<strong>Background:</strong> China faces a high prevalence of mental disorders but low treatment uptake, a gap driven by limited awareness and unevenly distributed mental health resources. While online psychotherapy has the potential to expand access, patient willingness remains insufficiently explored. <strong>Objective:</strong> This study aimed to investigate the willingness of Chinese patients with mental disorders to engage in online psychotherapy and to identify associated factors. <strong>Methods:</strong> A multicenter, cross-sectional survey was conducted using a structured questionnaire to assess the attitudes and willingness of patients with mental disorders in China to engage in online psychotherapy. Willingness to engage in online psychotherapy was assessed using a 0 to 100 rating scale, with higher scores indicating greater willingness. Univariate analysis, correlation analysis, and multivariate linear regression analyses were used to identify factors influencing willingness. <strong>Results:</strong> Among 361 eligible participants, the mean willingness score for online psychotherapy was 70 (SD 28.56). In total, 86.4% (n=312) of participants preferred short-term therapy (1 to 10 sessions), while 92.5% (n=334) expected the cost per session to remain less than CNY ¥400 (US $55.50). Participants most preferred therapist-guided online individual therapy (n=142, 39.3%). Convenience (124/361, 34.3%) and perceived anonymity (“no one will know about the illness”; 119/361, 33.0%) were the 2 most commonly reported perceived benefits of online psychotherapy. The leading barrier was concerns about data security and privacy (108/303, 35.6%), followed by difficulty in establishing therapeutic rapport (60/303, 19.8%). The regression analysis revealed that age, self-stigma, satisfaction with current psychiatric medications, and satisfaction with previous online psychotherapy significantly influenced patients’ willingness to seek online psychotherapy. <strong>Conclusions:</strong> This multicenter study reveals a high level of willingness to engage in online psychotherapy among Chinese patients, with self-stigma as a key barrier. These findings support the development of tailored services, stigma reduction interventions, and infrastructure investment to enhance mental health care delivery.
As cancer care becomes data-driven, artificial intelligence (AI) will play an increasingly central role across the treatment continuum, from biomarker identification and drug development to clinical trial recruitment and diagnostics. In this corner of healthcare, the ability of AI to interpret and annotate tumor sample slides that have been digitized is taking center stage. While the promise is great, and AI interpretation is already influencing some clinical care, it has not yet reached critical mass.
“There’s something like a billion slides created every year for diagnostic purposes, and today most of those, about 85%, are still read by a pathologist with a microscope on physical glass slides,” said David West, CEO and co-founder of digital pathology company Proscia. In practice, that means pathologists manually examine slides, identify cancer, grade tumors, and dictate reports in a traditional approach to diagnosing cancer that has seen little change in decades.
Mohamed Omar, MD Associate Professor Cedars-Sinai Medical Center
But that foundation is now shifting. Advances in slide scanning, cloud storage, and AI are turning digital pathology images into data that can be analyzed at scale. At Memorial Sloan Kettering Cancer Center, large archives of digitized slides helped launch Paige AI, one of the earliest companies to train deep learning systems on pathology images linked to clinical and genomic outcomes. This yielded the first U.S. Food and Drug Administration (FDA)-approved diagnostic using AI and digital pathology: Paige Prostate Detect. The company, which was acquired last year by AI-enabled precision medicine company Tempus, now combines Paige’s digital pathology-based AI with Tempus’s broad genomic sequencing data platform.
Researchers in the field say the implications of AI in digital pathology extend beyond image analysis. Mohamed Omar, MD, an associate professor of computational biology at Cedars-Sinai Medical Center, Los Angeles, noted that large language models can help clinicians navigate a research landscape that produces “hundreds of papers every single day” to inform ongoing cancer research. Multimodal AI tools promise to unlock even more insights from digital pathology data by combining it with genomic, radiomic, and clinical data to build powerful new models of both common and rare cancers for diagnosis, drug development, and clinical trial enrollment.
Razik Yousfi SVP and GM, Tempus
While adoption is in its early stages, the advent of faster and less expensive scanners is bringing digital pathology within reach of both regional and rural hospitals. Razik Yousfi, senior vice president and general manager of AI products at Tempus, and a co-founder of Paige, predicts that within the next 10 years, the majority of pathology workflows will be digital. The ultimate goal of the application of AI here is not to replace human pathologists, but to empower them with a capable assistant while spreading adoption beyond major medical centers.
Building the foundations
As the field of applying AI to digital pathology progresses, it needs to build the groundwork for a wider range of potential applications that could address rare cancers and other areas without an abundance of data. One such project is called Atlas, a collaboration between researchers in Korea, Germany, and the United States to build a foundation model trained using 1.2 million histopathology whole-slide images from 490,000 cases sourced from the Mayo Clinic and Charité – Universitätsmedizin Berlin.
Foundation models like Atlas allow large-scale pre-training of data to develop numerical representations called embeddings that capture both the structural and contextual features of slides in the dataset. Atlas incorporates a diversity of diseases, staining types, and scanners, and uses multiple image magnifications during training. This broad approach confers power and utility. It allows the digitized representations of the histology to be adapted, queried, or fine-tuned to very specific downstream tasks using much less data than would be needed to build a one-off model.
As such, a foundation model provides a reusable digitized computational backbone that can be tapped across a wide range of uses, like tumor classification, detection of morphologic structures, biomarker quantification, and outcome prediction. In short, foundational models make the process of querying digital pathology images more efficient compared with past approaches.
