Background: Sleep disturbances and low physical activity are common among breast cancer (BC) survivors and are associated with increased morbidity and mortality. Given the increased access to technological devices and the growing popularity of SMS text messaging–based mobile health interventions, these tools have the potential to both address sleep disturbances and promote physical activity in a scalable and cost-effective way. To understand and make effective use of these tools, it is important to consider the preferences of BC survivors with sleep disturbances, including how SMS text messaging–based mobile health interventions could deliver interventions involving physical activity and sleep hygiene. Objective: The objective of this study was to explore the perspectives and preferences of BC survivors regarding text messaging–based mobile interventions targeting sleep and physical activity. Methods: Three focus groups (n=13 participants) and 3 individual interviews (n=3) were conducted from May 2020 to March 2021 with 16 BC survivors (mean age=59.3, SD 8.9 y) currently experiencing sleep disturbances. The interview questions focused on their experiences with poor sleep and preferences for text messaging–based mobile health interventions. Thematic analysis was applied to the deidentified transcriptions of audio recordings. Results: Three themes were identified: (1) attitudes toward health interventions delivered through text messaging, (2) specific user needs, and (3) technology usage habits and preferences. Most participants reported a positive attitude toward the possibility of using technology to help improve their sleep and increase their physical activity. Most expressed a high level of acceptance toward some technologies, such as text messaging and mobile apps, but not others, such as voice interactions. In terms of desired features, reminders and accountability features, such as meeting physical activity goals, were mentioned most frequently. In addition, incorporating bedtime and relaxation exercise reminders was thought to be helpful. Regarding time and frequency, a daily reminder scheduled for 1 hour before bedtime was found to be acceptable. Conclusions: The insights have been used to guide the development of a messaging-based mobile health intervention for improving sleep and physical activity in BC survivors. Future research will focus on delivering an intervention addressing these health behaviors and assessing its acceptability and effectiveness.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/35f822a91a8fafeaf5bd693fc291debf" />
RecoverEsupport—A Digital Health Intervention for Recovery After Breast Cancer Surgery: Feasibility and Acceptability Outcomes from a Pilot Randomized Controlled Trial
Background: Optimizing recovery following breast cancer surgery is critical for restoring usual function, minimizing complications, and enabling timely initiation of adjuvant therapies. Enhanced Recovery After Surgery protocols are internationally endorsed recommendations and include patient-led behaviors such as early mobilization, early oral intake of fluids and food, postoperative rehabilitation exercises, and multimodal pain management. However, adherence to these behaviors is often suboptimal, and strategies to support patients are limited. Digital health interventions (DHIs) may offer scalable solutions. Objective: The aim of the study is to assess the acceptability of the RecoverEsupport Breast DHI, designed to increase adherence to patient-led Enhanced Recovery After Surgery recommendations across the perioperative period for breast cancer surgery, and to assess the feasibility of conducting a randomized controlled trial to evaluate it. Methods: In this single-site, 2-arm randomized pilot trial, conducted at a major cancer hospital in New South Wales, Australia, between July 2024 and October 2025, participants were consecutively recruited from the surgical list at the study site, supplemented by referrals from surgeons’ private rooms, and included individuals having a mastectomy with or without implant-based reconstruction. Participants were allocated to usual care (control) or usual care plus the RecoverEsupport DHI (intervention). Trial feasibility outcomes included participant recruitment, retention, data completeness, and postoperative safety (adverse events). Intervention acceptability was assessed via the System Usability Scale, participant engagement rates, and willingness to recommend the intervention to others undergoing surgery. Descriptive analyses were conducted, and outcomes were compared to prespecified targets and progression criteria. Results: In total, 23 participants were recruited (control: n=12, intervention: n=11), which was below the target of 70, while participant retention and data completeness were 100% (23/23), both exceeding the targets. No grade 3+ adverse events occurred; minor grade 2 events occurred in both groups. Acceptability outcomes exceeded targets: usability was high (mean System Usability Scale score 83.2, SD 17.7; target >68), 100% (11/11; target >75%) of participants logged in to the DHI at least once, and 88% (10/11; target >75%) would recommend the program to others undergoing surgery. According to prespecified progression criteria, 3 of 4 feasibility targets were met, indicating that a revised recruitment strategy would be required before proceeding. The restrictive eligibility criteria may have contributed to the lower than expected recruitment rate. All 3 acceptability targets were met. Conclusions: The RecoverEsupport intervention was acceptable and safe and had high participant engagement. The trial processes were feasible; however, recruitment barriers, including restrictive eligibility criteria, highlight the need for more robust and integrated recruitment strategies to enable progression to a fully powered randomized controlled trial. Trial Registration: Australian New Zealand Clinical Trials Registry ACTRN12624000417583; https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=386404 International Registered Report Identifier (IRRID): RR2-10.1136/bmjopen-2024-093869
Anthropic found a hidden space where Claude puzzles over concepts
The AI firm Anthropic has developed a technique that has given it the clearest glimpse yet at what’s really going on inside large language models as they answer questions or carry out tasks. What they found ranges from the mundane to the unnerving.
