It’s possible that I shall make an ass of myself. But in that case one can always get out of it with a little dialectic. I have, of course, so worded my proposition as to be right either way (K.Marx, Letter to F.Engels on the Indian Mutiny)
Powering microrobots has long been a chemical engineering problem. A 1.7 g microrobot can crawl for only about two minutes on commercial batteries. To run for two hours, it would need a battery weighing 94% of its total mass, with an energy density of 5382 Wh kg−1, far beyond today’s lithium-ion technology. Below 1 cm3 or 1 g, microbatteries lose energy density even faster because packaging and conductive materials take up a growing share of their mass.
From carrying batteries to walking on energy
A team led by Yue Gao at Fudan University’s Department of Macromolecular Science tried a different route. Instead of making batteries smaller and denser, the researchers asked whether the robot could harvest energy from the surface it walks on. Active materials such as aluminum, zinc, silicon, tin, and lead are common in pipes, buildings, data centers, and equipment. If a robot’s foot could trigger a chemical reaction on contact, its runtime would no longer be limited by onboard storage.
The team’s answer is an open electrochemical power system: a crosslinked potassium polyacrylate membrane attached to the microrobot’s foot. It converts chemical energy from the substrate and from atmospheric moisture and oxygen into electricity.
A skin that powers, grips, and senses
The membrane is made in one step by polymerizing acrylic acid monomer, crosslinker, initiator, and potassium hydroxide directly on the robot’s foot. A flexible platinum/carbon film acts as the reaction site for oxygen and water. The skin can be as thin as 135 micrometers and as small as 1 square millimeter. On zinc, it delivers a power density of 133 mW cm−2; on aluminum, 103 mW cm−2. These values are about an order of magnitude higher than those of typical microbatteries. The system also retains its current capability when scaled down to 0.0004 cm3. The design must satisfy several conflicting needs at once. It has to hold water, stay flexible, stick to surfaces, release easily, and sense the ground.
Holding water and surviving extreme temperatures
The membrane’s carboxylate groups, potassium ions, and hydroxide ions interact strongly with water molecules. Even at 20% relative humidity, similar to the Sahara Desert, it retains most of its water and continues to discharge after 48 hours. At −20°C, potassium ions disrupt the ordered hydrogen-bond network of water, keeping the membrane ionically conductive. The voltage drops by only about 0.2 V compared with room temperature. The crosslinked structure also allows operation at 80°C. After water evaporates, adding water restores performance.
Power on contact, off on lift-off
The membrane generates electricity when it touches an active substrate and stops when the foot lifts. The substrate loses electrons. On the platinum/carbon side, oxygen and water accept electrons to produce hydroxide ions. The reaction starts in about 0.1 milliseconds and stops when contact ends. Each walking step refreshes the reaction interface. Discharge products remain on the substrate rather than accumulating on the membrane, avoiding the performance decay seen in static batteries.
Sticky in the right direction
The membrane combines strong shear adhesion with easy normal detachment. Carboxylate groups create electrostatic adhesion to metal substrates. In the walking direction, adhesion is strong enough to prevent slipping. In the vertical direction, detachment resistance is low, so the robot can lift its foot efficiently. This allows the microrobot to crawl on inclined aluminum surfaces.
Sensing the ground through ions
The membrane can also sense the surface beneath it. Different materials cause different electrochemical reaction rates and hydroxide consumption, leading to directed migration of potassium ions and a characteristic potential signal. Rough surfaces deform the membrane and change ion distribution, producing another signal. Material changes generate stable square-wave signals, while roughness changes produce millivolt-level spikes. The two signals can be naturally decoupled. With an X-Y electrode array, the membrane can also locate where contact occurs. This helps the robot recognize its environment and actively seek energy-rich paths.
Satellite data could help governments decide where to build the data centres behind the growth of AI, by showing where renewable power, water and land are available and where drought and climate risks are highest, according to researchers at the University of Surrey.
Writing in Nature Reviews Clean Technology, the team argues that Earth observation (information about the planet's systems gathered largely by satellites), which is often underused, should become a formal part of how governments, regulators and investors plan, run and check the clean energy transition.
The call from Surrey researchers comes as demand floods the queue to connect to Britain's electricity grid. According to Ofgem, contracted demand in that queue rose from 41 gigawatts to 125 gigawatts between November 2024 and June 2025, driven largely by data centre projects.
