In a review, researchers looked at the growing field of protein engineering . There are two main branches, they say: one where a known, existing protein gets tweaked, and one where pure biophysical modeling is used. AI could be especially useful to both routes, thanks to its ability to wade through the mind-bending number of possible amino acid sequences to predict and find proteins far faster than humans can. One small protein with 100 amino acids has a number of possible sequences that’s more than one thousand time larger than the number of atoms in the universe!

Though AI is promising, the authors cautioned, “Even the most sophisticated machine learning models described, trained on large corpuses of data, are not so accurate that we can simply take a single output and be done.” In other words, even encouraging sequences may turn out to be duds, since only laboratory testing can confirm if they are stable, foldable, and functional.

Out of 2079 traded mammal species, 41% share at least one pathogen with humans , while only 6.4% of nontraded species do. And the longer a species has been traded and shipped, the more likely it is to carry pathogens that can infect us. “On average, a wild mammal species shares one additional pathogen with humans for every 10 years it is present in the global wildlife trade,” the team wrote. Furthermore, those that are sold alive are more likely to share pathogens with us. “It is important to understand that the probability of being infected by playing a piano with ivory keys or wearing fur is almost nonexistent,” explained the study’s lead author, Jérôme Gilpert, in a statement. “The problem lies at the beginning of the chain: Someone had to hunt the animal, skin it, transport it.”

The data suggest it’s not just about stopping illegal trade—while illegally traded species were more likely to host a potential human pathogen, numerous viruses with pandemic potential reside inside legally marketed species. “Cross-species pathogen transmission is an inherent consequence of diverse uses of wildlife by humans,” the team wrote, adding that the findings “underscore the need to strengthen biosurveillance and integrate zoonotic risk considerations into wildlife trade regulations to help prevent future pandemics.”

To measure this, researchers recorded the activity of over 700 neurons in the visual temporal cortex (VTC), a region involved in recognizing complex features and categorizing images, across 16 patients with previously implanted electrodes for epilepsy monitoring. The team tracked how individual brain cells responded as participants viewed a series of images of plants, animals, objects, and faces, and then later reconstructed them from memory.

Many of the same neurons were activated in both cases, with recall engaging 40% of the same visually responsive neurons active when an object was initially perceived. Based on this preserved structure, researchers were able to reconstruct which images patients were recalling from patterns of neural activity. In some cases, the reconstruction captured enough detail to distinguish general categories, like animals or faces. Still, the two processes were not identical, with recall producing less precise representations.

Exactly how these signals are generated and coordinated is still unclear, with authors noting that although the same regions of the brain are involved in both perception and recall, it’s not yet clear how the brain reactivates these patterns; current methods cannot fully distinguish which neurons are involved in each process.

With contributions from Hannah Richter and Ana Georgescu

Do you have a burning science question you can’t seem to find a good answer for? Submit it to Ask Science! Selected questions will receive responses from Science editors right here in ScienceAdviser.

Want more? Catch up on past issues of ScienceAdviser.

If you were forwarded this newsletter by a friend, you can subscribe for free here.