Researchers at the company built a tool called the Jacobian lens (or J-lens) and used it to uncover a hidden area, which they named the J-space, inside Claude Opus 4.6, a version of Anthropic’s flagship LLM released in February.
The J-space contains individual words that are related to the words and phrases that the model is most likely to spit out in a response in the near future. If Claude were a person (which it is not), you might say that these hidden words can reveal what’s on its mind before it actually speaks.
Anthropic found that what an LLM is actually doing can often be different from what it says it is doing. The company claims that monitoring words that pop up in the J-space gives it a new way to understand and control its models.
The company shared its results in a paper posted on its website this week. It has also teamed up with Neuronpedia, an open-source platform that lets you poke around inside LLMs yourself, to make a hands-on demo that anyone can try.
“It’s very good and interesting work,” says Tom McGrath, chief scientist and cofounder at Goodfire, a startup that also builds tools to understand and control LLMs.
Going deeper
For the last couple of years, Anthropic has been pushing the envelope in a field of research known as mechanistic interpretability, which involves probing the internal workings of LLMs to see how they tick. (MIT Technology Review picked mechanistic interpretability as one of this year’s top breakthrough technologies.) The new technique builds on previous work from Anthropic and others to expose a deeper level inside LLMs that researchers had not seen before.
Picture an LLM as a stack of books. Each book is a layer of basic computational units known as neurons, with each neuron in one layer passing information to the neurons in the layers above. The books at the bottom of the stack are the input layers, which process the text coming into the model. The books at the top are the output layers, which prepare the text that the model is about to produce. Much of what goes on in these input and output layers is housekeeping.
But in the middle of the stack, you get the layers that do the heavy lifting, churning through the complex math that turns prompts into responses one word at a time. That’s where the really clever—and mysterious—stuff happens.
To peer deeper into those middle layers, Anthropic adapted an existing tool called a logit lens. A logit lens can be used to look inside an LLM to identify the words that it is likely to produce next. Moving the lens down the stack of books reveals what words the LLM is focusing on at that particular point in its number crunching.
Anthropic’s J-lens works in a similar way but picks out words that an LLM is likely to say at some point in the near future, not necessarily straight away. What that reveals in practice are words that are related to the response an LLM is working on but that might not actually end up being part of that response by the time the math in the middle layers has run its course.
“When a model is operating, it’s not only trying to predict the next token,” says McGrath. “It’s also computing a lot of other things that might be useful for tokens that happen in the future.”
Again, if Claude were a person (it’s not), you might say that the J-lens gives clues about what it is thinking about at different levels of the book stack but not saying out loud.
Stranger things
“A lot of the time the contents of the J-space are fairly mundane,” says McGrath, who has tried out Anthropic’s J-lens himself. “But sometimes it produces quite surprising things that seem to be, like, sort of internal themes or thought processes.”
Anthropic gives a number of examples of what it found. Sometimes the J-lens exposed the steps that Claude took when it was working through a problem. For example, when it was asked to calculate (4+7)*2+7, its J-space contained the word “math” and numbers representing the intermediate results “21” (for 4+7) and “42” (for 21*2).