Globally, the International Energy Agency expects electricity use by data centres to double from 485 terawatt-hours in 2025 to 950 terawatt-hours by 2030, roughly the amount of electricity Japan uses in a year.
The authors say geospatial and satellite data can support more strategic data centre planning by combining information on renewable energy potential, land-surface temperature, open water and groundwater availability, drought exposure, biodiversity, land-use constraints and climate-risk projections.
Dr Ana Andries, Senior Lecturer in GIS, Remote Sensing and Environmental Assessment and lead author of the Comment from the University of Surrey, said:
"Every new data centre is a long-term decision about power, water and land, while also an increasing source of global carbon emissions. Satellites already give us detailed and regularly updated information on all three, yet that evidence is too often left out of the decisions that matter. We're asking governments to write it into the rules, so the infrastructure behind the AI boom goes where the grid, renewable energy potential and the landscape can support it."
The Comment also sets out how satellites and related geospatial data support planning, operation and monitoring. For planning, the Global Solar Atlas uses satellite data to map solar power potential worldwide, while MapYourGrid, an open-source community project, is mapping power lines, substations and power plants so planners can see where new generation can connect. Once infrastructure is running, thermal imaging from space could help identify faults such as hot spots on solar panels, while radar from Sentinel-1 satellites can map wind at the sea surface around offshore wind farms, including the slower air in the wakes the farms leave behind.
Water is one area where energy systems are exposed. For instance, this summer's drought pushed the Danube to record lows and forced Romania to shut down both reactors at the Cernavodă nuclear plant, which normally supplies about a fifth of the country's electricity. Climate data and satellite monitoring of river levels, reservoirs and drought can help estimate risks to hydropower and to power stations that rely on river water for cooling, and help protect a dispatchable electricity supply. For accountability, some methane-sensing satellites detected releases as small as 0.03 tonnes per hour in single-blind controlled tests, though performance varied between systems. This gives regulators a way to check emissions from oil and gas infrastructure. Satellite imagery has also been used to map mining expansion linked to clean-energy supply chains in the Democratic Republic of Congo.
The authors call on governments and regulators to require satellite evidence in renewable energy zoning, infrastructure permitting, climate-risk screening and environmental impact assessment, especially where projects affect land, water, biodiversity or grid access. Energy plans and regulatory reporting should also set out clear data standards, validation requirements and how uncertainty is reported. For investors and utilities, they argue, satellite data should become part of due diligence and asset-risk assessment. Satellite data should support these decisions, they add, not replace ground data, local knowledge or official reporting.
Delivering this will need capacity building, the authors say – energy ministries, regulators, utilities, local authorities and investors need trusted satellite products, practical guidance, standards, training and evidence that the approach is cost-effective.
Professor Ravi Silva, Distinguished Professor and Director of the Advanced Technology Institute at the University of Surrey and co-author of the Comment, said:
"Building a clean energy future requires thousands of decisions, from choosing locations for solar farms, wind turbines, transmission infrastructure and energy storage to ensuring these assets operate effectively over their lifetime. Better evidence helps make these decisions more efficient, cost-effective and reliable. Satellite data can help identify areas with strong solar and wind resources, map existing infrastructure, detect operational issues, and provide independent information to support monitoring and oversight."
###
Notes to editors
Citation: Andries, A., Xuan, J., Liu, L. et al. Leverage Earth observation data to support the clean energy transition. Nat. Rev. Clean Technol. (2026). https://doi.org/10.1038/s44359-026-00220-y
Primary school pupils who use an AI-supported learning tool learn arithmetic more quickly than children who do not use such a tool. This is the finding of a multi-year study by Radboud University involving nearly eight thousand children. However, the researchers warn that it's no miracle cure, as the teacher’s role remains essential even with AI.
The researchers examined the development of numeracy skills among 7,885 Dutch primary school pupils over a three-year period, when these systems were first used between 2014 and 2017. They compared classes that used an adaptive learning system with classes that taught numeracy without such a system. Such a system automatically adapts exercises to the level of an individual pupil. Pupils are given more difficult tasks if they perform well and simpler tasks if they get stuck. On average, pupils who used the system showed more positive growth in their maths performance over the years.
Large-scale study
‘It is one of the first studies to examine the long-term effects of adaptive learning tools. Until now, research into these kinds of systems has mainly focused on short-term effects, or has been conducted on a smaller scale,’ explains Susanne de Mooij, an education researcher and one of the study’s authors.
