Researchers have built a high-resolution brain-cell atlas that could do for neuroscience what the Human Genome Project did for genetics
By charting over 25 million individual brain cells across humans, macaques, and marmosets, a team of neuroscientists, computational biologists, and software engineers across 55 institutions, including Princeton produced a cellular roadmap of the basal ganglia, a deep-seated network of brain regions that aid movement, habit-formation, and decision-making.
"It's kind of like the Human Genome Project 1.0, or 0.1," said Fenna Krienen , assistant professor of neuroscience at Princeton University and a co-investigator in one of the founding consortia of the National Institutes of Health's BRAIN Initiative® Cell Atlas Network (BICAN), the network behind the atlas. "You had the first kind-of-complete thing, and then all the work that happened after that could map to that reference. That's what we were able to really achieve as part of BICAN."
The consortium chose the basal ganglia in part for its relevance to Parkinson's, Alzheimer's, and Huntington's diseases, Krienen said, and because the region spans parts of the brain with different developmental and evolutionary histories. "With one circuit, we can traverse those different axes," she said.
Now that the technical proof-of-principle is complete, the BICAN team will continue to map the rest of the brain across humans and other primates, in hopes of creating an entire brain-cell atlas for humans, macaques, and marmosets.
The collection of 27 papers was published November 12 across several Cell Press-affiliated journals.
The BICAN group is not unlike the basal ganglia itself: while the basal ganglia refers to a general region in the brain, it's really made up of a collection of smaller, interconnected subregions working together.
Similarly, BICAN is a coalition of consortia, with each consortium focusing on a different corner of the cellular brain puzzle.
Some groups focused on analyzing the diversity of brain cells in humans, others on how the developing brain changes in mice.
Krienen's lab was responsible for mapping the marmoset brain, as part of the Human and Mammalian Brain Atlas consortium led by Ed Lein and Hongkui Zeng at the Allen Institute.
Scientists know less about marmoset brain cells than mouse brain cells. But marmosets are primates, sharing a more recent common ancestor with humans. That makes them a key bridge for connecting findings in mice to people.
Marmosets also have a practical advantage: tiny brains. Whereas a macaque's brain weighs about as much as a roll of dimes, a marmoset's brain weighs only about 8 grams, roughly the same as a Sacagawea dollar coin. That's small enough to map in its entirety, while still being a primate brain that more closely reflects our own.
Krienen and her lab contributed two papers to the package of BICAN findings.
The main paper, led by Krienen's graduate student Shu (Daisy) Dan with collaborators at the Allen Institute, mapped the cells of the marmoset subcortex, the deep-brain territory that houses the basal ganglia, across three molecular layers: spatial transcriptomics; single-cell RNA sequencing; and ATAC sequencing.
Like toggling among street, satellite, and terrain views in Google Maps, each layer shows something different. Spatial transcriptomics shows where each cell sits and which of several hundred key genes it's using. Single-cell RNA sequencing reads out the activity of nearly all of a cell's genes, though not its location. And ATAC sequencing reveals which parts of a cell's DNA instruction manual are legible: since most DNA is tightly packed to save space, ATAC shows which "pages" are open and ready to be read.
"You're going to be able to zoom through, and for every region, for every cell type, you'll have representation," Krienen said. "We are going for depth and completeness."
The second paper, led by neuroscience graduate student Ning Wang, co-mentored by Krienen and fellow Princeton professor Jonathan Pillow , introduces PIANO, a deep-learning computational tool that helps stitch together data from millions of cells and multiple species.
"As BICAN datasets now scale to millions of cells and incorporate multiple species, we needed better approaches for data integration," Krienen said. "Ning used the cross-species basal ganglia dataset to test drive his algorithm, which is now being deployed across our consortium as we build the full whole-brain cross-species atlases."
Having the whole marmoset subcortex mapped gives researchers context the other species can't yet offer. If a cell type turns up in a human sample that seems to belong to the basal ganglia, the marmoset map can show whether it's genuinely new or simply came from a neighboring region during dissection.
"Without knowing what is around you, you don't necessarily know who you are," Krienen said.
That context is already paying off. The most extensive human brain atlas to date contained a catch-all group of neurons nicknamed "Splatter," drawn from many deep-brain regions sampled too sparsely to tell apart. Using the marmoset map, Dan and her colleagues sorted it into hundreds of distinct cell types, each with a specific home in the brain and most shared across species.
The larger goal of BICAN, though, is to map the entire brain, and for the marmoset, Krienen's team is nearly there. The lab has already collected data across the whole marmoset brain, though the new paper focuses on the subcortex. Because the marmoset brain is so small, "we actually could go a bit ahead of schedule," Krienen said.
Researchers from the BICAN community now aspire to apply the work in the future to help analyze how known human genetic risk factors for neuropsychiatric and neurodegenerative disease map onto other primate cell types, such as the marmoset's.
For now, though, Krienen and the BICAN community have established a unified reference system, which is one of the high-level goals of the project. Different laboratories often use different names for the same cell, or the same name for different cells. The community's work provides a common vocabulary for classifying cells across the brain, and sets the stage for mapping the rest of the brain using the same methods.
"What we have from our corner of this ecosystem is basically the map, the representative, the atlas, the template, the reference for these organisms that every study that comes after us can map to," Krienen said.
Shu Dan, Meghan A. Turner, Michael DeBerardine, Lakme Caceres, Victor E. Nieto-Caballero, Jessica Schembri, Kirsten Levandowski, Christopher Cardenas, Ning Wang, Chang Li, Soumyadeep Basu, Hongyu Zheng, Xinhao Liu, Thomas Höllt, Jessica Feliciano, Qiangge Zhang, Reilly Nakamoto, Delissa A. McMillen, Naomi Martin, Nasmil Valera Cuevas, Paul Olsen, Josh Nagra, Jazmin Campos, Marshall M. VanNess, Jack Waters, Shea Ransford, Zoe Juneau, Sam Hastings, Stuard Barta, Augustin Ruiz, Jeanelle Ariza, Benjamin J. Raphael, Hongkui Zeng, Ed S. Lein, Boudewijn Lelieveldt, Guoping Feng, Brian Long, Fenna M. Krienen. Spatial patterning of transcriptional and regulatory programs in the primate subcortex . Cell. November 12, 2026. DOI: 10.1016/j.cell.2026.10.007
Ning Wang, Christopher Cardenas, Victor E. Nieto-Caballero, David Turner, Hannah Feinberg, Dan Yuan, Nathaniel Scott, Michael DeBerardine, Shu Dan, Lakme Caceres, Jessica Schembri, Zizhen Yao, Changkyu Lee, Jonathan W. Pillow, Fenna M. Krienen. PIANO: Probabilistic Inference Autoencoder Networks for multi-Omics enables 10x faster single-cell integration and robust generative modeling of gene expression . Cell Reports Methods. November 12, 2026. DOI: 10.1016/j.crmeth.2026.101661
Research reported in these publications was supported by the NIH BRAIN Initiative under award numbers UM1MH130981, DP2MH140136, 5T32MH065214; as well as the Netherlands Organization for Scientific Research NWO:024.004.012; the Klingenstein–Simons Fellowship; and the McDonnell Fellows in Neuroscience at Princeton University.
