Scientists create unprecedented atlas of the human epigenome; understand the Salk Institute discovery

Scientists at the Salk Institute, in collaboration with the Arc Institute and other researchers, have revealed a significant advance in understanding the human genome. They completed the first comprehensive single-cell-scale atlas for two important epigenetic systems: the three-dimensional organization of the genome and DNA methylation.
Published in the journal Science on July 23, 2026, the study provides a detailed look at how the body’s cells use the same three-billion-letter genetic code to perform distinct functions. This process is orchestrated by epigenetics, which involves chemical markers and structures that control the activation or silencing of genes.
The atlas maps 86,689 cells, originating from 16 different human tissues, identifying 35 main cell types and 206 subtypes. The resource is freely accessible online and is part of the National Institutes of Health’s 4D Nucleome (NIH 4DN) program, which seeks to elucidate the spatial and temporal organization of the genome in the regulation of gene expression.
Salk’s research was published simultaneously with five other papers from the NIH 4DN program in Science and three in Science Advances, demonstrating the collaborative importance of the initiative. Simultaneously measuring the two epigenetic layers allowed scientists to compare the information each provides about cellular identity.
The importance of mapping the epigenome cell by cell
The Human Genome Project, completed in 2003, produced the linear DNA sequence. However, this sequence alone does not explain the cellular diversity of the body. This information is contained in the epigenome, through chemical modifications and structural folding that act as switches for genes, specifically for each cell type.
Two of the most important epigenetic characteristics are: DNA methylation, which consists of the addition of methyl groups to specific DNA bases, and the 3D organization of the genome, with loops, folds and compartments that bring distant regions of the DNA together. Although both influence gene expression, they have never been measured together in individual cells throughout the human body.
Joseph Ecker, professor and chair of the Salk International Council on Genetics at Salk and a Howard Hughes Medical Institute investigator, said there is an interest in understanding the organization of the genome at the individual cell level. He explained that this could provide a better understanding of how genetic variants impact disease, as some cell types may be more susceptible due to genomic organization.
Understand the link between non-coding DNA and the development of diseases
Most genetic variants linked to disease are located in the noncoding regions of the genome, a vast portion of DNA that does not produce proteins. This particularity has made it difficult to identify how a variant contributes to the disease, in which type of cell it acts and which genes are affected.
The new atlas offers a straightforward approach to solving this challenge. The team identified more than 1.36 million regions with differential methylation and 283,606 differential chromatin loops across cell types in the human body, using tissues from the heart, brain, lungs, stomach, skin and other organs.
By crossing this data with genetic variants known to increase disease risk, specific associations emerged. Variants linked to blood glucose regulation, for example, were concentrated in endocrine cells, while atrial fibrillation variants were found in cardiac muscle cells. Baldness variants appeared in skin fibroblasts, and bipolar disorder and schizophrenia variants in excitatory and inhibitory neurons.
Jesse Dixon, associate professor and Helen McLoraine Developmental Chair at Salk, said that much of the genetic variation that predisposes to disease is in noncoding parts of the genome. He added that adding the 3D genome aspect can fill this gap, connecting non-coding variations to the genes they affect in specific cells and tissues.
Surprising discoveries about the divergence between epigenetic systems
One of the study’s most surprising revelations is that DNA methylation and 3D genomic structure do not always provide the same information about a cell. In skeletal muscle samples, the team observed fibers that, due to their 3D organization, looked like mature, differentiated muscle cells, but still showed the methylation signature of muscle stem cells. The reverse rarely occurred. The most likely explanation, according to the researchers, is that these cells would be in an intermediate stage of differentiation, with the 3D architecture updating first, and methylation adjusting later.
Similar disagreements were noted in Schwann cells of the peripheral nervous system and in placental trophoblasts. This pattern suggests that different epigenetic features update on distinct time scales during cellular state transitions. Such a discovery could redefine how researchers classify “cell types” in adult tissues and how they track the movement of cells between disease states.
The atlas also revises a long-held belief about “non-CG methylation,” an unusual form of methylation previously believed to be restricted to brain cells and stem cells. The study demonstrated that it carries cell identity information in several human tissues, including muscle, pancreas and immune cell types, at smaller but biologically relevant levels.
Jingtian Zhou, co-first and co-corresponding author, former graduate researcher in Ecker’s lab and now leader of his own lab at Arc Institute, said the inconsistency between modalities can be used to determine which cell populations are alternating in adult tissues and diseases. He suggested this could, for example, expand understanding of cancer cell dynamics.
A global resource for scientific research and artificial intelligence tools
To ensure broad usability of the atlas, the team developed an interactive web browser that allows researchers to visualize DNA methylation and 3D chromatin contacts across all tissues, cell types and subtypes studied. The underlying data, which includes 195 billion methylation measurements and 18 billion chromatin contacts, is freely available.
This feature comes at a time when artificial intelligence tools are increasingly used to predict the functional impact of genetic variants. Atlases like this provide the cell-type-resolved, labeled training data that models need to make accurate predictions—a bottleneck that has historically limited the field.
In a companion article in the same issue of Science, a study led by Bing Ren of the New York Genome Center and Columbia University used inter-tissue methylation data from the atlas. Ren’s research has shown that a substantial portion of microglia, the brain’s resident immune cells, are replaced by cells that resemble blood monocytes between the ages of approximately 50 and 75 years. This finding challenges the long-held view that microglia persist from embryonic development throughout life.
Bing Ren, who also co-authored the Salk-led study, said DNA methylation patterns are specific to each cell type and analogous to a cellular barcode. He explained that comprehensive atlases of DNA methylation across tissues show that aged microglia in the human hippocampus more closely resemble peripheral blood monocytes than microglia from young adults, providing a crucial clue to the biological identity of these cells.
The NIH 4D Nucleome consortium, of which this study is part, aims to expand this type of mapping to the fourth dimension: time. A 4D understanding of the genome as its structure and chemistry changes as cells develop, age and respond to disease remains a primary goal. The inter-tissue atlas provides a reference framework that will serve as a basis for future time course studies.