‘Adaptive learning systems have been in use for almost fifteen years in more than half of all Dutch primary schools. Nevertheless, some fear that this type of adaptive learning tool leads to pupils learning less effectively, as digital learning materials are thought to be less effective than paper-based ones. This study shows that, in cognitive terms at least, this is not the case.’
Greater success at large schools
The effects found are modest, De Mooij emphasises. ‘We have found small effects, but generally a positive trend. Moreover, the positive effect is particularly evident in large schools and in vulnerable schools with many pupils from less advantaged socio-economic backgrounds. There is often greater variation between pupils in such settings. Adaptive technology can then help to better tailor teaching to individual needs,’ says De Mooij.
This means that an adaptive learning tool is primarily something that supports teachers, rather than replacing them. ‘A teacher cannot adapt the lesson content to thirty pupils at the same time. But with a tool like this, you can improve pupils’ maths skills whilst also reducing a teacher’s workload.’ The researchers emphasise that the teacher always remains in control: ‘The teacher draws up the lesson plan, decides what pupils learn and oversees what happens in the classroom. Such learning technology is fed solely by the data stored in the system, whereas a teacher has access to much more information.’
Researchers have created a closed-loop AI laboratory, capable of conducting research on brewer’s yeast. It could identify biological questions, recommend experiments and evaluate experimental outcomes. The image shows the robot scientist Eve, which was specifically designed for drug discovery, and which has now been updated with large language models and automated reasoning.
Credit: Credit: NIH Image Gallery/Chalmers University of Technology
Researchers at Chalmers University of Technology in Sweden have developed an AI scientist capable of generating scientific hypotheses, designing experiments, and interpreting results. Making new biological discoveries with minimal human intervention is an important advance in self-driving laboratories.
By combining advances in large language models, automated reasoning and laboratory automation, the researchers created a closed-loop AI laboratory, capable of conducting research on brewer’s yeast, Saccharomyces Cerevisiae. As the basis of the research, the AI was input scientific knowledge, including the yeast’s genome, metabolism and previous studies.
“It is too much information for a human to analyse, but our AI scientist could identify promising biological questions, recommend experiments to test them, evaluate experimental outcomes and iteratively refine its understanding based on new evidence. Rather than serving solely as decision supporting tools, the AI scientist actively generates new scientific knowledge”, says Ievgeniia Tiukova, postdoctoral researcher at the department of Life Sciences at Chalmers University of Technology, and one of the authors of the new study.
Integrating the thinking power of AI with an experimental capability is unusual, and cutting edge within a rapidly expanding area, where AI scientists autonomously perform extensive research. Tiukova compares the development to that of self-driving cars, where AI and machine learning are also used to process information, draw conclusions, and take action.
According to Ross King, Professor at the Department of Computer Science and Engineering at Chalmers and the University of Gothenburg, and the study’s senior author, autonomous laboratories will revolutionise research by systematically investigating biological systems much faster than is possible today.
“AI Scientists will collaborate with human scientists to accelerate discoveries across biology, medicine and biotechnology. Such AI systems have the potential to reduce the time required to explore complex scientific questions, and optimise the use of laboratory resources”, says Professor King.
Both authors emphasise that autonomous AI will, for now, augment rather than replace scientists, by increasingly undertaking the routine cycles of hypothesis generation and experimental testing.
“Human scientists remain essential for defining research priorities, interpreting broader scientific significance and ensuring ethical oversight. Future generations of autonomous discovery systems will become increasingly capable of collaborating with human scientists, becoming valuable partners in addressing some of the most challenging questions in biology and medicine”, says Professor King.
Captions: Top image: Researchers have created a closed-loop AI laboratory, capable of conducting research on brewer’s yeast. It could identify biological questions, recommend experiments and evaluate experimental outcomes. The images shows the robot scientist Eve, which was specifically designed for drug discovery, and which has now been updated with large language models and automated reasoning. Credit: NIH Image Gallery/Chalmers University of Technology
Second image: Ievgeniia Tiukova and the Robot scientist Eve
About AI scientists
Professor Ross King was the first to develop the concept of a general-purpose robot scientist. His first robot scientist, Adam, was designed to autonomously carry out scientific experiments and generate new knowledge. He later developed a second robot scientist, Eve, which was specifically designed for drug discovery.
The underlying concept of using robotic systems to automate and accelerate scientific discovery has since been extended to other areas of research, including chemistry and other specialised scientific tasks.