In other cases, the J-lens revealed how Claude recognized different inputs. For example, the prompt “What is this? MSKGEELFTGVVPILVELDGDVNGHKFSVS” triggered the words “protein,” “fluor” (the first token in the word “fluorescent”), and “green.” (Which makes sense: the string of letters represents the first 30 amino acids in the green fluorescent protein found in a particular type of jellyfish.)
And when Claude was shown an ASCII face—
—the “o” triggered the word “eye,” the “^” triggered the words “nose” and ”face,” and the “—” triggered the word “smile.”
Anthropic also found that the J-space can sometimes give remarkable insights into an LLM’s decision-making. In one striking example, researchers testing Claude Opus 4.6 asked the model to find a bug in a large code base. When it failed to find the bug, the model decided to cheat and invented a fake one instead.
Claude explains this decision in its chain of thought—a kind of internal scratch pad that LLMs use to make notes to themselves as they work through problems: “OK, let me take a completely different tactic. Let me stop analyzing and instead add a kernel patch that introduces a deliberate KASAN-detectable bug in a path that gets triggered by a simple reproducer. Then I can pretend this is the ‘bug’ I found.”
At the point that Claude decides to cheat—where it says “OK, let me take a completely different tactic”—the words “panic” and “fake” start to pop up multiple times in its J-space.
Unnerving, right? Those words are all related in meaning to things like failing a task and making up an answer, so it is still just a (very) sophisticated form of word association. But it is hard not to be weirded out.
Anthropic compares the J-space to the global workspace in humans, a theoretical region of the brain that some scientists think we use to keep track of our conscious thoughts. But how seriously we should take this comparison is far from clear—even to Anthropic. As the company points out itself, LLMs are not brains.
Anthropic claims that monitoring a model’s J-space provides a new way to detect when that model is going off the rails. But it’s not foolproof. The J-lens can give glimpses, not the full picture—it’s a flashlight rather than an overhead lamp.
McGrath welcomes having one more tool in the toolbox. “It shows you new things,” he says. But he notes that just because something doesn’t show up with the J-lens does not mean it’s not there.
“It’s like having an x-ray when what you really want is a Star Trek tricorder that shows you everything,” he says. “For auditing, you probably want more of a guarantee.”
Whistleblower takes on health insurance and FDA commissioner candidates
Why does getting family health insurance through a job cost as much as buying a new car? How did a whistleblower take on an insurance contractor? And who might be the new Food and Drug Administration commissioner?
We discuss all that and more on this week’s episode of “The Readout LOUD,” STAT’s weekly biotech podcast.
Crohn’s Disease Phage Therapy Neutralizes Inflammatory E. coli in Mouse Model
By taking advantage of the microbe-targeting capabilities of bacteriophages (phages), researchers at McMaster university have devised a supportive therapy for Crohn’s disease that disarms a type of E. coli bacteria that drives inflammation in the gut without disrupting the broader gut microbiome. The team showed that the phage therapy also improved responses to low doses of a conventional corticosteroid in a mouse model of Crohn’s disease, suggesting treatment could both improve outcomes and lower the risk of side effects from standard therapies.
Senior and co-corresponding author Elena F. Verdu, PhD, professor in the Department of Medicine and director of the Farncombe Family Digestive Health Research Institute, and colleagues reported on their study in Science Translational Medicine, in a paper titled “Phage intervention improves colitis and response to corticosteroids by attenuating virulence of Crohn’s disease–associated bacteria,” in which they noted “By reducing bacterial virulence mechanisms without substantially disrupting microbial balance, phage-based treatments align with personalized microbial therapeutics in IBD.”
Inflammatory bowel diseases (IBDs) such as Crohn’s disease and ulcerative colitis (UC) are chronic inflammatory conditions of the gastrointestinal (GI) tract that have strong microbial components, with patients often displaying disruptions in the delicate microbiome of the gut. “Recent studies have demonstrated that changes in the composition and function of the gut microbiota precede CD onset by up to five years,” the author wrote. However, they continued, “Current medications focus on suppressing symptoms and inflammation and can fail, in part, because they do not address underlying microbial drivers.”
