An international team of researchers, including Earth-Life Science Institute (ELSI) at the Institute of Science Tokyo, has developed a protein language model that brings together two fundamental sources of information about proteins: their amino acid sequences and their three-dimensional structures. The model provides researchers with a new way to map relationships across the protein universe and investigate how proteins have evolved over billions of years.
The research was led by Prof. Rachel Kolodny and PhD candidate Guy Yanai of the University of Haifa, Prof. Nir Ben-Tal and graduate student Gabriel Axel of Tel Aviv University, and Specially Appointed Associate Professor Liam M. Longo of ELSI. Kolodny also spent five months as a visiting researcher at ELSI developing approaches to analyze the new model. The findings were published in Proceedings of the National Academy of Sciences (PNAS).
Thousands of protein families are responsible for carrying out nearly every function within living cells. A fundamental question in evolutionary biochemistry is how these proteins are related to one another and where they came from in the first place.
Scientists traditionally organise proteins into hierarchical groups based on their relatedness, somewhat like the genus and species classifications used for living organisms. These carefully curated systems contain decades of scientific knowledge, but advances in artificial intelligence are now creating new ways of exploring relationships across the vast protein universe.
Protein language models can convert a protein into a numerical representation known as an "embedding". One way to think of an embedding is as a kind of postcode: proteins with similar properties tend to receive nearby addresses. Researchers can then visualise these relationships to produce a "protein world map".
However, there is a complication. Proteins contain information in both their amino acid sequences and their three-dimensional structures, and the relationship between the two is not straightforward. Proteins with unrelated sequences can sometimes adopt similar structures, while similar or even identical sequences can produce very different structures.
Most protein language models have approached protein sequence and structure separately. Even models that use both kinds of information do not necessarily place the sequence and structure of the same protein at the same location on a protein map.
The researchers developed a model called Contrastive Learning Sequence-Structure, or CLSS, designed to produce highly similar embeddings for both the sequence and structure representations of a protein.
CLSS uses an approach called contrastive learning. During training, the model receives protein sequences and their corresponding structures and learns to produce similar embeddings for sequence-structure pairs while separating unrelated pairs. The result is a shared map in which a protein representation occupies a similar location within the protein world map, regardless of whether its sequence or its structure was used.
When compared with other state-of-the-art protein language models, CLSS successfully brought sequence and structure information together in a cohesive map. Its representations also closely reproduced relationships recorded in the expert-curated ECOD and CATH protein classification systems, even though those classifications were not provided to the model during training.
The model also performed strongly in classification tests, demonstrating that combining sequence and structure information can produce more informative representations of proteins.
AI can trawl vast amounts of health data which is not specifically related to migraines, but which can be used to identify people predisposed to the condition.
A new study from NTNU - the Norwegian University of Science and Technology - reveals that migraines may leave a biological pattern throughout the body, traces of which can be identified by artificial intelligence.
Dr Chi Kyung Kim, CMIO at Korea University Guro Hospital, shares how hospital-startup collaborations leverage agentic AI to monitor home exercise, diet and medication routines.
Insilico Medicine, a clinical-stage biotechnology company powered by generative AI, today announced that it has dosed the first patient with Rentosertib (known as ISM001-055 / INS018_055) in GENESIS-IPF-3, a Phase III clinical trial (NCT07687459, CTR20262475) at Peking Union Medical College Hospital, while Shanghai Pulmonary Hospital has also enrolled its first patient on the same day.
The Phase III clinical trial is a prospective, multi-center, randomized, double-blind, placebo-controlled, parallel-group study designed to systematically evaluate the efficacy and safety of once-daily Rentosertib administered over 52 weeks. The study is led by Professor Zuojun Xu of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences as the Leading Principal Investigator (Leading PI), with Academician Nanshan Zhong of the Chinese Academy of Engineering, a renowned expert in respiratory medicine, and President Chang Chen of Shanghai Pulmonary Hospital serving as Co-Leading Principal Investigators (Co-Leading PIs).
As Leading PI of the study, Professor Zuojun Xu from Peking Union Medical College Hospital, Chinese Academy of Medical Sciences, commented: "TNIK, the target driven by AI, had never previously been linked to fibrosis. This perhaps indicates that AI is carving out a path distinct from traditional research paradigms in target discovery for complex diseases. Dosing the first patient marks another key clinical milestone for Rentosertib; after preliminary findings in the 12 week duration phase IIa study showed promising results, including improvements in lung function among patients with idiopathic pulmonary fibrosis (IPF). This Phase III study is designed to validate the phase IIa study findings in a larger cohort of idiopathic pulmonary fibrosis patients treated for a longer duration of 52 weeks. From initiating Phase III to obtaining final regulatory approval, it will take three to four years under favorable conditions. We look forward to Rentosertib achieving its primary endpoints and successfully securing market approval to benefit IPF patients."
Till today, Insilico has published multiple peer-reviewed papers to document the milestones of Rentosertib along the R&D journey, all the way from target identification to Phase IIa positive results. In 2024, a Nature Biotechnology publication demonstrated the AI-driven early discovery process of Rentosertib. In 2025, a Nature Medicine study reported Phase IIa results of Rentosertib, demonstrating a promising dose-dependent efficacy trend. Most recently, in 2026, a Nature Biotechnology study revealed a consistent reduction in biological age after Rentosertib dosage, across six independent biological aging clocks.