Andrew P. Norgan, MD, PhD CMO, Mayo Clinic
“In the case of pathology, the successful AI models developed using ‘conventional’ neural network approaches before the advent of FMs (foundation models) typically required huge amounts of training data to achieve high performance and generalizability—the ability to work across datasets distinct from the training data,” said lead Atlas researcher Andrew P. Norgan, MD, PhD, CMO of Mayo Clinic Digital Pathology and assistant professor of laboratory medicine and pathology. “We think of FMs as [an] enabler that allows model development in pathology … to move from artisanal or craft processes to more scalable and reproducible processes that should allow for the rapid development of high-quality models to address problems in pathology.”
At Paige AI, the company’s early work resulted in the first FDA-approved AI diagnostic, Paige Prostate Detect. Its algorithm was built using a technique called multiple instance learning instead of traditional supervised neural network techniques that require detailed human annotation of slides, a time-consuming and expensive method that could expose the learning to human error. The difference between the two methods is that traditional neural networks expose AI to a slide with cancer and tell it that there is cancer present. In multiple instance learning, the model is shown unannotated slides and is tasked with finding the cancer.
Even this approach, however, required a very large dataset. It became apparent to company leaders that the heavy lifting required to get Paige Prostate Detect to work wasn’t scalable.
“We had kind of cracked this recipe,” said Yousfi. “We know how to use a lot of GPU (graphics processing unit) compute, and if we get a ton of data and a lot of compute, we can build anything. But GPU infrastructure is very expensive, and it takes a lot of time to train a very large system.”
Perhaps the most important factor moving Paige away from this model is that it will not work when there is only a small amount of data available. This blocks the ability to train AI to recognize rare cancers for which sample counts are low. The company needed a different approach.
“We had this idea [for] a new system that was basically trained on all of the images we had access to, independent of the organ and indication and tissue and task,” Yousfi said. “Back then, we didn’t know what that thing was called. But ultimately, that became what everyone is calling today a foundation model.”
Originally trained on 200,000 slides, Paige’s new model now includes 3.5 million images and roughly two billion parameters, making it the backbone for other downstream applications the company builds today. This ability to use foundation models as the AI and data encyclopedia for smaller applications will ultimately propel the field of digital pathology forward by widening the playing field.
Going multimodal
To address more complex predictive problems, additional data types can be integrated. Clinical, radiologic, or genomic data can be combined with morphologic embeddings or used during training to help the model learn which tissue features carry a signal of disease or identify a biomarker. These approaches aim to support precision oncology by making morphologic data computable and aligning slide-derived features with other cancer-focused datasets. “These approaches can surface subtle or ‘latent’ patterns in pathology slides and align them with other data sources,” Norgan said. Pathologist and oncology care teams can then evaluate and interpret the features identified by the models within the clinical and biological context.
“In this way, pathologists and oncology teams use these outputs as decision-support tools, while clinical judgment remains central to diagnostic interpretation and therapeutic decision making,” Norgan added.
Atlas has now been succeeded by Atlas2, which was trained on 5.5 million pathology images and is now a two billion-parameter model, making it one of the largest pathology foundation models to date. The team has explored distilling methods to create smaller, more efficient, and targeted versions of the model that retain performance, with an eye toward finding a balance between scale and deployability.
Proscia is embarking on a different multimodal approach that combines vision models with language models, with the intent of creating methods to query the morphology of digitized slides. Their efforts in vision-language models (VLMs) combine textual data with visual data and allow the model to describe the morphology of a slide, answer questions about what it contains, find images in a database based on a text query, and even follow multimodal instructions such as “circle the tumor area on this image.”
In short, a VLM can be engaged in the same way you can engage a human. “I could go ask a pathologist to point out all the areas of tumor-infiltrating lymphocytes,” West said. “Now, because language-vision models are encoding language and images in the same space, they can do that, too. You can ask the model to describe what is happening in an image, and it will tell you exactly what it sees.”
At Cedars-Sinai, Omar’s work with large language models takes a less direct route of leveraging queries to gather information from research studies or even images. “Basically, you could go to the tool, ask questions, and the tool will provide you with pieces of code,” he explained. “These pieces of code are what you use on the slide to get more information.”
Atlas provides a similar function at the Mayo Clinic, Norgan noted. Because the model-generated embeddings in the digitized slide also encode semantic information, the Atlas team is now building a slide search function, which would allow researchers or clinicians to identify and access slides, or regions of slides, with related features.
Democratizing care
Although it will take time to disseminate the tools needed for AI-enabled digitized models of cancer care to smaller health systems, the future is now at Moffitt Cancer Center, where the research hospital is engaged in a top-to-bottom digitization of its system.
Marilyn Bui, MD, PhD Senior Member Moffitt Cancer Center
According to Marilyn Bui, MD, PhD, senior member of the departments of pathology and machine learning, the comprehensive cancer center plans for full digital adoption across clinical and research labs by 2027. Last August, it entered a multi-year collaboration with integrated AI and digital pathology company PathAI to deploy its cloud-based digital pathology image management system for both research and clinical applications.
Within the pathology department, the transition will mean that all glass slides will be scanned and reviewed digitally, providing the basis for applying AI computational tools to assist pathologists. Bui said that the cancer center is accelerating its move toward clinical AI adoption: “Just today I received an email asking which AI algorithms we plan to incorporate for clinical utility—prostate cancer, breast cancer, general tumor detection,” she said. “For us, it’s no longer just research.”