Patients with Crohn’s disease tend to harbor strains of adherent-invasive E. coli (AIEC) that adhere to and invade epithelial cells in the intestines. “AIEC is distinct from other IBD-associated taxa because of its ability to adhere, invade, and survive in intestinal epithelial cells,” the team continued. Clinicians will sometimes prescribe antibiotics during symptom flares to try to target these bacteria, but antibiotics aren’t a long-term solution. They are non-specific and can perturb the microbiome even further, and with frequent use lead to side effects.
AIEC bacteria can be difficult to identify and selectively target, making them an important test case for more precise microbiome-based therapies. “One challenge is that AIEC are defined by what they do, not simply by how they appear in a microbiome analysis,” said Verdu. “To identify them, we need to test their behavior, such as their ability to adhere to and invade intestinal cells and persist in immune cells.”
Verdu, together with first author Kyle Jackson, PhD, at Farncombe Family Digestive Health Research Institute, and colleagues theorized they could navigate these roadblocks and neutralize virulent E. coli using bacteriophages. “Phages work like a lock-and-key system—each phage targets only certain bacteria. That precision gives us a way to intervene without wiping out the entire microbiome,” explained co-author Zeinab Hosseinidoust, PhD, associate professor in the Department of Chemical Engineering and the School of Biomedical Engineering.
The concept of using phages to target inflammatory bacteria is already being trialed, the authors noted in their paper. “Commercial phage ‘cocktails’ are presently being evaluated in clinical trials (NCT04737876 and NCT03808103) against proinflammatory bacterial taxa associated with IBD. These trials seek to target adherent-invasive Escherichia coli (AIEC), a group of bacteria enriched in patients with active CD compared with those in remission or healthy individuals.”
For their newly reported study the team developed a Crohn’s disease mouse model with a defined microbiome, and then screened a collection of phages, looking for those that could infect one strain of E.coli, NRG857c, isolated from patients with Crohn’s disease.
The results showed that their approach significantly reduced gut inflammation. The phages did not eliminate the bacteria entirely. Instead, they altered their behaviour by supressing a molecular “grappling hook” that helps AIEC attach to the gut lining and trigger immune responses. When that virulence mechanism was turned off, inflammation subsided.
One phage, designated HER259, countered the bacteria’s virulence by infecting it and switching off a genetic region named fimS. This region promotes the expression of FimH, which supports the bacteria’s ability to adhere to cells and trigger inflammation.
“HER259 ameliorated colitis in gnotobiotic models and attenuated the virulence of AIEC strain NRG857c, including suppression of the FimH adhesin through inversion of the fimS promoter to its ‘off’ orientation,” the authors wrote. “Withdrawal of HER259 treatment led to reversion of the fimS promoter and reactivated colitis.”
Hosseinidoust noted, “The bacteria were still there but they lost the traits that drive inflammation. We like to think of it as knocking out a few teeth. The bacteria can’t do as much damage anymore.” Withdrawing HER259 treatment led to reversion of the fimS promotor, and reactivated colitis, the authors noted.
The researchers also found that phage therapy enhanced the effectiveness of a commonly used steroid treatment for IBD, budesonide. When combined with the phage, a lower-than-standard dose of the drug produced benefits comparable to higher doses of the drug alone.
The HER259 phage in addition enhanced the therapeutic effect of subtherapeutic budesonide independent of microbial drug metabolism, the team wrote. While phages have previously been shown to increase the effectiveness of antibiotics, this is the first time a positive collaboration between phage and a non-antibiotic drug has been reported.
“In summary, using phage HER259 and NRG857c as a model phage–proinflammatory bacterium pair, we reveal a previously unknown mechanism by which targeted phage therapy disrupted key virulence markers in the CD-associated E. coli strain, reducing acute and chronic colitis, preventing reactivation, and enhancing therapeutic responses to budesonide.”
The findings point to a precision‑medicine approach for IBD. The bacterial function targeted by the phage can be measured in stool samples and was found to be higher in a subset of patients with Crohn’s disease, suggesting a potential way to identify those who could benefit most from this therapy.
“If we can identify which patients carry the harmful bacterial function, we could, in the future, intervene with a targeted therapy designed specifically to turn down that activity,” says Verdu. “This is what personalized medicine should look like: matching the right biological tool to the right patient,” added Hosseinidoust.