An artificial intelligence model that analyzes women's past and recent annual 3D mammograms is more effective at predicting five-year risk of developing breast cancer than a tool that uses only the most recent, single 3D mammogram, as well as an AI model that analyzes 2D mammograms, a new study shows.
Researchers at NYU Langone Health and its Perlmutter Cancer Center developed the deep-learning tool, called NYU-DRP, using a woman's 3D mammograms taken over multiple years, also known as longitudinal digital breast tomosynthesis (longitudinal DBT).
Published in the American Journal of Roentgenology online Aug. 12, the study showed that NYU-DRP was superior to other test models at predicting a woman's risk of breast cancer after five years, correctly ranking those at higher risk 72 percent of the time. Single DBT and AI-assisted 2D testing correctly predicted higher-risk cases 70 percent and 68 percent of the time, respectively.
Artificial intelligence (AI) holds profound potential to reshape public health. Although its application accelerates threat detection and targeted interventions, the rapid deployment has outpaced the regulatory, validation, and equity safeguards standard to traditional health interventions.
In a recent analytic essay, Dr. Terry Adirim from the Department of Pediatrics and the Department of Preventive Medicine and Biostatistics, Uniformed Services University of the Health Sciences, Bethesda, MD, and Dr. Amy Molten, MD, from the Department of Pediatrics, Tufts University School of Medicine, Boston, MA, examined the ethical challenges of AI use in public health, highlighting the potential for these systems to reinforce existing health inequities. The authors contend that responsible AI deployment requires governance addressing the needs of historically marginalized populations. The study was published online in the American Journal of Public Health on September 9, 2026.
"The most fundamental problem with AI in public health work is structural rather than technical," says Dr. Adirim.
The analysis identifies critical ethical risks like discrimination, surveillance, and privacy violations, across historically marginalized populations, including children, minoritized communities, Indigenous peoples, and people with disabilities.
While analyzing diverse datasets, AI can produce biased outputs. Furthermore, using commercial data like location tracking blurs the line between public health surveillance and consumer privacy, while behavioral AI tools risk manipulation, misinformation, and compromised autonomy.
AI that can sift through a patient’s health history and previous interactions could enable more personalized and continuous care. But that poses new questions around patient privacy and data security, says HealthTap CEO Sean Mehra.
The human genome contains approximately 3 billion DNA letters, creating more than 9 billion possible single-letter changes. Testing the effects of each change in a laboratory would be practically impossible. Google DeepMind's new AlphaGenome Atlas, available beginning today, gives scientists a comprehensive, searchable resource designed to accelerate understanding of the human genome.
Stowers Institute for Medical Research Investigator Julia Zeitlinger, Ph.D., partnered with the Google DeepMind team led by Vice President of Science and Chief Scientist Žiga Avsec, Ph.D., to map and interpret the patterns in DNA that regulate biological processes inside cells. At the same time, additional scientific collaborators from leading institutions across the United States and England helped test how the new resource could be used to identify impactful genetic variation in humans. The work is now available as a preprint on bioRxV.
The one-petabyte resource contains artificial intelligence-generated predictions for the molecular effects of more than 9 billion possible changes, creating what Google DeepMind describes as the most comprehensive catalogue of its kind.
Zeitlinger has made significant contributions to the field of gene regulation and computational biology. In 2019, in an international collaboration that included Avsec, Zeitlinger and her team at the Stowers Institute developed a powerfulAI framework, BPNet.This framework is now widely used to extract and dissect the DNA sequences that explain genome-wide biological data. Just last month, her lab unveiled anew AI method, PISA,which generates high-resolution visualizations of what AI models have learned from DNA.
Stowers Institute Bioinformatics Scientist and Zeitlinger Lab member, Melanie Weilert, served as a lead author on the AlphaGenome project. With her deep expertise in interpretating AI models, she helped build the AlphaGenome Atlas resource to ask one of biology's biggest questions: How does a cell know which genes to turn on and off?
"This is a very difficult problem because every cell type speaks a slightly different language, making it hard to know which rules are general," Zeitlinger said. "With AlphaGenome, we can quickly query many cell types and look for general patterns by which genes are activated and repressed."
As documentation automation spreads, health systems must preserve clinician verification, define accountability and give frontline nurses authority from vendor selection through ongoing oversight.
Diabetes is one of the most common chronic diseases, affecting around 66 million adults in Europe - a figure projected to rise to more than 72 million by 2050. For people with type 1 diabetes, keeping glucose levels within a safe range requires continuous monitoring and careful insulin management. Yet even with experience and careful treatment, unexpected fluctuations can occur, potentially leading to hypoglycemia or hyperglycemia.
Glucose levels are influenced by far more than food and insulin. Physical activity, stress, hormonal changes, sleep quality, circadian rhythms and individual differences in metabolism can all play a role. Anticipating fluctuations could therefore help assess risks and support better-informed decisions about diabetes management.
Continuous glucose monitoring systems already record glucose levels every few minutes, generating vast amounts of data. Researchers at Kaunas University of Technology (KTU) are now exploring whether artificial intelligence (AI) can uncover patterns within these data. Their approach brings together glucose readings with information on insulin delivery, carbohydrate intake and physical activity.
Health systems are testing whether nursing documentation automation improves presence, note quality, wellbeing and retention – not just reducing the minutes required to finish a note.
Organizations need safeguards in place as agents become more autonomous. Ankit Jain, cofounder and CEO of Infinitus, discusses ways to prevent unintended or unauthorized AI actions from affecting patient care.