More about the research:
The study Agentic AI integrated with scientific knowledge: laboratory validation in systems biology was published in the Journal of the Royal Society Interface.
The authors of the study are Daniel Brunnsåker, Alexander H. Gower, Prajakta Naval, Erik Y. Bjurström, Filip Kronström, Ievgeniia A. Tiukova and Ross D. King. The researchers are affiliated with Chalmers University of Technology, the University of Gothenburg and the University of Cambridge in the UK.
The study has received funding from the Wallenberg AI, Autonomous Systems and Software Program (WASP), the UK Engineering and Physical Sciences Research Council, Chalmers AI Research Centre (CHAIR), and the Swedish Research Council for Sustainable Development, Formas.
The robot scientist Eve, which was specifically designed for drug discovery, has now been updated with large language models and automated reasoning.
Self-driven biological discovery through automated hypothesis generation and experimental validation
AI-driven analytics and data-mining strategies for drug discovery
New volume of SLAS Discovery highlights innovative screening platforms, AI-driven analytics and data-mining strategies for treating oncology, infectious disease, immunology and computational biology
SLAS (Society for Laboratory Automation and Screening)
Oak Brook, IL – Volume 42 of SLAS Discovery includes one review, eight original research articles and 1 short communication for oncology, infectious disease, immunology and computational biology research through innovative screening platforms for accelerating the discovery of targeted therapies, diagnostic biomarkers, and mechanistic insights across diverse disease areas.
Review
ADCs for Colorectal Carcinoma: Decoding Clinical Evidence for Molecular Design Innovation This review examines the rapidly evolving landscape of antibody-drug conjugates (ADCs) in colorectal cancer (CRC), highlighting the recent approval of T-DXd for HER2-positive disease as a milestone entry into this treatment arena. The authors explore clinical performance, design considerations and future directions, positioning ADCs as a promising targeted option for CRC patients with unmet clinical needs.
Original Research
Identification of AMPD2 Allosteric Inhibitors with Novel Mechanism of Action by Fragment Merging Approach Using an X-ray fragment screening approach, researchers identified a series of allosteric inhibitors that selectively target Adenosine monophosphate deaminase 2 (AMPD2) over other AMPD isozymes, overcoming the poor selectivity of traditional orthosteric inhibitors. Through iterative fragment merging and optimization, the team developed potent compounds (10g and 10h) that bind a previously uncharacterized allosteric site, offering valuable tool compounds for studying AMPD2's roles in nucleotide metabolism, energy homeostasis, and immune oncology.
Thermodynamic Profiling and Fragment Screening of GPCRS Using Grating-Coupled Interferometry Researchers demonstrate that grating-coupled interferometry (GCI) offers a powerful biosensor platform for characterizing GPCR–ligand interactions, providing high-quality kinetic data comparable to established Biacore technology while enabling rapid affinity and thermodynamic profiling from single-concentration injections. Using the adenosine A2Areceptor as a model, the team validated the approach through kinetic fragment screening of a 704-member library, identifying specific binders confirmed by nanoDSF and establishing GCI as an information-rich tool for early-stage GPCR drug discovery.
Identification of Novel Diagnostic Biomarkers and Host-Directed Drug Screening for Mycobacterium avium Infection: a Multi-Omics and Artificial Intelligence Study By integrating single-cell RNA sequencing with machine learning, researchers mapped the immune landscape of Mycobacterium avium (MAV) infection. They uncovered a monocyte-driven MIF-APP signaling axis that recruits and "locks" macrophages into a hyper-inflammatory state, explaining the paradox of high inflammation but poor pathogen clearance. The study also developed a five-gene diagnostic signature (AUC > 0.88) and identified Wogonin as a potential host-directed therapeutic that targets STAT3 and TNF to break the immune impasse.
Development of a Rapid and Sensitive EnLIGHT OMEGA Assay for Extracellular AGR2 Detection In Biological Fluids Researchers developed a novel homogeneous OMEGA assay to detect extracellular AGR2 (eAGR2), a protein strongly linked to tumor progression and cancer aggressiveness, with a broad dynamic range (0.02–300 ng/mL) and no washing or separation steps required. This sensitive, rapid method outperforms conventional ELISA and is applicable across diverse biological fluids, offering a valuable tool for cancer biology, drug resistance and biomarker discovery studies.