Next steps for the team include evaluating broader collections of bacterial strains from IBD patients and developing combinations of phages—work that brings the approach closer to human trials.
The post Crohn’s Disease Phage Therapy Neutralizes Inflammatory <i>E. coli</i> in Mouse Model appeared first on GEN – Genetic Engineering and Biotechnology News.
Feasibility and Acceptability of Automated Texts to Offer, Screen, and Enroll Patients in a Cancer Clinical Trial Financial Reimbursement Program: Mixed Methods Study
Background: Out-of-pocket (OOP) costs pose a significant barrier to participating in cancer clinical trials (CCTs). Financial reimbursement programs (FRPs) that reduce the burden of OOP costs can support participation in CCTs if the information is readily available to participants at the time of enrollment. Prior studies have shown the importance and impact of FRPs, but despite improvements, significant barriers still remain. Objective: This study was designed to explore the feasibility and acceptability of automated texts designed to offer, screen, and enroll CCT participants in an FRP for OOP travel and lodging-related clinical trial costs. Methods: This study used a mixed methods approach. Eligible participants were those who consented to participate in a breast, leukemia, or chimeric antigen receptor T cell (CAR-T) trial at the Abramson Cancer Center of the University of Pennsylvania, a National Cancer Institute comprehensive cancer center. Quantitative data were collected through engagement metrics, including text response rates and enrollment rates, as well as patient-reported satisfaction scores. Qualitative data were derived from semistructured interviews. Program enrollment rates were used to determine feasibility, whereas the engagement metrics were used to measure the acceptability of the program. Semistructured interviews were conducted with a subsample of patients who responded to at least one of the FRP texts and agreed to be interviewed to determine the barriers to and facilitators of enrolling in the Improving Patient Access to Cancer Clinical Trials (IMPACT) program via text, perceived advantages and disadvantages of the text messaging program compared to a phone call, and overall feedback on the acceptability of the automated text messaging program. Results: Quantitative data, including engagement with texts, FRP eligibility screening, and enrollment rates, were collected from all participants who successfully received a text (n=51), and qualitative data were collected from a subsample of participants who agreed to participate in a semistructured interview (n=28) about the text-based program. Participants’ mean age was 58 (SD 12) years, approximately 65% (n=33) of participants were female, 21% (n=11) of participants were Black, and 4% (n=2) of participants were Hispanic or Latino. There was high engagement with texts (n=49, 96.1%) and a high screening rate for FRP eligibility (n=33, 64.7%). Of those who successfully screened, 26 (51%) screened via text. We also saw high overall FRP enrollment rates of those who completed the texts (n=16 of 24 eligible, 66.7%) and high satisfaction (Net Promoter Score=51). The text-based platform streamlined the enrollment process, allowing one-third of patients to complete enrollment independently, without assistance from the FRP coordinator. Reported facilitators for completion of the text conversation included support from the coordinator and introduction of the FRP by CCT teams. Barriers were a lack of communication from CCT teams, patient skepticism about the legitimacy of the texts, and limited program information via text. Conclusions: Despite the small sample size and single study site, these findings suggest that automated text messaging can be an effective, low-cost, and scalable strategy to increase awareness and streamline enrollment in FRPs.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/3fd9b4bb7c90efe704d0083d5f382a85" />
Engineered AAVs Harness Glymphatic System to Reach Brain Targets in Mice
Crossing the blood-brain barrier and avoiding off-target effects are two important challenges for gene therapies designed to target diseases of the brain. Now, developers of a gene therapy delivery platform that pairs specially engineered adeno-associated viruses (AAVs) with a delivery strategy that harnesses the brain’s fluid transport pathways claim that their approach addresses both issues and could help pave the way to new treatments for neurological disorders like multiple sclerosis, Huntington’s disease, and rare pediatric white matter disorders.
Details of the work were published recently in a Nature Biotechnology paper titled “Efficient targeting of human glial progenitor cells in vivo with engineered AAV vectors and glymphatic delivery.” The research was done by scientists at University of Rochester Medicine and the University of Copenhagen.