Moffitt is taking a hybrid approach to algorithm development and deployment within the system. Some AI tools will come from commercial vendors and will be validated internally, while others will be developed by investigators through the center’s translational pathology work. Taking this approach will allow it to apply AI to both common cancers and the rare tumor types Moffitt frequently encounters.
While the digital initiative will be transformational, Bui emphasized that the goal is not to replace pathologists but to enhance their capabilities. She prefers to refer to AI as augmented intelligence to reflect this. “Artificial intelligence suggests a robot replacing us,” she said. “But what we mean is augmented intelligence—tools that assist and enhance our ability to make clinical decisions.”
Further, Moffitt intends to integrate digitized slide data with genomic, proteomic, and clinical outcome data to build a multimodal data environment that could advance precision oncology. “Digital pathology and AI will allow us to extract far more information from tissue samples,” Bui said, “making our diagnoses more actionable for the clinical team and ultimately improving patient care.”
The promise of AI in oncology isn’t just better algorithms, it’s broader access. The maturation of computational pathology and its dissemination from large cancer centers like Moffitt to regional and rural health systems has the potential to provide levels of care typically only available at large research hospitals in community settings as well.
“It’s about democratizing access to care,” said Omar. “For a person in Maine or Wisconsin or another place to have access to the same high-quality care that you would get from a larger academic medical center in LA or New York, slides have to be digitized.”
Over the next 10 years, there could be a compelling business case for hospitals to embrace digital pathology. As the cost of scanners comes down and a broad range of diagnostic tools becomes available, digitizing routine H&E slides could become common.
While genetic cancer testing can cost hundreds of dollars, Omar pointed out that pathology slides “cost $5 [and] they are available universally, in all patients with cancer.” As AI models increasingly identify genomic-level insights directly from those inexpensive images, it represents a “huge win for accessibility, making AI work for patients who cannot afford genetic tests,” Omar said. If there is broad adoption of digital pathology “it is very easy to roll out any kind of AI models and computational tools across the board, across situations and locations that don’t have access to care.”
“At the end of the day, all slides will be digitized,” he concluded. “It’s just a matter of time.”
Chris Anderson, a Maine native, has been a B2B editor for more than 25 years. He was the founding editor of Security Systems News and Drug Discovery News, and led the print launch and expanded coverage as editor in chief of Clinical OMICs, now named Inside Precision Medicine.
Background: Digital mental health technologies (DMHTs) are playing an increasing role in mental health services. The quality of evidence for DMHTs is variable, and there are concerns that evidence is not sufficient to support decision-making. Objective: This study used a cross-sectional analysis of evidence supporting DMHTs included in National Institute for Health and Care Excellence (NICE) evaluations to examine the strength of evidence available for decision-making. Methods: We identified all NICE evaluations relating to DMHTs by reviewing details of published NICE evaluations on the NICE website. From each of these evaluations, we identified included DMHTs and reviewed committee documentation to identify studies that provided supporting evidence for each of these technologies. We extracted information on a series of items relating to study quality and summarized the characteristics of evidence both at the level of individual studies and across the package of evidence from multiple studies supporting DMHTs. We also identified key evidence gaps in available evidence. Results: We included nine NICE evaluations relating to anxiety, depression, psychosis, insomnia, attention deficit hyperactivity disorder (ADHD), and tic disorders. These evaluations included 30 DMHTs and referenced 78 supporting studies. We identified common evidence gaps relating to effectiveness compared to relevant comparators, use of appropriate outcomes, including health-related quality of life, cost of delivery, and impact on resource use, and reporting of adverse events. Conclusions: Our study highlights that some DMHTs have been supported by high-quality studies and that evidence to support DMHTs is likely to be developed across a series of studies. However, there are often key evidence gaps that need to be addressed to provide a stronger case for adoption. Developers should ensure that they consider these gaps while planning evidence generation, and where possible, address them earlier in the product lifecycle.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/825f13db8cbad54213afa5c433d7adde" />
Researchers at the Institute for Biomedical Sciences at Georgia State University have developed a vaccine platform designed to provide protection against a range of influenza virus infections by targeting conserved viral structures and inducing immunity at mucosal surfaces. The study, published in ACS Nano, details how the development of the novel vaccine uses cell-derived extracellular vesicles (EVs) engineered to display multiple influenza hemagglutinins (HAs) in an inverted configuration, which allows the immune system to recognize conserved regions shared across viral strains. In mouse models, the vaccine elicited cross-reactive antibodies, cellular immune responses, and mucosal immunity, providing protection against heterosubtypic H5N1 and H7N9 influenza virus challenges following intranasal administration.
“The influenza virus is smart. They have evolved to evade the immune system by hiding their critical conserved structures, rendering these elements poorly immunogenic,” said senior author Bao-Zhong Wang, PhD, a professor at the Institute for Biomedical Sciences at Georgia State.
The vaccine’s design centers on two key features: the use of extracellular vesicles (EVs) as a delivery platform and the inversion of HA proteins on their surface. EVs are natural nanoparticles involved in cell-to-cell communication and have been researched extensively for therapeutic delivery due to their biocompatibility. In this study, they were engineered to display multiple HA subtypes simultaneously. The HAs were presented in an upside-down orientation, which partially shields the highly variable head domain while exposing the conserved stalk domain. This structural arrangement directs the immune response toward regions less prone to mutation, enabling broader protection across influenza strains.
