WASHINGTON, DC — The Editorial Board of the Proceedings of the National Academy of Sciences (PNAS) has announced the recipients of the 2025 Cozzarelli Prize, recognizing six exceptional research papers published in the journal during the past year. The award celebrates outstanding scientific achievements across the disciplines represented by the National Academy of Sciences.
Each year, the prize is given to one paper in each of the six major research categories that structure the academy. The winning papers were selected from more than 3,600 studies published in PNAS throughout 2025. In addition to the winners, the editorial board also named six finalist papers—one in each category—acknowledging their significant contributions to scientific advancement. The honorees will be formally recognized at the NAS Annual Meeting scheduled for April 2026.
Established in 2005 as the “Paper of the Year Prize,” the award was renamed in 2007 to commemorate former PNAS editor-in-chief Nicholas R. Cozzarelli. Since then, the Cozzarelli Prize has become one of the journal’s highest distinctions, honoring research noted for both originality and scientific impact.
In the category of Physical and Mathematical Sciences, the winning study explored how specialized machine-learning techniques can uncover previously hidden physical principles in dusty plasmas. Dusty plasmas consist of ions, electrons, and electrically charged dust particles and are frequently found in cosmic environments. Because numerous forces influence the motion of these particles, accurately modeling such systems has long posed a challenge for physicists.
The research team developed a machine-learning system trained on three-dimensional trajectories of dust particles observed in laboratory experiments. The algorithm successfully identified the forces acting between particles with remarkable precision. This approach allowed researchers to estimate particle charge in ways that diverged from existing theoretical expectations. Because dusty plasmas represent a type of many-body system—where large numbers of interacting components produce complex behavior—the findings suggest that machine learning could eventually help scientists derive new physical laws governing similarly complex systems, from colloidal fluids to planetary formations.
In Biological Sciences, the winning paper examined one of the most pivotal events in the history of life: the emergence of eukaryotic cells. Eukaryotes differ from simpler prokaryotic organisms because they possess internal structures such as a nucleus and specialized organelles. These cellular innovations ultimately enabled the evolution of complex multicellular organisms.
To investigate how this transformation occurred, the researchers analyzed genetic and protein characteristics from 6,519 species across the tree of life. Their analysis revealed that in early prokaryotic organisms, the length of genes and their corresponding proteins tended to increase together over evolutionary time. However, once eukaryotic cells appeared, protein size stabilized at around 500 amino acids while gene length continued to expand. The increase in gene size was mainly due to noncoding sequences that regulate gene activity. The authors propose that this pattern reflects a biological compromise: producing extremely large proteins demands substantial energy, so evolution favored maintaining manageable protein sizes while allowing regulatory DNA to expand.
The Engineering and Applied Sciences prize went to a study introducing a new design for absorbing low-frequency sound. Traditional acoustic metamaterials that target low-frequency noise often rely on air resonance inside rigid structures, which can be bulky and only effective within narrow frequency ranges.
The award-winning researchers created a composite material that combines air resonance with structural damping. The design features a soft viscoelastic shell enclosed within a rigid frame. As sound waves interact with the structure, the shell deforms and dissipates energy through viscoelastic damping. Tests showed that the device could absorb more than 97 percent of sound within a broad low-frequency band between 227 and 329 hertz. This approach could lead to more compact and adaptable technologies for noise reduction in buildings, transportation systems, and industrial settings.
In the Biomedical Sciences category, the prize recognized a study investigating a potential strategy to protect the brain from damage associated with neurodegenerative disease and traumatic injury. The research focused on the blood–brain barrier, a critical protective layer that regulates the passage of substances between the bloodstream and the brain. When this barrier deteriorates, it can contribute to neurological decline seen in conditions such as Alzheimer’s disease and traumatic brain injury.
The researchers examined an enzyme known as 15-hydroxyprostaglandin dehydrogenase (15-PGDH), which breaks down anti-inflammatory molecules. Their findings revealed elevated levels of this enzyme in mouse models of Alzheimer’s disease and traumatic brain injury, as well as in human brain samples affected by these conditions. By inhibiting the enzyme in mice, the team was able to reduce oxidative stress and inflammation while preserving the integrity of the blood–brain barrier. The treatment also protected cognitive performance. Notably, improvements occurred even without reducing amyloid plaques in the Alzheimer’s model, suggesting an alternative therapeutic pathway that could help safeguard cognitive health.
The Cozzarelli Prize in Behavioral and Social Sciences went to a study exploring the implications of advanced artificial intelligence for survey-based research. Surveys remain a cornerstone method across many fields, from public health to political science. However, the study demonstrated that autonomous AI agents powered by large language models can convincingly complete online surveys.
These agents can generate coherent responses, adopt specific demographic personas, and even imitate vocabulary patterns and spelling errors typical of particular education levels. They also remember earlier answers and integrate complex reasoning across multiple questions. The research warns that AI systems may soon be able to evade current detection tools designed to identify automated responses. Even without malicious intent, AI-generated participants could infer a study’s hypothesis and unintentionally produce biased results. In more concerning scenarios, coordinated use of such agents could manipulate public opinion polling or disrupt data collection used in health monitoring and social research.
In the Applied Biological, Agricultural, and Environmental Sciences category, the winning paper addressed the growing threat of plastic pollution in marine ecosystems. Scientists conducted a large-scale analysis using data from more than 10,000 necropsies of marine animals, including seabirds, marine mammals, and sea turtles. The goal was to assess how ingesting macroplastics—larger plastic debris—affects survival rates.
The researchers found evidence of plastic ingestion in 35 percent of seabirds, 12 percent of marine mammals, and 47 percent of sea turtles examined. Using statistical modeling, the team estimated mortality risks based on the number, volume, and type of plastic pieces consumed. Their analysis suggested that ingesting approximately 23 pieces of plastic poses a 90 percent mortality risk for seabirds. For marine mammals, the threshold was around 29 pieces, while sea turtles could face similar risk at about 405 pieces due to differences in body size and physiology. The results provide a framework for assessing how plastic pollution affects marine wildlife and could guide conservation strategies aimed at reducing ecological damage.
Beyond the winning papers, PNAS also recognized six additional studies as finalists across the six categories. These papers address topics ranging from air–sea gas exchange and transmissible cancers to electric-field sensitivity in biological membranes and rising riverine heat waves.
Published by the National Academy of Sciences, the Proceedings of the National Academy of Sciences is among the most widely cited multidisciplinary scientific journals worldwide. The journal features research spanning biological, physical, and social sciences, as well as mathematics and applied disciplines. New articles are released online daily and compiled into weekly issues, continuing a long tradition of presenting influential scientific discoveries to the global research community.







Leave a Reply