Use t Ttests to Analyze Counts of Cells in Two States A comparative analysis of statistical tests for count data found that t tests perform as well as or better than specialized count-based methods at maintaining false-positive rates, with no disadvantage in detecting real differences. The findings reassure researchers that converting count data to percentages and analyzing with t tests is a valid and effective approach under typical wet-lab conditions.
In Silico Prioritization and Cheminformatics Identify Structurally Diverse Small-Molecule Inhibitors Using a computational data-mining strategy applied to a previously deposited screen of nearly 300,000 small molecules, researchers identified three distinct chemical scaffolds that inhibit Lassa virus cell entry with potencies as low as 10 nM and strong selectivity over related viruses. These compounds act at the membrane fusion stage by targeting pH-sensitive regions of the viral glycoprotein, highlighting the power of combined computational and experimental approaches for antiviral drug discovery.
CellVision: A Deep Learning Based Image Analysis Platform to Accelerate Immuno-Plaque Assay Data Processing for Dengue Vaccine Development Researchers have developed CellVision, a deep learning-based workflow that fully automates viral plaque counting in high-throughput immuno-plaque assays, accurately segmenting fused plaques and distinguishing them from other objects without any manual image review. Integrated into Merck’s (& Co., Inc.) µPlaque assay to support an investigational dengue vaccine, CellVision outperformed a commercial alternative and sets a new standard for AI-powered analysis in antiviral vaccine discovery.
Tumor-Versus-Nonmalignant Quantitative Drug Sensitivity Profiling Identifies capivasertib as a Selective Therapeutic Candidate for Nasopharyngeal Carcinoma Using a tumor-normal-paired high-throughput drug screening approach against EBV-positive and EBV-negative nasopharyngeal carcinoma (NPC) cell lines, researchers identified the AKT inhibitor capivasertib as a highly selective anti-NPC agent that spares normal epithelial cells. Capivasertib synergized with platinum-based chemotherapy, enhanced radiosensitivity, and significantly prolonged survival in combination with cisplatin in xenograft models, providing a strong rationale for its clinical evaluation in advanced NPC.
Short Communication
Development of a High-Throughput TR-FRET Assay for Identification of Small Molecule Inhibitors of the LILRB4 (ILT3)-SCG2 Immune Checkpoint Interaction Researchers have developed a high-throughput TR-FRET assay to interrogate the interaction between the immune checkpoint LILRB4 (ILT3) and its ligand SCG2, a pathway driving myeloid-mediated immunosuppression in the tumor microenvironment. Pilot screening identified two compounds, BMS-813,160 and PSB-603, that dose-dependently inhibit this interaction with micromolar potency, providing the first small-molecule modulators of the LILRB4-SCG2 axis and a foundation for targeting myeloid-driven immunosuppression.
SLAS Discovery reports how scientists develop and use novel technologies and/or approaches to provide and characterize computational, chemical and biological tools to understand and treat human disease. The journal focuses on drug discovery sciences with a strong record of scientific rigor and impact, reporting on research that:
Enables and improves target validation
Shares and/or compare novel methods for medium- or high-throughput screening
Evaluates current drug discovery technologies
Provides novel research tools
Incorporates research approaches that enhance depth of knowledge and drug discovery success
SLAS (Society for Laboratory Automation and Screening) is an international professional society of academic, industry and government life sciences researchers and the developers and providers of laboratory automation technology. The SLAS mission is to bring together researchers in academia, industry and government to advance life sciences discovery and technology via education, knowledge exchange and global community building.
SLAS Discovery: Advancing the Science of Drug Discovery, 2025 Impact Factor 3.3. Editor-in-Chief Robert M. Campbell, PhD, Grove Biopharma, Inc., Chicago, IL (USA)
###
Journal
SLAS DISCOVERY
Motion Style Slider: novel framework enabling fine control of nuances in character motion
Directors and animators can now modulate the degree of emotion, energy, and expression without intermediate motion-capture data
Novel framework enables fine control over the intensity and expressiveness of AI-generated human motion, as reported by researchers from Science Tokyo. The “Motion Style Slider” framework enables animators to control the style and adjust a virtual character’s motion, from subtle to overly exaggerated, based on only two motion samples. The proposed approach achieves consistent style control and smooth transitions without the need for extensive motion-capture datasets.