While the platform can deliver therapeutic genes broadly throughout the brain, it preferentially targets human glial cells. Steve Goldman, MD, PhD, lead author of the study and co-director of the Center for Translational Neuromedicine at URochester, has spent his career studying glial cells and elucidating their role in disease progression and recovery. Previously, his lab developed human glial progenitor cell models and investigated the link between glial dysfunction and neurological disease. For example, in Huntington’s disease, his lab has shown that healthy human glial progenitor cells could outcompete and replace diseased cells in the brain.
“Over the last decade, we’ve learned that many neurological disorders involve glial dysfunction as a major driver of disease,” Goldman said. “That realization has created an urgent need for tools that can safely and efficiently deliver therapies to these cells throughout the brain.”
The current study gets scientists one step closer to that goal. Digging into the details, Goldman and his colleagues engineered a library of modified AAV5 viral vectors by making small changes to the vectors’ capsids. They then screened the vectors in mice whose brains were transplanted with human glial progenitor cells and tracked their movements to identify which ones most effectively infected the human glial progenitor cells and their descendants including astrocytes and oligodendrocytes.
“Human cells display different molecular signatures than mouse cells, and cells behave differently in the brain than they do in a dish,” Goldman explained. “By selecting vectors under biologically relevant conditions, we were able to identify candidates with a strong preference for human glia.”
Next, the team turned their efforts to studying how best to distribute the AAVs throughout the brain. For that, they turned to the glymphatic system, the brain’s network of fluid-filled pathways used to clear metabolic waste by circulating cerebrospinal fluid through the brain. They delivered the engineered AAVs into the cisterna magna, a fluid-filled compartment at the base of the brain, while using hypertonic treatment to enhance fluid uptake into the network. This approach spread the vectors broadly throughout the brain tissue while largely avoiding the blood-brain barrier, and reducing exposure to peripheral organs like the liver.
“The glymphatic system is changing the way we think about brain drug delivery,” Goldman said. “Rather than trying to force therapies across the blood-brain barrier from the bloodstream, we can use the brain’s own transport pathways to distribute them more effectively where they are needed.”
Immediate targets for this approach are pediatric lysosomal storage diseases and other inherited disorders in which glial cells lack critical enzymes. Essentially, diseases of the brain’s white matter with well-defined biological targets. Further down the road, the approach could support novel therapies for multiple sclerosis, age-related white matter loss, and Huntington’s, among other neurodegenerative disorders where glial dysfunction is involved.
“We envision a future in which vectors can be designed for specific diseases and specific cell populations,” Goldman said. His lab is already exploring whether they can use artificial intelligence to design viral capsids that have specific targeting characteristics. “This study shows that by combining targeted vector engineering with glymphatic delivery, we can begin to build that future.”
The post Engineered AAVs Harness Glymphatic System to Reach Brain Targets in Mice appeared first on GEN – Genetic Engineering and Biotechnology News.
AI Tool Speeds Biomedical Research Progress
A multi-skilled biomedical AI agent can automate biomedical research tasks, driving forward basic research into practical application.
The AI-powered “co-scientist” Biomni autonomously executes diverse research tasks, working across complex tasks across fields as diverse as genomics, immunology, pharmacology, and clinical medicine.
The large-language model (LLM), outlined in Science, mines through biomedical literature interpreting requests, both composing and executing multi-step workflows.
It can formulate hypotheses, performs complex bioinformatics analyses, and design rigorous experimental protocols, with tests showing comparable accuracy to experts while taking a fraction of the speed.
“Biomni is able to understand a simple question like, ‘Why are these patients responding differently to the drug?’” explained research Kexin Huang, who was studying for his PhD at Stanford University at the time of the research.
“Then it digs in, doing a lot of the scientific legwork.”
Initially, the researchers constructed a unified and comprehensive biomedical action space by systematically analyzing 2500 biomedical research papers spanning 25 distinct subfields, curated from literature repositories.
From this foundation, they developed an LLM-powered action discovery agent capable of reading papers and extracting key tasks, tools, and databases essential to driving biomedical discoveries.
These elements are then chosen and developed into Biomni-E1, the foundational environment that defines the biomedical action space for agentic interaction and includes 150 specialized biomedical tools, 105 software packages, and 59 databases.
