Ученые из Йельского университета разработали систему для улучшения экстренной помощи и новый метод ускорения научных расчетов

This month’s “Insights & Outcomes” brings the summer sizzle with news about the adaptability of gut bacteria, a statistics success story, and a Yale professor who led a high-profile discussion on the national research agenda. There’s also news about a Yale-developed system for improving emergency care. As always, you can find more science and medicine research news on Yale News’ Science & Technology and Health & Medicine pages. How gut bacteria reorganize under stress When their food supplies become scarce, bacteria must adapt quickly to survive. One way they do it is by reorganizing a key gene-regulating protein into tiny droplet-like structures inside the cell. A new Yale-led study reveals how nutrient scarcity drives droplet formation. The research, led by molecular biologist Aimilia Krypotou in the laboratory of geneticist Eduardo A. Groisman, the Waldemar Von Zedtwitz Professor of Microbial Pathogenesis at Yale School of Medicine (YSM), uncovered the molecular mechanism that controls the formation of these droplet structures (known as biomolecular condensates) in the gut bacterium Bacteroides thetaiotaomicron, which is commonly found in healthy individuals. The study is published in The EMBO Journal.  Previous work in the Groisman lab showed that the bacterial protein Rho — which is known to play a part in transcription termination — forms condensates when nutrients run low, enabling the bacterium to continue thriving in the mammalian gut. But up to now, researchers didn’t know how this process was controlled. “People have known about biomolecular condensates for some time now,” Groisman said. “Having established that the bacterial protein Rho forms condensates and that these condensates are critical for the normal lifestyle of the organism, we wanted to understand the formation of Rho condensates.” The new study identified a collection of short amino-acid sequences within Rho that govern interactions within Rho and between Rho and other molecules – and thus promote or restrain condensate formation. The researchers found that RNA is required to drive condensate formation and that ppGpp — a stress-signaling molecule produced during bacterial starvation — binds directly to Rho and promotes condensate formation. Together, these mechanisms allow the bacterium to respond rapidly to changing environmental conditions, particularly during periods of nutrient deprivation — which our gut bacteria face daily. The findings provide new insight into how cells organize and regulate complex biological processes by revealing a molecular blueprint for how environmental signals reshape gene regulation and cellular behavior. Improving the safety of emergency care Emergency departments (ED) are uniquely high-risk environments. Traditionally, quality and safety work in hospitals has often been retrospective, with clinicians reviewing cases after something goes wrong and then trying to improve. In a new study, Yale researchers turned that process into a continuous learning system. They developed a new model called SENTRY, which brings together multiple safety signals and organizes them into a structured review and tracking process. The goal is not simply to identify individual gaps but to recognize patterns across the system and use those patterns to redesign care. “In simple terms, we built a system that helps the ED learn from itself,” said Suresh K. Pavuluri, assistant professor of emergency medicine and associate director of quality and patient safety in the Department of Emergency Medicine at YSM and first author of the study published in New England Journal of Medicine Catalyst. In the ED, many quality and safety events are not isolated. A medication delay, a missed diagnosis, a handoff problem, or a delay in recognizing deterioration may seem like separate events, but when you look across many cases, patterns begin to emerge. For this study, researchers moved beyond reviewing cases one at a time and created an infrastructure that could detect recurring vulnerabilities, prioritize the highest-risk problems, and support system-level solutions. The researchers found that a structured safety surveillance system can identify patterns that may otherwise remain hidden when cases are reviewed individually. Their findings suggest that safety improves when health systems stop relying only on individual vigilance and instead build systems that make the safer action easier and more reliable. Other Yale authors include Rohit Sangal, Arjun Venkatesh, Reinier van Tonder, and John Sather. For simulations, there’s no place like NOEM Many computing problems in science and engineering are governed by partial differential equations — mathematical rules that describe how quantities such as heat, pressure, stress, or fluid flow change across space and time. For decades, one of the most trusted ways to solve these equations has been what is known as the finite element method, which breaks a complex system into many small pieces. This approach is powerful and flexible, but it is prohibitively expensive when a system contains fine-scale features, complex geometries, or many repeated components. Emerging machine learning methods offer the promise of faster simulations, but they often require costly training data and may be difficult to reuse across large systems. In a new study in Nature Computational Science, Lu Lu, an assistant professor of statistics and data science in FAS and of chemical and environmental engineering in the Yale School of Engineering & Applied Science, and postdoctoral researcher Weihang Ouyang introduced the neural-operator element method (NOEM) — a framework designed to combine the strengths of both approaches. Instead of replacing the finite element method with a machine learning model, NOEM embeds machine learning directly into the finite element framework. Complex subdomains that would normally require many small finite elements are replaced by a single neural-operator element, while the rest of the system continues to be modeled using conventional finite elements. In this way, NOEM retains the modeling flexibility and scalability of finite element methods while gaining the efficiency of learned neural operators. For the study, the team successfully demonstrated NOEM across a variety of problems and simulations. “The significance of NOEM is that it offers a path toward faster and more scalable scientific computing without forcing researchers and engineers to choose between classical numerical methods and machine learning,” Lu said. “Many real-world systems — from engineered materials to buildings, energy devices, and environmental systems — are assembled from repeated or standardized components. NOEM suggests that these components could eventually be represented by pretrained libraries of neural-operator elements, allowing high-fidelity simulations to be built more efficiently by interconnecting reusable computational modules.” Yale researcher moderates ‘State of the Science’ panel Brandon Ogbunu, an associate professor of ecology and evolutionary biology in the Faculty of Arts and Sciences (FAS) recently moderated a panel discussion in Washington, D.C., at the third annual State of the Science address, which was given by National Academy of Sciences President Marcia McNutt. The panel included national research leaders and policymakers from academia, government, the private sector, and philanthropy, including Stacie Bloom, president and CEO of the Alfred P. Sloan Foundation; Roy Blunt, chairman of leadership strategies advisory services for HB Strategies and a former U.S. senator from Missouri; Elizabeth Bruce, director of science and innovation policy at Microsoft; Ethan Klein, U.S. Chief Technology Officer and associate director of the White House Office of Science and Technology Policy; and Geri Richmond, vice president of research and innovation at the University of Oregon. Panelists discussed how the U.S. research enterprise can meet current challenges and persevere in making essential contributions to the nation in the years ahead. “I believe it is important to create focused moments when the American scientific community — including scientists and citizen scientists — can engage directly with the public,” Ogbunu said. “Opportunities for this kind of exchange are rare, but they are increasingly essential.” The State of the Science is an initiative of the Strategic Council for Research Excellence, Integrity, and Trust. It is supported by the Gordon and Betty Moore Foundation, the Annenberg Public Policy Center at the University of Pennsylvania, and the National Academy of Sciences Whitney and Elizabeth MacMillan Fund. Karen Guzman, Meg Dalton, and Jim Shelton contributed to this report. Research Redux: How AI is changing the routine mammogram Beyond the brain: Organs help shape the nervous systems that control them Growing a new ‘leaf’ that harnesses sun, water and carbon dioxide to make liquid fuel Is it time to expand our thinking about dark matter? A new study says yes