“These (vaccine responses) highlight that the inverted HA is a smarter strategy for inducing protective immunity to the conserved HA stalk. Meanwhile, cell-origin EVs are a biocompatible platform for mucosal vaccine delivery. Using EVs simultaneously displaying multiple inverted HAs is a powerful approach for developing universal influenza vaccines,” Wang added.
To test their vaccine design the researchers immunized mice intranasally with the EV-based vaccine, allowing researchers to assess mucosal immunity in addition to systemic responses. The data showed that the vaccine induced cross-reactive antibodies targeting HA stalks, virus-specific T cell responses, and a balanced Th1/Th2 immune profile. Importantly, the vaccinated mice were fully protected against lethal infections from reassortant H5N1 and H7N9 viruses.
The new vaccine designed was based on previous research into EV-based vaccine delivery and different strategies to target the HA stalk. Earlier studies had shown that EVs could serve as adjuvants and antigen carriers for intranasal vaccines. In addition, other research seeking to develop a universal influenza vaccine has focused on the conserved HA stalk domain, which evolves more slowly than the immunodominant head. As the researchers noted, “the conserved HA stalk domain has emerged as a promising candidate for a universal influenza vaccine due to its low evolutionary rate and greater tolerance to mutations.” However, previous approaches had used isolated HA stalk constructs, but faced shortcoming related to structural stability and immunogenicity.
By preserving the full HA ectodomain while inverting its orientation, the new design addressed these limitations. “Our findings suggest that utilizing the entire HA ectodomain as an immunogen, while hiding the HA head and increasing exposure of the HA stalk, is an effective strategy to induce robust immune responses targeting conserved HA epitopes,” the researchers wrote, noting that this approach allows the immune system to access structurally intact conserved regions.
If this vaccine design can be shown effective in humans, it could provide a vaccine with broader and longer-lasting protection against influenza, and reduce the need for frequent reformulations. The use of intranasal delivery could also change how vaccines are administered by targeting immune responses at the site of viral entry. “Mucosal vaccination effectively induces local immune responses, protecting against respiratory virus infections at the site of invasion,” the team noted.
Next steps for the team include continued characterization of the immune responses induced by the vaccine, to include the specificity and neutralizing capacity of antibodies, as well as evaluation of anti-EV immunity with repeated dosing. More animal studies, and eventually clinical trials, will be needed to assess safety, scalability, and efficacy in humans.
Background: The universal adoption of mobile technologies by households has created an opportunity to provide families with young children with access to high-quality oral health information at convenient times and locations. Using community agencies (eg, Head Start and public health programs) that offer parenting education is an effective approach to reaching families in low-income households. Objective: This study aimed to explore the extent to which a coordinated, in-person oral health prevention intervention, together with an accompanying smartphone app, BeReadyToSmile, is feasible to implement among caregivers of young children. Methods: The BeReadyToSmile program targeted parents of children aged 0 to 6 years attending parenting education classes. This study was designed as a single-group pre-post feasibility study that included quantitative surveys and open-ended feedback. A total of 30 parents attended an in-person session on child oral health and were invited to use the BeReadyToSmile smartphone app. Preintervention and postintervention surveys were administered to assess pediatric oral health knowledge, attitudes toward child toothbrushing, brushing intention, brushing efficacy, program satisfaction, and ease of use. Results: Significant effects were observed on parent-reported pediatric oral health knowledge, attitudes toward brushing, brushing intention, and toothbrushing efficacy. Out of the 30 parents invited to use the BeReadyToSmile app, 1 (3%) completed no sessions and 20 (67%) completed all sessions. Participants rated the app highly on measures of satisfaction and use. We found significant increases in pediatric oral health knowledge (.004), child brushing attitudes and intention (=.01), and parental efficacy regarding child toothbrushing (=.03). Conclusions: Caregivers reported positive experiences with the implementation of BeReadyToSmile, indicating the overall feasibility of delivering oral health prevention to households with young children both in person and through a facilitated smartphone app. Further studies should include a larger and more diverse sample, randomized comparison conditions, and a longer follow-up period to assess outcomes. Trial Registration: ClinicalTrials.gov NCT03637309;
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/337d501308d195f5f436e32d7939b2c5" />
Background: Health care professionals’ perceptions of telemedicine, its usability, and the presence of organizational barriers are important determinants of the successful implementation of digital solutions in health care. In Kazakhstan, the use of international assessment instruments requires contextual adaptation. The Telehealth Usability Questionnaire-Model for Assessment of Telemedicine-Kazakhstan version (TUQ-MAST-KZ) questionnaire was previously developed and psychometrically validated by integrating elements of the TUQ and MAST frameworks to assess perceptions of telemedicine within the national context. Objective: The aim of this study was to conduct the first pilot application of the TUQ-MAST-KZ questionnaire with physicians in Kazakhstan and perform an initial assessment of the organizational, technical, and educational aspects of telemedicine implementation. Methods: This cross-sectional study involved an anonymous online survey using the TUQ-MAST-KZ questionnaire, which covers perceptions of telemedicine, formats of use, platform usability, communication-related aspects, telemonitoring, organizational conditions, and implementation barriers. Responses from 156 physicians were analyzed. Stratified nonparametric comparisons were performed by sex, age group, work experience (years), and workplace, adjusted for multiple comparisons. Results: The most used telemedicine formats were telephone consultations (78/156, 50%), video