishes, they lack the sensory signals and connections of a living brain. A few years ago, organoids were transplanted into rats: the cells occupied a third of the cortex, but rat neurons mature faster than human ones and did not allow the human cells to develop normally.

The next step was genetically modified mice that do not develop most of the cortex and hippocampus. Human organoids were transplanted into the space that was left. The result: successful engraftment in 25 of 29 cases. Over 3 months, the human tissue grew almost fivefold and occupied more than 90% of the cortex.

Different cell types appeared in the graft, including rare neurons that are almost impossible to obtain in a laboratory dish. The characteristic layered structure of a mature cortex did not form, but human neurons created numerous connections with the mouse nervous system—and this makes it possible to observe human neural circuits in a living brain.

The first test used a cerebral palsy model. After oxygen deprivation, signs of damage appeared in the human tissue, and the mice developed characteristic problems with gait and coordination, like people with cerebral palsy. Next, the researchers plan to study dementia and genetic forms of autism using organoids made from cells from specific patients.

On the ethics question, experts say that what matters is not the proportion of human cells, but the abilities acquired. So far, mice with transplants perform worse than ordinary mice in memory and fine motor skills tests, but better than mice without a cortex—the human tissue partially compensates for impairments, but does not make the animals smarter. Not yet.

Video straight from a mouse's brain

A long-standing challenge in neuroscience is to reconstruct what a person or animal saw at a particular moment from brain activity. Functional MRI is usually used for this, capturing overall activity signals from brain regions. Accuracy remains modest.

A new study went deeper: instead of fMRI, the researchers recorded the activity of individual neurons in the visual cortex of mice. Yes, it's invasive, but these are mice. The advantage is that it reveals how the brain encodes visual information at the cellular level.

For reconstruction, the researchers used a dynamic neural network model from the Sensorium 2023 competition—it predicts the responses of individual neurons to video, taking into account the animal's movement and changes in pupil diameter. The process: first, they trained the model; then they took the actual neural activity during viewing of a new video (active cells were identified by calcium increases), and the algorithm gradually changed the pixels in a blank video until the difference between the predicted and actual activity was minimized.

The result: the system reconstructed a 10-second clip that was not in the training data—using only brain activity. The more neurons included in the analysis, the more accurate the result. The resolution and scene coverage are still limited, but the direction is clear.

The main takeaway is not about spying on dreams—the brain does not store an exact video recording; it constantly interprets and filters signals. Now there is a tool that makes it possible to compare “what the animal saw” with “how the brain encoded it”—and find the difference.

Bonus: the best news from the previous edition

QWERTY viewers voted for the story about prost-6 photoswitchable molecules: they can be dropped into the eye or injected, and when illuminated, they activate surviving neural circuits in the retina through a special protein. In experiments on zebrafish and mice, the molecules restored responses to light without implants or gene therapy. The method does not stop photoreceptors from dying, but offers a way to restore light sensitivity where the receptors have already died.

This article is based on publications in Nature, Science Advances, and open preprints from research groups. The editorial team of seo-mind42.ru.