In the production of videogames and animated movies, directors and animators are constantly fine-tuning how a character’s movements should appear. In many cases, the directors require subtle changes, such as making a character appear a little more restrained or a bit more expressive, and subjectively communicate the degree or extent to which the change needs to be applied. Commercially available artificial intelligence (AI) tools can generate stylized animations for virtual characters, transforming a neutral motion into one that conveys particular styles like joy, anger, exhaustion, or confidence. However, a more refined modulation demands more than simply applying a style; it requires precisely controlling how much of that style comes through.
Most existing tools fall short in this regard, offering only fixed versions of a given emotion or expression with no way to increase or reduce the intensity. The main reason behind this is that AI-based motion-generation systems are often trained using fixed examples of the same action, namely a neutral motion and fully stylized versions obtained through motion-capture technology. Creating additional recordings at multiple intermediate levels of expression would require multiple long motion-capture sessions, making it costly and impractical for real-world production.
To address this problem, a research team comprising Specially Appointed Assistant Professor Chen-Chieh Liao and Professor Hideki Koike of the School of Computing, Institute of Science Tokyo (Science Tokyo), Japan, together with researchers from Cygames, Inc., Japan, has developed the Motion Style Slider, a new AI framework for continuous control of the motion styles intensity in generated human animation. Their work was presented at the European Conference on Computer Vision 2026 (ECCV 2026) held on September 10, 2026.
The proposed framework needs to learn from only two endpoints: a neutral motion and a stylized version of the same action, such as ‘neutral walking’ and ‘happy walking.’ The style intensity ‘α,’ with values such as 0, 0.5, 1.0, and 1.5, was used to represent low, medium, and high style intensities of motion. “Given these endpoints, the model generates a continuous family of motions controlled by a user-defined scalar value, ranging from neutral behavior through the target style and into stronger reactions that lie beyond the input range. This type of control matches the way artists and directors naturally ask for a motion to be 'a little more' or 'a little less' expressive,” explains Liao. Figure 1 illustrates how changing the slider value α continuously adjusts the strength of the motion style.
The researchers implemented their innovative approach within a diffusion-based motion generation framework. The system learns the ‘direction’ in a learned motion-style embedding space along which motion changes between the neutral and stylized endpoints, without relying on fixed style labels. As shown in Figure 2, information representing this style-change direction is combined with information about the motion content and the user-specified slider value before being fed into a pretrained motion-generation model.
The team evaluated the method using multiple benchmark motion datasets and an additional dataset specifically created to assess performance beyond the styles seen during training. The method was compared against established techniques like Multi-condition Motion Latent Diffusion Model and DeepMotionEditing (Aberman et al.). “Compared with existing methods, the proposed strategy achieved favorable results in the consistent control of style intensity, smooth transitions, and extrapolation to highly expressive motion,” remarks Liao. Representative examples comparing the proposed method with existing approaches are shown in Figure 3.
To further validate the framework’s output, the researchers conducted a Likert-scale user study with 11 university participants who rated the differences in style intensity among motions created by the Motion Style Slider as clearly distinguishable, while their naturalness remained comparable.
The Motion Style Slider provides an intuitive and practical framework that can consistently modulate motion style and intensity, transitions, and nuances of AI-generated human motion. This technology could advance animation workflows in games, films, and other forms of virtual content, letting directors and designers refine character performances according to their exact creative vision.
***
About Institute of Science Tokyo (Science Tokyo)
Institute of Science Tokyo (Science Tokyo) was established on October 1, 2024, following the merger between Tokyo Medical and Dental University (TMDU) and Tokyo Institute of Technology (Tokyo Tech), with the mission of “Advancing science and human wellbeing to create value for and with society.”
Motion Style Slider: Endpoint-Supervised Continuous Style Control for Human Motion Diffusion
The model derives the direction of style change from the stylized and neutral motions, then feeds style information corresponding to the user-specified slider value α, together with motion-content information, into a pretrained motion-generation model.
Examples generated by the proposed method (OURS) and existing methods (ABERMAN, MCM-LDM) as style intensity is varied for the same motion content. The proposed method changes the motion progressively in accordance with the specified style intensity.
Credit
Chen-Chieh Liao
Wearable AI forecasts prolonged sitting in women with chronic pelvic pain
Personalized models could help time prompts to move
The Mount Sinai Hospital / Mount Sinai School of Medicine
[New York, NY] September 30,2026—Researchers at the Icahn School of Medicine at Mount Sinai have developed an artificial intelligence (AI) approach using data from wearable devices that can forecast upcoming periods of prolonged sitting in women with chronic pelvic pain disorders.