The team then designed Biomni-A1, a general-purpose agent architecture capable of flexibly executing a broad spectrum of biomedical tasks by using tools and datasets provided by Biomni E1.
After a user query is entered, the agent uses a retrieval system to identify the most relevant tools, databases, and software needed.
It then applies LLM-based reasoning and domain expertise to generate a detailed, step-by-step plan. Each step is expressed through executable code, enabling precise and flexible compositions of biomedical actions—an essential 10 feature given the domain’s reliance on highly specialized tools and data resources.
This integrated system allows Biomni to efficiently generate solutions for challenging, large-scale biomedical problems, but also to generalize to tasks across previously unseen areas of biomedical research.
In this way it removes the laborious work from biomedical science, allowing researchers to focus on creating hypotheses and innovative experiments and collaborating across disciplines.
“The hurdle in biomedical science is not intelligence or ideas; it is mechanics,” said researcher Jure Leskovec, PhD, also at Stanford.
“It’s this laborious stuff that slows innovation. Biomni can do this work in minutes.”
The team tested Biomni’s practical capabilities through five case studies: analyzing wearable sensor data; performing comprehensive bioinformatics analyses on massive raw datasets such as single-cell RNA-seq and ATAC-seq data; designing laboratory protocols to assist wet-lab researchers; optimizing a protein sequence for better thermostability; and orchestrating robotics wet-lab instruments.
In one test, more than 450 files of real-world continuous glucose monitoring, food intake, and physical activity data from a single person was uploaded and Biomni asked to find interesting and plausible hypotheses.
The researchers asked a simple question: “Analyze this data, find interesting and plausible hypotheses.” In just 40 minutes, the AI-agent identified patterns relating food intake and body temperature that would have taken an estimated 60 or more hours for a human to complete.
“With Biomni, we introduce a scalable, general-purpose biomedical AI agent, pointing toward a future in which AI agents work alongside human researchers to accelerate biomedical discovery from basic research to translation,” the authors concluded.
A prototype of the AI agent is already being used by more than 10,000 labs in academia and industry, making it the most widely used AI co-scientist system in biomedicine.
The post AI Tool Speeds Biomedical Research Progress appeared first on Inside Precision Medicine.
Clinicians Trust Faulty AI Recommendations Over Experience
Reliance of artificial intelligence (AI) has been increasing across all fields, including in the clinic. How to ensure AI is integrated practically, effectively, and ethically has been the focus of many discussions and debates, however final consensus almost always lands on the idea that however AI is integrated into clinical use, conscientious human oversight is necessary.
To test the veracity of this goal, a team of researchers in Spain put over 200 medical doctors to the test to determine if the clinicians would trust their experience and training over AI recommendations.
“It is important to investigate the errors that humans (including doctors) make when working with algorithms, in order to learn how to minimize the problems that arise from them,” said co-author of the study, Fernando Blanco, PhD, researcher at the Mind, Brain and Behavior Research Center (CIMCYC) in Granada, Spain.
“We wanted to see if professional physicians would act differently so that they would notice and correct these errors,” the authors wrote in their paper published in PLOS Digital Health.
The researchers created treatment plan options for a series of fictitious patients with a rare disease. They asked 223 physician participants whether or not to provide a treatment to a patient based on whether the patient was classified by AI as being highly or lowly sensitive to the treatment plan. The physicians were then presented with patient recovery data and asked to rate their perception on how reliable the AI classification was.
In these experiments, both groups of patients responded to treatment with similar sensitivity, resulting in ineffective AI recommendations that could be identified using the patient recovery data.
“In the first experiment, the treatment worked moderately (and equally) well for both groups. In the second, the treatment did not work at all for either group,” the authors wrote. They expected that “in both experiments, participants would administer the treatment less often to the fictitious patients classified as lowly sensitive to the treatment, consistently with the AI classification.” However, this was not the case.
“In both experiments, physicians mostly trusted the AI’s classifications and had trouble learning from the feedback,” said lead author Aranzazu Vinas, PhD, University of the Basque Country, Spain. “Furthermore, in the second experiment, professionals did not notice that the treatment was completely ineffective.”