consultations (69/156, 44.2%), chats and messaging applications (57/156, 36.5%), and mobile apps (48/156, 30.8%). The Kazakhstan National Telemedicine Network was used by 14.7% (23/156). Wearable devices were used by 5.8% (9/156). Telemedicine technologies incorporating artificial intelligence elements were used regularly by 13.5% (21/156) and occasionally by 32.1% (50/156) and not used by 50.6% (79/156). Positive ratings were as follows: 48.7% (76/156) regarding the simplicity and intuitiveness of telemedicine platforms; 56.4% (88/156) regarding the timeliness of patient condition monitoring; 51.9% (81/156) regarding the effectiveness of telemedicine for the management of patients with chronic diseases. The potential usefulness of telemonitoring for earlier detection of deterioration of a patient’s condition was rated as fairly or very high by 48.7% (76/156); 41% (64/156) rated it as moderate. Only 35.9% (56/156) positively rated the connection’s reliability and stability. Regarding the accuracy of wearable device data transmission, 57.1% (89/156) responded neutrally, potentially indicating ambiguity in perception, limited personal experience, or difficulty evaluating this aspect. Readiness to recommend telemonitoring at the national level was more often rated as moderate, high, or very high (78/156, 50%; 42/156, 26.9%; 14/156, 9%, respectively). Conclusions: This pilot application of the TUQ-MAST-KZ questionnaire showed a generally moderately positive perception of telemedicine by physicians, who recognized its potential clinical and organizational value. However, we identified substantial technical and institutional barriers, including connection instability, concerns about the accuracy of data transmission, insufficient process formalization, and a need for additional training. These preliminary findings should be interpreted in light of the pilot study design; however, they may serve to inform future larger-scale research and the development of organizational measures related to physician training, protocol standardization, and infrastructure support for telemedicine implementation.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/804a11a35fc36f67f33a929fd1435eba" />
Background: The widespread prevalence of chronic pain (CP) significantly impacts daily functioning worldwide. In mainland China, maintaining engagement in biopsychosocial interventions remains challenging. Gamification, designed based on self-determination theory, can enhance motivation, while machine learning (ML) algorithms can assist clinicians in dynamically optimizing pain management. Objective: This study aimed to (1) evaluate the preliminary effectiveness of a gamified pain management (GPM) program on CP and psychological outcomes and (2) identify key factors of significant pain improvements through the application of ML to guide intervention adjustments. Methods: A single-arm, pre-post study was conducted with 16 participants with CP in mainland China, recruited via social media using convenience sampling. Participants engaged in a 10-week web-based GPM intervention consisting of education, physical activities, and gamified elements, including points, avatars, and feedback. Primary outcomes were pain intensity and interference measured by the Brief Pain Inventory. Secondary outcomes included anxiety, depression, and quality of life. Analysis included paired tests, and ML models were trained to predict clinically meaningful pain reductions. Shapley additive explanations, least absolute shrinkage and selection operator regression, association rule mining, and Kaplan-Meier survival analysis were used to identify key predictors and optimal sessions and intervention durations across subgroups. Results: A total of 16 participants were engaged, with a mean age of 27.63 (SD 9.584) years. Results from paired tests reported significant improvements in pain intensity (decreased by 27.3%, 95% CI 1.061 to 3.064; =.001), pain interference (decreased by 27.3%, 95% CI 8.159-17.216; <.001), and psychological distress, including anxiety (=3.538, 95% CI 0.969 to 3.906; =.003) and depression (=4.559, 95% CI 2.230 to 6.145; <.001). The gradient boosting model demonstrated the highest predictive accuracy (area under the curve=0.89 and accuracy=0.82). Least absolute shrinkage and selection operator regression identified session 3 (β=−0.45, 95% CI −0.68 to −0.22; <.001) and session 5 (β=−0.32, 95% CI −0.59 to −0.05; =.02) as most predictive of clinical success, while association rule mining revealed effective session combinations for different patient subgroups. Time-to-event analyses indicated that individuals with low back pain and higher baseline severity required longer intervention durations for improvement (5 wk; =.03). Conclusions: This pilot study presents an innovative method that combines ML with dynamic engagement data from a GPM program during interventions, rather than relying on static baseline data in prior studies. The results show preliminary efficacy and identify specific optimal session combinations and personalized treatment durations for different pain subgroups. These exploratory findings contribute to the field by providing a data-driven method for adaptive, personalized digital health interventions that move beyond one-size-fits-all strategies, potentially enabling clinicians to modify content and dosage to improve engagement and outcomes if validated in larger sample trials. Trial Registration: Chinese Clinical Trial Registry ChiCTR2400094247; https://www.chictr.org.cn/showprojEN.html?proj=245138
Dating back more than a century, biobanks have outgrown their beginnings as small, local collections to become large, global facilities that store and handle millions of samples and serve thousands of researchers at any given time. Over the years, biobanks have transformed from passive repositories into active research infrastructures that are increasingly bridging the gap between medical research and clinical applications.
“Today’s biobanks have evolved far beyond sample storage,” said Yan Zhang, PhD, president of proteomic sciences at Thermo Fisher Scientific. “They are automated, digitally connected systems integrated with hospitals and health networks to ensure appropriate consent, longitudinal clinical context, and the ability to re-engage participants over time.”
Yan Zhang, PhD President Thermo Fisher Scientific
As safeguards of clinical samples, biobanks fulfill a central role in the advancement of precision medicine. Access to the right samples can make or break a research project, with most researchers reporting that they have had to limit their scope of work because of difficulties obtaining the samples they need.
