Dispatches

The week’s most astounding developments from the neobiological frontier.

December 8, 2022

Machine learning predicts drug side effects

Designing a drug is a fickle art. It relies not just on answering the obvious—does the medicine actually work—but addressing many other fundamental questions as well, like solubility, gut absorption, delivery methods, metabolism, excretion, and others. Side effects are another big risk, especially when they are severe, but sometimes those aren’t discovered until a drug is already on the market, which first and foremost is a tragedy for the people affected. It is also a financial disaster for drug developers, who spend on average $1.3 billion developing and bringing a drug to market. Now researchers at National University of Asuncion in San Lorenzo, Paraguay, have developed an interpretable machine learning framework that can predict the side effects of drugs based solely on the information collected in clinical trials, possibly saving time, money, and human lives. Cell Reports Methods

Neuroimaging-based indicators of pain

Two standard neuroimaging techniques may be able to distinguish different levels of pain in people, something normally assessed through self-reporting. In an analysis of 366 people with EEG and 399 people with fMRI, researchers at the Chinese Academy of Sciences in Beijing found a correlation between certain patterns of brain activation and self-reported pain in response to laser heat pulses of varying intensity. These neural indicators are unique to pain discriminability and do not predict pain sensitivity, nor the ability to distinguish levels of non-painful touch, sight, and sound. Nevertheless it lays the groundwork for objective pain assessment, which could be useful for treating chronic pain, where impaired pain discriminability has emerged as an early symptom and pain discrimination training as a promising therapy. Neuroimaging could also help identify pain in small children, people who are non-communicative, or people who have disorders of consciousness. Cell Reports Medicine

Ultra-processed food and cognitive decline

A survey of 10,775 ethnically diverse adult men and women working as public servants in six Brazilian cities shows that eating more ultra-processed foods was associated with greater cognitive decline. The work, part of the Brazilian Longitudinal Study of Adult Health, was led by researchers at the University of São Paulo Medical School. Compared to the people who ate the lowest amount of ultra-processed foods (0–20 percent of their daily calories), the people in the trial who ate more showed a 28 percent faster rate of global cognitive decline and a 25 percent faster rate of executive function decline over 6–10 years of follow-up. Limiting ultra-processed food consumption, particularly in middle-aged adults, may be an efficient way to prevent cognitive decline, the researchers write. JAMA Neurology

Artificial faces and their human consequences

Psychologists at the University of London have conducted a series of experiments involving generative adversarial networks (GANs)—deep neural networks trained on a large dataset of real faces that are capable of generating impressively realistic novel faces. They found that the computer-generated faces were more likely to be perceived as real than real faces (something other researchers have already found), and they discovered that people displayed increased social conformity toward those faces, trusting them more even though they weren’t real. However when people were informed that a face was computer generated, that eroded the trust. “Our findings point to potentially far-reaching consequences for the pervasive use of GAN faces in a culture powered by images at unprecedented levels,” the researchers write. iScience

Magnetic micro robots may make milk safer

One of the most common and costly diseases on dairy farms is bovine mastitis, an inflammation of the mammary gland caused by infections with Staphylococcus aureus or other things. It impacts milk production, can expose humans to bacterial toxins through the milk, and it’s hard to treat other than by resorting to antibiotics. Researchers at the University of Chemistry and Technology Prague have developed a magnetically controlled robotic technology they call “MagRobots” loaded with antibodies against S. aureus that can remove the bacteria from milk without affecting other naturally occurring microbes, they report. Small

The magic of reindeer antlers

Human fetuses have a remarkable ability to heal cuts without scarring, but as adults we lose this ability. The same is true for most mammals who typically regrow skin through a scarring mechanism whereby cells known as fibroblasts deposit extracellular matrix, express inflammatory proteins, and heal the skin while forming scars. However, the fuzzy “velvet” skin of reindeer antlers has a remarkable regenerative healing property. The animals are able to regrow their functional ornaments at the explosive rate of up to 1 centimeter per day each year. Now, researchers at the University of Calgary in Canada have uncovered the regenerative secret of reindeer antlers. The cells in the antlers, like human fetal fibroblasts, restrain the immune response as they promote regeneration. This could lead to better ways of treating wounds in adults. And wouldn’t that be a magical gift? Cell