Radically different
Category Added in a WPeMatico Campaign
Scientists at the Salk Institute and the Arc Institute, along with their collaborators, unveiled the first body-wide single-cell atlas of two major epigenetic systems: three-dimensional genome folding and DNA methylation, measured simultaneously in the same cells.
The atlas spans 86,689 cells from 16 human tissues, revealing 35 major cell types and 206 subtypes, and is freely available online. The work is part of the National Institutes of Health’s 4D Nucleome (NIH 4DN) program, which aims to understand how the genome is organized in space and time to regulate gene expression in health and disease.
Because the two epigenetic layers were measured together, the researchers could compare what each layer says about a cell’s identity. And while often the two pictures agree, they found that sometimes they do not.
The Salk paper “Human body single-cell atlas of 3D genome organization and DNA methylation” was published alongside five other NIH 4DN papers in Science, and three others in Science Advances.
The Human Genome Project, completed in 2003, produced a linear read of the three billion DNA letters in the human body. But the letters alone don’t explain how a single genome produces hundreds of different cell types. That information lives in the epigenome in the form of chemical modifications and structural folds layered on top of the DNA sequence, where they can switch genes “on” and “off” in patterns specific to each cell type.
![Caption: Salk scientists Jingtian Zhou (left), Jesse Dixon (center), and Joseph Ecker (right) profiled 86,689 cells across 16 human tissues, linking cell-type-specific epigenetic features to disease risk and revealing that a cell’s 3D genome and DNA methylation don’t always tell the same story. [Salk Institute]](https://www.genengnews.com/wp-content/uploads/2026/07/260723-pr-ecker-dixon-authors-232x300.jpg)
Two of the most consequential epigenetic features are 1) DNA methylation, where small chemical groups called methyl groups are attached to specific DNA bases, and 2) 3D genome organization, where intricate loops, folds, and compartments bring distant stretches of DNA into contact. Both influence gene expression, but they had never been measured together in single cells across the human body.
“There has been an appreciation for trying to understand, at the individual cell level, how the genome is organized, so that we can get a better idea of how genetic variants impact disease,” said co-corresponding author Joseph Ecker, PhD, a professor and Salk International Council Chair in Genetics at Salk and a Howard Hughes Medical Institute investigator. “Some cell types may be more vulnerable than others to genetic variants, because the genome is organized differently in different cell types—and whether a variant matters can depend on that organization.”
Why is noncoding DNA relevant in disease?
Most disease-associated genetic variants fall in the noncoding regions of the genome. That has made it difficult to figure out how a variant contributes to disease, which cell type it acts in, and what gene it ultimately affects.
The new atlas identifies more than 1.36 million differentially methylated regions and 283,606 differential chromatin loops across the human body’s cell types, using tissues from the heart, brain, lungs, stomach, skin, and more. When the researchers overlaid genetic variants known to raise disease risk, specific pairings emerged like variants for blood-glucose regulation concentrated in endocrine cells, atrial fibrillation variants in heart muscle cells, balding variants in skin fibroblasts, and bipolar disorder and schizophrenia variants in excitatory and inhibitory neurons.
“A lot of the genetic variation that predisposes someone to disease is in noncoding parts of the genome,” said co-corresponding author Jesse Dixon, MD, PhD, associate professor and Helen McLoraine Developmental Chair at Salk. “By adding in the 3D genome aspect, we can potentially bridge that gap—connecting noncoding variations with the genes they affect in specific cells and tissues.”
What happens when two epigenetic lenses disagree?
One of the study’s most surprising findings is that DNA methylation and 3D genome structure don’t always tell the same story about a cell. In skeletal muscle, the team found fibers that look like mature, differentiated muscle cells by their 3D genome folding, but still carry the methylation signature of muscle stem cells. The reverse almost never happens. The most plausible explanation, they explained, is that these cells are caught mid-differentiation, with 3D architecture updating first and methylation catching up.
Similar mismatches appeared in Schwann cells of the peripheral nervous system and in placental trophoblasts. The pattern suggests that different epigenetic features update on different time scales during cell state transitions—a finding that could reshape how researchers define “cell type” in adult tissues and how they track cells moving between states in disease.
The atlas also revises a long-standing assumption about “non-CG methylation,” an unusual form of methylation previously thought to be largely confined to brain cells and stem cells. The study shows that it carries cell-identity information across many human tissues, including muscle, pancreas, and immune cell types, at lower but biologically meaningful levels.