“Robust, population-scale biobanking enables precision medicine to move from isolated findings toward broader clinical relevance,” said Zhang. “Modern biobanks combine genomics, proteomics, and other high-dimensional omics platforms with robust data architecture, high-performance computing, and artificial intelligence (AI)-driven modeling. Dedicated data science teams integrate molecular data, longitudinal health records, and curated public datasets to generate biologically meaningful interpretations.”
Biobanks now provide the infrastructure needed to support population-scale, longitudinal studies that allow scientists to uncover molecular drivers of disease and understand their evolution over time to ultimately identify biomarkers, develop targeted treatments, and inform clinical decisions.
“We’re seeing researchers design studies with scale in mind,” Zhang noted. “They’re combining proteomics, genomics, and clinical data to generate insights that are both statistically powerful and relevant to real-world populations. There’s also a clear shift from searching for a single biomarker to building a more complete, systems-level understanding of disease.”
To navigate today’s rapidly shifting landscape and meet their core purpose of supporting cutting-edge clinical research, biobanks have to keep up with fast-moving targets. Going forward, moving from initial discovery to translation will remain the number one challenge in precision medicine. “Generating discovery insight is no longer the limiting factor,” said Zhang. “Validating, standardizing, and implementing those insights at scale is.”
A matter of scale
Martin K. Rutter, MD Deputy Chief Scientist UK Biobank
One of the most transformative shifts in biobanking over the past decade has been an exponential increase in the scale of data collection and sample storage. At the forefront of this expansion is the UK Biobank, which currently stores around 18 million samples from 500,000 participants, together with imaging and biomarker data, healthcare records, questionnaires, physical measurements, demographics, lifestyle, and environmental data collected over the course of 20 years. This depth of phenotyping is what makes the data so valuable to researchers worldwide, said Martin K. Rutter, MD, professor of cardiometabolic medicine at the University of Manchester and deputy chief scientist at the UK Biobank. “When you link all that together, you can get amazing insights into the biology of disease.”
To keep up with increasing storage needs and researcher requests, the UK Biobank is now getting ready to move more than 10 million samples currently stored in its main laboratory to a new building in central Manchester by the end of the year. The new storage facility is designed to quadruple sample retrieval speed while making the whole infrastructure more energy-efficient and environmentally friendly.
The scale at which facilities like the UK Biobank operate today would have been unthinkable when it was established two decades ago. Such massive growth has been driven by rapid technological advances across genomics, transcriptomics, and proteomics, with costs continuing to fall while coverage, speed, and accuracy keep surging.
Partnerships with the pharmaceutical industry have also been instrumental in nurturing this exponential growth. This can be seen in initiatives like the UK Biobank Pharma Proteomics Project (UKB-PPP), a collaboration between the UK Biobank and 14 biopharmaceutical companies with the goal of analyzing proteomics data from 600,000 samples.
In the long run, scale provides the backbone to enable increasingly ambitious, statistically powerful studies. However, as they grow, biobanks face the challenge of navigating a constantly shifting landscape while making sure the samples and data they collect, store, and maintain are valuable to the entire research community they serve.
“Our job is to make the data available to researchers,” said Rutter. “We are involved now more than ever in connecting with research teams and trying to understand what their needs are.”
Through surveys and consultations, the UK Biobank actively gathers information to design prospective data collection programs that anticipate researcher needs. Next year, the biobank is planning a repeat assessment of its whole cohort, focusing on measurements of aging. The goal is to support researchers looking into causal pathways and mechanisms driving age-related diseases, empowering the development of preventive interventions and new diagnostics and treatments for age-related conditions.
Keeping pace with the evolving demands of researchers, industry, and the broader public is essential for biobanks to secure the funding necessary not only to operate but also to expand such vast enterprises, which remains a major challenge across this resource-intensive field.
Diversity takes the spotlight
Historically, samples collected by biobanks are biased in favor of participants who are white, middle-class, and have a higher education. This creates major disparities in the applicability of clinical research. In fact, studies have shown that patients from non-European ancestry backgrounds have not benefited equally from precision drugs approved by the U.S. Food and Drug Administration (FDA) to treat a range of cancer indications.
Even within biobanks dedicated to sampling the population of a specific region, ethnic minorities, low-income, or elderly people are often underrepresented, skewing results against the real-world populations they strive to serve. As the research community increasingly recognizes the importance of more diverse and representative patient cohorts, demand is rising for resources that address these barriers.
Representation is at the heart of All of Us, a program launched by the National Institutes of Health in 2018 to address the gap present at the time in many biobanks and sample repositories. This precision medicine initiative was designed to enroll participants who reflect the full range of populations found within the U.S., including individuals of varied ancestry backgrounds as well as those living in rural commmunities, which are rarely represented in biorepositories due in part to longstanding barriers to research participation, such as the logistical challenges of collecting samples and data from participants in remote locations.
Joshua C. Denny, MD CEO All of Us
“A lack of diversity impoverishes discovery and applicability of findings for all,” said Joshua C. Denny, MD, CEO of the All of Us Research Program.
For instance, data collected by All of Us has been used to investigate APOL1 gene variants linked to kidney disease, which are more common among people of West African ancestry. This research led to the identification of a novel APOL1 variant that can reduce the risk of kidney disease in individuals carrying high-risk variants.