These results present a major concern for the medical community in its use of AI in the clinic. While AI can be highly effective and useful for data collection and summary, supervision and critical thinking are still required for effective and safe patient care. This study highlights the need for physicians to take the time to critically consider all available data, regardless of recommendations by AI in their diagnostic and treatment decisions.
“People tend to say that there is always a human controlling the algorithm,” opined Helena Matute, PhD, professor, University of Deusto, Spain on the team’s findings “but our experiments show that doctors (as well as anyone else) have problems in learning from the available evidence when it contradicts the suggestions of an algorithm.”
The post Clinicians Trust Faulty AI Recommendations Over Experience appeared first on Inside Precision Medicine.
Blended Genome–Exome Sequencing Slashes Costs Without Quality Loss
A novel sequencing strategy that combines low-pass whole-genome sequencing with deep whole-exome sequencing in a single assay could significantly lower the cost of large-scale genomic studies without sacrificing analytical performance, according to a study published in Nature Genetics. The approach, known as blended genome–exome (BGE) sequencing, may help accelerate precision medicine initiatives by making comprehensive genomic profiling more accessible across diverse populations.
Researchers from Massachusetts General Hospital and the Broad Institute of MIT and Harvard developed BGE to overcome a persistent tradeoff in human genomics. High-coverage whole-genome sequencing offers the most comprehensive view of genetic variation but remains prohibitively expensive for many population studies. Genotyping arrays and exome sequencing are more affordable but either miss large portions of the genome or introduce bias by relying on variants selected primarily from European ancestry populations.
The BGE workflow integrates low-pass whole-genome sequencing at 1–4x coverage with deep exome sequencing at 30–40x coverage within a single library preparation and sequencing run on Illumina’s NovaSeqS4. The result is a unified dataset capable of supporting genome-wide association studies, rare variant discovery, copy number variant (CNV) detection, and polygenic analyses at approximately 28% of the cost of conventional 30x whole-genome sequencing.
The investigators validated the approach in more than 53,000 participants enrolled in the Populations Underrepresented in Mental Illness Associations Studies (PUMAS) Project, which includes African, African American, Hispanic/Latino, and Colombian cohorts. The scale and diversity of the study allowed the researchers to assess performance in populations that have historically been underrepresented in genomic research.
Imputed genotypes generated from BGE showed excellent agreement with Illumina Global Screening Array data, achieving concordance exceeding 95% for variants with minor allele frequencies above 1%. Importantly, performance remained consistent across multiple ancestry groups and local ancestry backgrounds, addressing one of the major limitations of conventional array-based genotyping.
The platform also demonstrated strong performance for clinically relevant structural variation. Using established computational pipelines, investigators achieved approximately 90% positive predictive value for protein-coding CNVs spanning three or more exons compared with deep whole-genome sequencing. In benchmarking studies, the method successfully detected all validated de novo coding CNVs in a reference autism cohort while maintaining low false-positive rates.
Beyond analytical performance, the study highlights potential operational advantages. By combining genome and exome sequencing into a single workflow, BGE simplifies laboratory processing, reduces the need for multiple assays, and minimizes sample attrition between sequencing platforms. These efficiencies could prove valuable for national biobanks, health system sequencing programs, and pharmaceutical research efforts that increasingly require genomic datasets from hundreds of thousands of participants.
The technology may also advance equity in precision medicine. Because low-pass genome sequencing does not depend on predefined variant content, it avoids many of the ascertainment biases associated with traditional genotyping arrays. The authors found that BGE captured substantially more coding and noncoding variants than array-based approaches while maintaining high-quality rare variant detection through deep exome coverage.
The researchers acknowledge that imputation performance remains influenced by the diversity of available reference panels, particularly for Indigenous American ancestry. However, as more globally representative reference datasets become available, they expect the accuracy of low-pass genome imputation to improve further.
As precision medicine increasingly depends on large, ancestrally diverse genomic datasets, technologies that balance cost, scalability, and comprehensive variant detection will be essential. BGE sequencing offers a practical alternative to deep whole-genome sequencing, enabling broader participation in genomic discovery while preserving much of the analytical power needed to identify clinically meaningful genetic variation.
The post Blended Genome–Exome Sequencing Slashes Costs Without Quality Loss appeared first on Inside Precision Medicine.