“The inconsistency between modalities may be further used to determine what cell populations are switching between each other in adult tissues and diseases, which could, for example, expand our understanding of cancer cell dynamics,” said co-first and co-corresponding author Jingtian Zhou, PhD, a former graduate researcher in Ecker’s lab who now leads his own lab at the Arc Institute.
A public resource for scientists and artificial intelligence
To make the atlas broadly usable, the team built an interactive web browser that lets researchers visualize DNA methylation and 3D chromatin contacts across every tissue, cell type, and subtype in the study. The underlying data, including 195 billion methylation measurements and 18 billion chromatin contacts, are freely available.
The resource arrives as artificial intelligence tools are increasingly used to predict the functional impact of genetic variants. Atlases like this one can provide the labeled, cell-type-resolved training data that models need to make accurate predictions—a bottleneck that has historically limited the field.
For example, in a companion paper in the same issue of Science, a study led by Bing Ren, PhD, from the New York Genome Center and Columbia University used the atlas’ cross-tissue methylation data to show that a substantial fraction of the brain’s resident immune cells, called microglia, are replaced by cells resembling blood monocytes between roughly ages 50 and 75. The finding challenges the long-held view that microglia persist from embryonic development throughout the life span.
“DNA methylation patterns are specific to each cell type and analogous to a cellular barcode,” said Ren, who also co-authored the Salk-led study. “The comprehensive cross-tissue DNA methylation atlases show that the aging microglia in the human hippocampus more closely match the monocytes from peripheral blood than microglia from young adults, providing a crucial clue for the biological identity of these cells.”
The NIH 4D Nucleome consortium, of which this study is a part, aims to extend this kind of mapping into the fourth dimension: time. A 4D understanding of the genome—how its structure and chemistry change as cells develop, age, and respond to disease—remains a major goal, and the cross-tissue atlas provides reference scaffolding that future time-course studies will build on.
Along with scientists from the Salk Institute and Arc Institute, investigators from the University of California, San Diego, Columbia University, New York Genome Center, University of California, Los Angeles, Harvard, Henan University in China, Vanderbilt University, Stanford University, Broad Institute, University of Sheffield in the U.K., Yale, University of Florida, University of Freiburg in Germany, University of Graz in Austria, and Nanchang University in China; and Chongyuan Luo also contributed to the Science paper.
The post Single-Cell Atlas Simultaneously Maps 3D Genome Architecture and DNA Methylation appeared first on GEN – Genetic Engineering and Biotechnology News.
Designing and developing a new medicine is an expensive, failure-prone scientific challenge. A new drug can take many years to develop, at the cost of a significant investment. And even then, most possible candidates never reach the patient. For biologic medicines, therapies made from engineered proteins rather than synthetic chemistry (which are often used to treat conditions across most major acute and chronic diseases), the complexity is even greater.
Scientists explore vast quantities of possible molecules, looking for the rare few that will bind to the right target, remain stable in the human body, and be manufacturable at scale. Today, AI is speeding up these processes and has quickly become a core part of the infrastructure in pharmaceutical R&D.
AI-assisted design is a growing part of how biologic drug candidates are developed, and companies like AstraZeneca are actively building its engineering teams to push this further. “Everything we do, whether it’s design, make, test, or analyze, is now computationally enhanced,” says Puja Sapra, senior vice president and head of R&D biologics engineering and oncology targeted discovery at AstraZeneca. “The cycle times are getting shorter while productivity and innovation increase.”
Sapra explains that AstraZeneca’s approach follows a build-measure-learn loop. AI generates or prioritizes candidate molecules computationally, predicting which designs are most likely to succeed. Scientists then focus lab resources only on the top-ranked candidates. This leads to a tighter feedback cycle with fewer dead ends, faster iteration, and the ability to go after disease targets that were previously considered untreatable by medicine. Because the number of possible molecular combinations far exceeds what any human team can systematically explore, using AI to narrow and refine the options for testing has become a major focus in biologics drug design.