The program has so far enrolled about 870,000 participants across all U.S. states, with about 80% of them representing communities that have historically been underrepresented in biomedical research. This has been achieved by emphasizing accessibility and flexible participation models; participants can enroll digitally and choose whether to share access to their electronic health records, donate biospecimens, and complete demographics and lifestyle surveys. They may also opt to provide saliva samples, simplifying logistics in rural areas with limited access to blood collection facilities.
“What works in a rural location is different from what works in a big city like New York,” said Denny. Whether it comes to location, age, or language, he emphasized the importance of adapting how the program approaches and engages each population.
Democratizing access to patient data across the research ecosystem is another major biobanking challenge that All of Us is committed to addressing. The program has established a streamlined access model that enables researchers to access the data they need in less than two hours if they belong to one of the 1,300 already approved institutions across the world. Together with central data storage and cloud-based analysis tools, their setup is designed to make the data accessible to researchers lacking the resources and local infrastructure for high-performance computing.
Towards global integration
With precision medicine studies steadily escalating both in size and complexity, researchers increasingly seek to bring together data stored across diverse biobanks to power larger, more ambitious studies with broader scientific and societal impact. However, building the infrastructure needed to enable cross-biobank studies is still a challenge, starting with convening stakeholders to harmonize data collection standards and establish international guidelines.
Anticipating this need, in 2013 the European Union established the Biobanking and Biomolecular Resources Research Infrastructure – European Research Infrastructure Consortium (BBMRI-ERIC), which currently coordinates the activity of about 500 biobanks across 32 countries.
Jens K. Habermann, MD, PhD Director General BBMRI-ERIC
“Precision medicine can only move forward with a strong starting point for research,” said Jens K. Habermann, MD, PhD, professor for translational surgical oncology and biobanking at the University of Lübeck and director general of the BBMRI-ERIC. “It can be very difficult for scientists to get all the information they need in one place, and this is what biobanks can enable.”
Pulling together data from all its members, the BBMRI-ERIC has set up a central catalogue for biobanks, biomolecular resources, and other data and sample collections, which users can employ to identify relevant resources and build virtual cohorts tailored to their research needs. The consortium also works with international committees to set guidelines and support members working towards compliance with international standards.
Despite ongoing progress, there are still obstacles ahead when it comes to harmonizing biobanking practices worldwide, including data collection, annotation, storage, and sharing. Tackling differences in data protection, consent, ethical standards, and regulatory requirements across borders will be another necessary step towards broader standardization. Finally, biobanks will need to invest in cybersecurity to ensure patient data can be shared between institutions safely.
Funding will be key to successfully addressing all these challenges. On this front, biobanks face the difficult task of maintaining their existing infrastructure, staying up to date and relevant to the research community, and investing in cross-biobank initiatives. All this must be balanced with growing financial pressure on research centers, hospitals, and the governments supporting them.
As part of its 10-year roadmap, the BBMRI-ERIC is setting the goal of forming international networks that bring together more diverse biobank types, such as environmental, wildlife, veterinary, and plant biodiversity repositories. The overarching aim is to move towards a One Health approach to biobanking, where samples and data that expand beyond monitoring human populations are brought together to tackle overlapping challenges that simultaneously affect human, animal, and environmental health.
Data-driven horizons
As the field forges ahead, biobanks are undergoing broad transformations in the way they operate. On the technology side, these changes are being propelled by the rise of multi-omics techniques in precision medicine research, as well as by rising demand from the research community for non-invasive patient monitoring data and longitudinal sample collection. All of these will be critical for the development of the next generation of personalized therapies and diagnostics.
“Over the next decade, biobanks are expected to become increasingly integrated into clinical and translational workflows,” said Zhang. “Proteomics, in particular, will play a growing role in helping us understand the dynamic biology of disease, enabling earlier detection, better prediction of recurrence, and more precise therapeutic strategies.”
A key driver of this shift will be AI. No longer just a supporting tool, AI is now becoming an integral part of biobank operations, contributing to real-time sample monitoring, predictive maintenance, risk management, and decision making.
On the data analysis side, Zhang has seen how AI is redirecting the focus from data generation to data interpretation. She said, “Biobanking has already enabled the collection of high-quality biospecimens linked to large-scale molecular and clinical datasets. The challenge now is extracting meaningful biological insight from that complexity.”
Although still in its early days, AI is becoming central to how researchers make use of biobank data, noted Rutter. Drawing from the UK Biobank data, recent studies have developed AI models that can predict a patient’s risk of stroke based on retinal images, calculate the risk of future disease by looking at an individual’s disease history, or spot neurodegenerative diseases like Alzheimer’s and Parkinson’s early using brain scans and physical activity data.
Going forward, Rutter expects to see biobanks moving away from static cohorts and in favor of continuous data collection, enabling more powerful predictions. For example, the UK Biobank is developing a mobile app that can track a participant’s physical activity and monitor their location and sleep patterns, offering an in-depth look at how a variety of factors affect their health with much more accuracy than self-reported surveys.
Over time, all these advances will steer clinical practice from treatment to prevention, allowing healthcare professionals to act early in the patient journey, when interventions are most effective, and eventually, even before disease develops. Ultimately, addressing complex diseases will require coordinated contributions from all stakeholders, including AI innovators, drug developers, clinicians, technology providers, and policymakers.
“The next decade will be incredibly exciting,” said Denny. “It will be all about leveraging the huge scale of resources that are just emerging today.”
Clara Rodríguez Fernández is a science journalist specializing in biotechnology, medicine, deeptech, and startup innovation. She previously worked as a reporter at Sifted and editor at Labiotech, and she holds an MRes degree in bioengineering from Imperial College London.
