Clinical sleep deprivation is sometimes used as a short-term solution to rescue people from severe depression because disrupting sleep can trigger mania and reverse depression in people. The same is true in mice. Depriving mice of sleep causes them to become hyperactive, more social, more sexual, and more likely to fight. But we’ve never understood exactly how this works in the brain. Now neuroscientists at Northwestern University in Evanston, Illinois, have uncovered key parts of this underlying mechanism in mice. They found that acute sleep loss in mice causes the neurotransmitter dopamine to be released in certain parts of their brains, inducing the formation of dendritic spines, which form network-like connections between neurons. That corresponds with elevated mood and behavioral changes in the animals. But optically disassembling these new connections reverses the effect—all of which, they write, could provide the critical insight that could “pave the path to developing new antidepressants and mood stabilizers.” Neuron

The flatland fallacy holds that people tend to oversimplify complex landscapes. We purposely perceive with far less complexity in our minds what’s really in front of us: Instead of the studded peaks of the Himalayas, we see nothing but a smooth, glacier-carved flatland, and we think we can cross it with ease. Nothing would seem a more appropriate cautionary tale for drug design than this fallacy because nothing is simple in drug design. Even small molecules can be inordinately complex, with seemingly insignificant alterations to a chemical translating into major changes in its activity, solubility, metabolism, toxicity, or ease of synthesis. That’s where seasoned experts come in. Medicinal chemists in particular apply their hard-won knowledge to deciding what makes a good “lead compound” and how to refine them into effective drugs. Could AI do the same thing? That’s what researchers at the Novartis Institutes for BioMedical Research in Basel, Switzerland, and Microsoft Research’s AI4Science in Cambridge, England, are hoping. They trained an AI to do drug design using feedback from 35 Novartis chemists over several months, and they are making their model freely available under an open source license this week. Nature Communications

Animal models have shown that sleep deprivation promotes the accumulation of amyloid-beta and tau in the brain—the two proteins whose aggregation is associated with Alzheimer’s disease. Now doctors at Monash University in Victoria, Australia, have shown in a retrospective human study that not getting enough of a particular type of sleep could indeed be a risk factor for dementia. What happens when we sleep is that the brain undergoes a sort of maintenance cycle called “glymphatic clearance” where glial cells help to clear metabolic waste like amyloid-beta. This takes place during the deepest non-REM sleep (called slow-wave sleep), which can decline as we age. Looking at 346 people from the Framingham Heart Study (mean age 69) who completed two overnight polysomnography (sleep) studies, the doctors found that for every percentage a person’s slow-wave sleep decreased, they had a 27 percent increased risk of dementia. The good news: Sleep may be a modifiable risk factor. JAMA Neurology

Researchers at the Howard Hughes Medical Institute’s Janelia Research Campus in Ashburn, Virginia, and Beth Israel Deaconess Medical Center in Boston are announcing this week that rodents have powerful imaginations. That doesn’t mean we have newfound appreciation for their cleverness—we already knew they were. It means they can draw “cognitive maps” in their brain’s hippocampus region that represent familiar things, places, or memories—and voluntarily recall them later. They demonstrated this using brain-machine interfaces and mini rat-sized VR headsets to see how they virtually navigate a familiar setting. Science

Many experts have come to the realization that the health hype around happiness may be overrated, and now psychologists at Texas A&M University are saying anger may actually be underrated. Like all emotions, anger may be nothing more than a natural psychological reaction to the problems around us—perhaps even an evolved tool to help us solve them. To test this idea, the researchers conducted a series of puzzle-solving experiments with more than 1,000 people. They found anger increased the number of puzzles a person could solve, decreased reaction times, and increased scores on certain video games. “Our research adds to the growing evidence that a mix of positive and negative emotions promotes wellbeing, and that using negative emotions as tools can be particularly effective in some situations,” one of the researchers said in a statement to the press. Journal of Personality and Social Psychology