Beyond accelerating timelines, AI is also being applied to the discovery of entirely new classes of medicines. Traditional biologics typically target one disease pathway. The next generation of drugs can hit multiple targets simultaneously or precisely deliver therapeutic payloads to specific cells. Achieving this requires optimization across many variables at once. Looking ahead AI-driven models could help design these increasingly complex, multi-specific biologics, explains Puja Sapra. “For example,” she continues, “such models could help identify which two or three targets to prioritize based on the underlying biology, then optimize across multiple parameters to balance a molecule’s potency, stability, manufacturability, and safety.” “Drugging the undruggable is becoming a reality,” Sapra says. “These technologies will eventually enable us to develop medicines against targets once thought impossible to reach. The potential for benefit to patients is remarkable.”
McKinsey estimates that generative AI, combined with other computational tools, could cut drug discovery timelines by as much as 50%. But every AI model is only as good as its training data. In drug discovery, that means ample quantities of high-quality biological data. Experiments can provide a rich source of such data. Whether they succeed or fail, each experiment generates a signal about what does and does not work.
“Data is our differentiator,” says Sapra, explaining how the company’s datasets are proprietary and multimodal and include molecular structures, binding measurements, safety profiles, and manufacturing outcomes. “We’ve built an intentionally diverse portfolio across multiple disease areas and drug types. All of that data empowers us to fine-tune frontier AI models with richer, more representative training sets.” She continues, “Further, we have invested in deep screening technologies to generate additional datasets required in volume to constantly refine and validate our models.”
To bring all of that data together in one place, AstraZeneca is building what it calls a “lab of the future” facility in Kendall Square, Cambridge, Massachusetts where AI and robotic automation will be able to form a continuous, closed-loop discovery system. “Where a self-driving car uses sensors and models to navigate its environment, this system uses AI to make predictions, robotic systems to execute experiments, and instruments to generate data,” explains Sapra. That data feeds directly back into the models, accelerating each subsequent cycle.
“Throughout, scientists will remain central to the process, providing the oversight, judgement, and strategic direction that ensure outputs are explainable, tolerable, and directed toward potential patient benefit,” she adds.
Eventually, automated high-throughput systems will be able to make and evaluate thousands of molecular interactions on a weekly basis. “This will generate AI-ready data at a scale that traditional workflows cannot match,” Sapra says. “Robotic sample handling, automated quality checks, and integrated data pipelines also have the potential to help accelerate early drug development timelines significantly.”
Ultimately, Sapra says, the end-state vision for AI in biologic drug discovery is what the field calls “de novo” design. For this, the goal is for AI to generate entirely new protein sequences that precisely fit the desired drug properties. This includes designing the structure, predicting safety, how it will behave in the body and how to make it manufacturable.
“The field is making great progress toward a completely AI-generated biologic, designed from scratch all the way to a clinical candidate,” Sapra says. “As we continue to leverage frontier models and fine-tune them with the right datasets, we bring ourselves closer to this reality. I believe it will come. It’s a matter of time.”
Several key elements are needed to reach this point, however. First is richer and more standardized training data across the industry. Second, robust evaluation benchmarks for AI-generated candidates. And third, teams that know how to work at the intersection of machine learning and biology. Of all the prerequisites, however, safety prediction may be the most consequential, and perhaps the least discussed, Sapra says.
“One of the hardest problems in de novo design is predicting whether a computationally generated molecule will be safe in the human body,” Sapra explains. AstraZeneca is tackling this with what amounts to virtual clinical trials. These are advanced cell systems and micro-scale organ models that function as physical testbeds, paired with AI that learns from their outputs.
“These systems have the potential to generate enhanced biological signals without traditional testing bottlenecks, and they’re a critical missing piece in closing the loop between AI-generated designs and clinical-ready candidates,” Sapra adds.
A shift currently underway is the move toward agentic AI systems that can simultaneously generate molecule candidates and predict how efficacious and safe they are likely to be. These autonomous workflows can connect disease-level insights directly to molecule design, bridging what were previously separate data silos. “The complexity of the biology goes hand-in-hand with the design of the molecule,” summarizes Sapra.
The transformation underway in biologics is not just about technology. “With more autonomous systems, human oversight remains at the heart of this approach—ensuring explainable and ethical AI for the benefit of patients,” says Sapra.
For scientists, working with AI is a collaborative process. “Scientists will work hand-in-hand with these model systems,” she says. “There will be a world where models will design molecules, then scientists will work with the systems to test those molecules and put all that data together.” Through this process of human checks, balances, and judgement calls, the models will evolve and constantly improve, ultimately with potential to benefit patients.