Gilead Sciences has agreed to acquire German-based Tubulis for up to $5 billion, the companies said today, in a deal designed to expand the buyer’s antibody–drug conjugate (ADC) capabilities with a focus on fighting cancer.
Headquartered in Munich, privately held Tubulis has developed next-generation ADC candidates based on its own conjugation, linker and payload technologies intended to more selectively deliver diverse payloads to tumors deemed to be of high unmet need. The companies said Tubulis’ programs and platforms have broad potential across multiple tumor types, complementing Gilead’s development and commercialization expertise in oncology.
“We like the strategic fit and deal terms of the Tubulis (private) acquisition,” Daina M. Graybosch, PhD, senior managing director, immuno-oncology and a senior research analyst at Leerink Partners, wrote this morning in a research note. “This is more than an oncology bolt-on; we see real platform value in application of Tubulis’ ADC technologies to other therapeutic areas, namely virology.”
Tubulis’ lead pipeline candidate, TUB-040, is a sodium-dependent phosphate transport protein 2B (NaPi2b)-targeting topoisomerase-I inhibitor (TOPO1i) ADC that is now under study in the Phase Ib/II NAPISTAR1-01 trial (NCT06303505) assessing its safety, pharmacokinetics, and preliminary efficacy as a treatment for platinum-resistant ovarian cancer and non-small cell lung cancer (NSCLC).
In October at the European Society for Medical Oncology (ESMO), Graybosch noted, Tubulis presented data for TUB-040 showing a confirmed 50% overall response rate (ORR) and a 60% unconfirmed ORR across dose levels and irrespective of target antigen—results that were competitive with more mature datasets from leading TOPO1i ADCs.
“Though the dataset was early, and our primary outgoing question was how durability would mature, we suspect that Gilead saw durability maturing positively in their diligence,” Graybosch added. “If TUB-040 proves active in NSCLC, the program could complement their Trodelvy and IO [immune-oncology] lung programs. We wonder if Gilead saw early clinical NSCLC data in their diligence and if excitement around the emerging signal drove some of Tubulis’ valuation.”
Another Tubulis pipeline candidate, TUB-030, is a 5T4-targeting ADC that according to the companies has shown promising initial clinical data across various solid tumor types. TUB-030 is currently under study in the Phase I/IIa 5-STAR 1-01 trial (NCT06657222), a first-in-human study which aims to evaluate the safety, tolerability, pharmacokinetics, and efficacy of TUB-030 as a monotherapy in patients with advanced solid tumors. Tubulis has said it is developing TUB-030 for up to 13 undisclosed solid tumor indications.
Partners since 2024
The acquisition deal follows a two-year, up-to-$465 million collaboration with Tubulis launched in December 2024. Gilead gained access to Tubulis’ Tubutecan and Alco5 platforms after signing an exclusive option and license agreement to discover and develop an ADC against a solid tumor target.
At the time, Gilead agreed to pay Tubulis $20 million upfront, received an option that if exercised would have given Tubulis an additional $30 million—plus up to $415 million in payments tied to achieving development and commercialization milestones, as well as mid-single to low double-digit tiered royalties on sales of marketed products resulting from the collaboration.
“Today’s agreement follows a two-year collaboration with Tubulis, which has given us strong conviction in their programs and research capabilities,” Gilead Chairman and CEO Daniel O’Day said in a statement. “The agreement to acquire Tubulis is a significant milestone in Gilead’s progress in oncology. The company brings a clinical-stage candidate that is a potential new treatment for ovarian cancer, as well as a next-generation ADC platform and a promising early pipeline.”
“Bringing this potential into Gilead would further expand what is already the strongest and most diverse pipeline in our company’s history,” O’Day declared.
Investors appeared less enthusiastic about the acquisition, as shares of Gilead dipped 1.7% in early Tuesday trading to $137.80 as of 12:01 p.m. ET.
Tubulis is Gilead’s third announced acquisition this year. The biotech giant announced plans in March to buy Ouro Medicines for up to $2.18 billion, and in February agreed to acquire Arcellx for up to $7.8 billion—for which it agreed last week to extend its tender offer until 5 p.m. ET on April 24.
Under the acquisition deal, Gilead agreed to acquire all of the outstanding equity of Tubulis for $3.15 billion in upfront cash payable at closing, and up to $1.85 billion in payments tied to milestones.
The transaction is expected to close in the second quarter subject to expiration or termination of specified regulatory filings and other customary conditions.
Upon closing of the deal, Tubulis will operate as a dedicated ADC research organization within Gilead, with the Munich site serving as a hub for ADC innovation, building on its integrated discovery, manufacturing, and clinical capabilities to advance next generation ADCs.
Gilead said it plans to finance the transaction with a combination of cash on hand and senior unsecured notes. Gilead finished 2025 with $10.605 billion of cash, cash equivalents and marketable debt securities, up from $9.991 billion as of December 31, 2024.
You’re reading the web edition of D.C. Diagnosis, STAT’s twice-weekly newsletter about the politics and policy of health and medicine. Sign up here to receive it in your inbox on Tuesdays and Thursdays.
The 2027 budget that the Trump administration released on Friday is in many ways a repeat of last year’s proposal: It includes deep cuts to the National Institutes of Health, the elimination of a health research agency, and the creation of a new agency devoted to chronic diseases called the Administration for a Healthy America.