The findings, published in the September 30 online issue of npj Women’s Health[DOI: 10.1038/s44294-026-00156-5], could inform personalized digital health tools that prompt the individual to move, such as by taking a short walk, at the right time while minimizing unnecessary alerts.
Chronic pelvic pain affects an estimated one in seven women and frequently occurs in people with conditions such as endometriosis, adenomyosis, and uterine fibroids. They are often associated with prolonged sitting because of pain, fatigue, and other symptoms that impact daily life. While regular movement can help manage symptoms and improve overall health, generic advice to "sit less and move more" often fails to account for the realities of living with these conditions.
The study shows that wearable devices may do more than count steps. Using data collected over time from wearables of women with chronic pelvic pain, the researchers developed a forecasting model that can identify when prolonged sedentary periods are likely to occur during waking hours. This could allow a future digital intervention to deliver a brief reminder to stand up or take a short walk before prolonged inactivity begins.
"Our goal was to determine whether everyday wearable devices could serve as an early-warning system for prolonged sitting in women living with chronic pelvic pain," says senior author Ipek Ensari, PhD, Assistant Professor of Artificial Intelligence and Human Health at the Icahn School of Medicine and a member of the Hasso Plattner Institute of Digital Health at Mount Sinai. "Rather than offering generic advice after the fact, we wanted to determine whether we could anticipate these moments and support people with simple, well-timed prompts that fit naturally into their daily lives."
The research team analyzed wearable data from 134 women with chronic pelvic pain disorders, primarily endometriosis, along with 61 healthy participants as a comparison group. Participants wore Fitbit devices for up to 90 days, generating minute-by-minute information about physical activity, heart rate, and sleep.
Using approximately 10 days of each participant’s data, the team trained personalized forecasting models to predict activity levels one hour ahead. They then tested whether those forecasts could identify 15-minute periods of sedentary behavior during waking hours, a timeframe that could allow a brief movement break, which the researchers called an “exercise snack.”
The work challenged the assumption that health AI must be increasingly complex. Relatively simple, interpretable models forecasted prolonged sitting as accurately as more computationally intensive deep-learning approaches evaluated in the study.
"We were surprised by how well the simplest models performed," says lead author Jannes Jegminat, PhD, a former postdoctoral research fellow at the Icahn School of Medicine at Mount Sinai. "More complex AI is not always better. Lightweight, interpretable models can accurately forecast sedentary behavior while being practical enough to run directly on a person's own device, which also helps protect privacy."
Making future tools more feasible to run directly on a person’s phone or wearable could reduce computational demands and the need to transmit sensitive data to remote servers, say the investigators.
The models also remained robust even in instances of incomplete data, which can happen when participants remove their devices or forget to synchronize them. This suggests that everyday data from wearables may support meaningful predictions under real-world conditions rather than only in controlled laboratory settings.
"This study suggests that predicting prolonged sitting is feasible, even if the individual has chronic conditions that might impact their daily routine," says Dr. Ensari. "The next step is determining whether delivering personalized movement prompts based on those predictions actually helps reduce sedentary time, improves symptoms, and enhances quality of life. Those questions will require prospective clinical trials."
The researchers believe the work could ultimately support digital health tools that feel less like constant reminders and more like a personalized coach, delivering only a small number of meaningful prompts each day while minimizing unnecessary notifications that often lead users to ignore health apps.
Beyond chronic pelvic pain, the approach may also apply to other chronic conditions in which prolonged sitting contributes to poorer health outcomes.
The research team is now working to incorporate the forecasting framework into a just-in-time adaptive intervention, which will test whether personalized, AI-guided movement prompts can reduce sedentary time and improve symptoms among women living with chronic pelvic pain disorders.
The paper is titled "Robust forecasting of sedentary bouts in chronic pelvic pain disorders for on-device learning and real-time deployment."
The authors, as listed in the journal, are Jannes Jegminat, Samia Shahnawaz, Jovita Rodrigues, Matteo Danieletto, Kyle Landell, Gabriele Campanella, Carol Ewing Garber, Zahi A. Fayad, and Ipek Ensari.
The work was funded in part by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under Award Number R01HD108263, and by the Clinical and Translational Science Awards (CTSA) grant UL1TR004419 from the National Center for Advancing Translational Sciences.