For engineers, designing and building effective systems ready for human-AI collaboration will mean ensuring high levels of model transparency and explainability. According to Sapra, AstraZeneca’s engineering teams include data scientists, automation specialists, and AI engineers, who are developing systems that act as “thinking partners” rather than black boxes. “Engineers are designing systems that generate, validate, and learn at speed. And the problems are genuinely hard: Multimodal data fusion, closed-loop optimization, uncertainty quantification, and interpretability at the point of clinical decision-making,” she adds.
In taking on such technically demanding challenges, engineers and scientists have the opportunity to contribute to the research and development of potentially life-changing treatments for many diseases, says Sapra. “The biologic medicines we can develop today, and those we’ll design tomorrow, depend on combining world-class AI and engineering talent with deep scientific expertise.”
This article has been initiated and funded by AstraZeneca. Z4-85058, July 2026.
This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. It was researched, designed, and written by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.
Nature Medicine, Published online: 23 July 2026; doi:10.1038/s41591-026-04563-8
Early diagnosis of Alzheimer’s disease is important to ensure timely and accurate treatment. We show that a blood-based circular RNA signature can accurately detect Alzheimer’s disease as well as, or better than, current biomarkers and even predict who will develop symptoms before they appear. This less invasive approach could offer earlier diagnosis and improved monitoring of Alzheimer’s disease.
While Alzheimer’s disease is the most common cause of dementia, many of the molecular mechanisms that drive its progression remain poorly understood. While researchers have cataloged changes in gene activity across different brain cell types, a key unanswered question has been how the genome’s 3D organization influences those changes. Now, researchers have linked alterations in genome folding to disrupted gene regulation in Alzheimer’s disease, providing a new layer of insight into the biology of neurodegeneration.
The findings, published in Science in the paper “Single-cell multiomics connects 3D genome and transcriptome alterations in Alzheimer’s disease,” were reported by researchers from Carnegie Mellon University’s School of Computer Science, the University of Pittsburgh School of Medicine, the University of Washington, and collaborating institutions. Using single-cell multiomics, spatial transcriptomics, and artificial intelligence (AI), the team generated a multiscale view connecting genome structure, gene expression, and tissue organization in Alzheimer’s disease.
To investigate the role of genome architecture in Alzheimer’s disease, the researchers analyzed postmortem prefrontal cortex tissue from individuals with and without the disease. They used GAGE-seq (genome architecture and gene expression by sequencing), a technique that measures both gene expression and physical genome contacts in the same single cell. The team combined those data with chromatin accessibility data, spatial transcriptomic maps, and a transformer-based AI model called Hicformer, which integrates DNA sequence and 3D genome features to predict cell-type-specific gene activity.
The study revealed widespread changes in chromatin organization across major brain cell types. According to the paper, Alzheimer’s disease was associated with “reduced short-range interactions and increased longer-range interactions” within the genome. Active and inactive genomic regions also exhibited increased mixing, consistent with weaker compartment segregation. The researchers linked these structural changes to cell type–specific alterations in gene expression programs involved in disease-relevant pathways.
Researchers also observed weakening of promoter-proximal interactions and changes in regulatory elements, alongside evidence of senescence-related activation in microglia and sex-dependent dysregulation of X-linked genes in females. Integrating the molecular data with spatial transcriptomics revealed altered cellular neighborhoods and disrupted coordination of gene programs within diseased brain tissue. The authors wrote that the results connect “genome structure, gene regulation, and tissue organization through a unified multimodal analysis.”
Their predictive model Hicformer also demonstrated that “3D genome features provide information beyond DNA sequence alone for explaining AD-relevant gene expression, enabling prioritization of distal regulatory elements whose effects are mediated through chromatin contacts,” the authors wrote.
“Measuring gene activity and genome folding in the same cell allows us to directly connect chromosome structure with disease-related gene programs,” said Yang Zhang, PhD, a project scientist in Carnegie Mellon’s Computational Biology Department and co-lead author. “Across several kinds of brain cells, this paired view revealed a consistent signature of 3D genome reorganization in Alzheimer’s disease and helped us prioritize regulatory regions for future mechanistic and therapeutic investigation.”
The researchers concluded that genome folding represents a previously underappreciated regulatory layer associated with Alzheimer’s pathology. By creating a detailed map linking 3D genome remodeling to gene expression and tissue organization, the study provides a framework for future experiments aimed at determining which structural changes contribute directly to disease progression. This may also provide clues to future therapeutic focuses.
“Alzheimer’s disease cannot be understood one layer at a time,” said senior author Jian Ma, PhD, the Ray and Stephanie Lane Professor of Computational Biology at Carnegie Mellon University. “The genome’s 3D structure is a fundamental regulatory layer that helps to connect DNA sequence to gene activity. By integrating genome folding, cell state, and tissue context, we can move beyond cataloging disease-associated changes toward understanding how they fit together and which mechanisms to test next,” said Ma. “Alzheimer’s disease cannot be understood one layer at a time.”
The post Single-Cell Maps Reveal Genome Reorganization in Alzheimer’s Brain Cells appeared first on GEN – Genetic Engineering and Biotechnology News.
Background: Diabetic foot ulcers (DFUs) are a leading cause of hospitalizations, amputations, and health care costs among individuals with diabetes, often due to delayed detection and treatment. Early identification of skin changes is critical for preventing ulcer progression, yet daily visual foot self-inspections are often limited by impaired mobility, visual deficits, and poor compliance. Although some foot-focused remote patient monitoring technologies are in use, most of these approaches do not provide standardized, full-color visual documentation, which can affect clinicians’ ability to interpret findings using familiar visual cues. Objective: This formative evaluation aimed to assess the usability, compliance, and feasibility of an in-home full-color imaging scale for the remote monitoring of patients with DFUs to inform the design of larger prospective studies. Methods: We conducted a retrospective formative feasibility evaluation analyzing 7 months of preexisting data from adults with a history of DFUs who had enrolled in an outpatient remote monitoring program. During this program, participants were provided with an internet-connected in-home scale that captured high-resolution, full-color scans of the plantar surface of the feet during daily use. Scans were securely transmitted to a HIPAA (Health Insurance Portability and Accountability Act)-compliant web portal for podiatrist review. Usability was assessed through compliance with daily scanning, patient and physician surveys, and descriptive review of clinical workflows. Results: A total of 21 participants aged 43 to 86 years (median 65, IQR 59-74; mean 65.3, SD 10.6 years) were retrospectively analyzed. Participants demonstrated high compliance and feasibility of home use, with device use on 82.5% (3638/4410) of possible days, a median use of 27 days (IQR 21-30) per 30-day period, and a mean use of 23 (SD 6.4) days per 30-day period. Participants reported high satisfaction and minimal burden associated with daily use, with most rating the device as physically easy to use (median 7.0, IQR 7-7; mean 6.9, SD 0.45), expressing willingness to use it daily (median 7.0, IQR 7-7; mean 6.9, SD 0.31), and indicating trust in the system (median 7.0, IQR 7-7; mean 6.9, SD 0.32) on a 7-point Likert scale (1=worst; 7=best). Clinician usability was supported by podiatrist review of 3295 scans, of which 3220 (97.7%) were rated as usable for clinical assessment. Three representative cases illustrated real-world device use and longitudinal image review, demonstrating the ability to visualize skin changes such as foreign body presence, new ulceration, and early skin breakdown. Conclusions: This retrospective formative evaluation demonstrates the preliminary usability and feasibility of an in-home, full-color imaging scale for daily remote visual monitoring of diabetic feet in a real-world outpatient setting. High compliance and usability in clinical workflows suggest that this approach may support early identification of foot-related concerns. Future prospective studies are warranted to evaluate the impact of this technology on clinical outcomes, including ulcer progression and health care use.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/62c36c98001a2405fd323033480cbaa8" />
In a win for peptide proponents and for health secretary Robert F. Kennedy Jr., an advisory panel to the Food and Drug Administration recommended on Thursday that compounding pharmacies be allowed to manufacture the drugs BPC-157, KPV, TB-500, and MOTS-c.
The FDA panel will vote on additional compounds on Friday.
The votes are not binding, and the FDA will ultimately decide whether to lift current restrictions on these peptides and make them more easily accessible to people who’ve been turning to the little-researched but highly popular drugs to improve their health. It’s unusual for the FDA to go against the recommendations of the Pharmacy Compounding Advisory Committee (PCAC), but it has happened at least once before.