About Mount Sinai's WindreichDepartment of Artificial Intelligence and Human Health
Led by Girish N. Nadkarni, MD, MPH, Chair of the Windreich Department of Artificial Intelligence and Human Health, Icahn School of Medicine at Mount Sinai—an international authority on the safe, effective, and ethical use of AI in health care—Mount Sinai’s Windreich Department of Artificial Intelligence and Human Health is the firstof its kind at a U.S. medical school, pioneering transformative advancements at the intersection of artificial intelligence and human health.
The Department is committed to leveraging AI in a responsible, effective, ethical, and safe manner to transform research, clinical care, education, and operations. By bringing together world-class AI expertise, cutting-edge infrastructure, and unparalleled computational power, the department is advancing breakthroughs in multi-scale, multimodal data integration while streamlining pathways for rapid testing and translation into practice.
The Department benefits from dynamic collaborations across Mount Sinai, including with the Hasso Plattner Institute for Digital Health at Mount Sinai—a partnership between the Hasso Plattner Institute for Digital Engineering in Potsdam, Germany, and the Mount Sinai Health System—which complements its mission by advancing data-driven approaches to improve patient care and health outcomes.
At the heart of this innovation is the renowned Icahn School of Medicine at Mount Sinai, which serves as a central hub for learning and collaboration. This unique integration enables dynamic partnerships across institutes, academic departments, hospitals, and outpatient centers, driving progress in disease prevention, improving treatments for complex illnesses, and elevating quality of life on a global scale.
In 2024, the Department's innovative NutriScan AI application, developed by the Mount Sinai Health System Clinical Data Science team in partnership with Department faculty, earned Mount Sinai Health System the prestigious Hearst Health Prize. NutriScan is designed to facilitate faster identification and treatment of malnutrition in hospitalized patients. This machine learning tool improves malnutrition diagnosis rates and resource utilization, demonstrating the impactful application of AI in health care.
For more information on Mount Sinai's Windreich Department of Artificial Intelligence and Human Health, visit: ai.mssm.edu
About the Hasso Plattner Institute at Mount Sinai
At the Hasso Plattner Institute for Digital Health at Mount Sinai, the tools of data science, biomedical and digital engineering, and medical expertise are used to improve and extend lives. The Institute represents a collaboration between the Hasso Plattner Institute for Digital Engineering in Potsdam, Germany, and the Mount Sinai Health System.
Under the leadership of Girish N. Nadkarni, MD, MPH, Director of the Hasso Plattner Institute for Digital Health at Mount Sinai, and Professor Lothar Wieler, a globally recognized expert in public health and digital transformation, they jointly oversee the partnership, driving innovations that positively impact patient lives while transforming how people think about personal health and health systems.
The Hasso Plattner Institute for Digital Health at Mount Sinai receives generous support from the Hasso Plattner Foundation. Current research programs and machine learning efforts focus on improving the ability to diagnose and treat patients.
About the Icahn School of Medicine at Mount Sinai
The Icahn School of Medicine at Mount Sinai is internationally renowned for its outstanding research, educational, and clinical care programs. It is the sole academic partner for the seven member hospitals* of the Mount Sinai Health System, one of the largest academic health systems in the United States, providing care to New York City’s large and diverse patient population.
The Icahn School of Medicine at Mount Sinai offers highly competitive MD, PhD, MD-PhD, and master’s degree programs, with enrollment of more than 1,200 students. It has the largest graduate medical education program in the country, with more than 2,700 clinical residents and fellows training throughout the Health System. The Graduate School of Biomedical Sciences offers 13 degree-granting programs, conducts innovative basic and translational research, and trains more than 470 postdoctoral research fellows.
Ranked 11th nationwide in National Institutes of Health (NIH) funding, the Icahn School of Medicine at Mount Sinai is among the 90th percentile of U.S. private medical schools in Sponsored Programs Direct Expenditures per Principal Investigator, according to the Association of American Medical Colleges. More than 6,900 scientists, educators, and clinicianswork across dozens of academic departmentsand multidisciplinary institutes with an emphasis on translational research and therapeutics. Through Mount Sinai Innovation Partners (MSIP), the Health System facilitates the real-world application and commercialization of medical breakthroughs made at Mount Sinai.
* Mount Sinai Health System member hospitals: The Mount Sinai Hospital; Mount Sinai Brooklyn; Mount Sinai Morningside; Mount Sinai Queens; Mount Sinai South Nassau; Mount Sinai West; and New York Eye and Ear Infirmary of Mount Sinai.
