Summary: Identifying the “type” of a neuron has traditionally been like identifying a tree by its leaves—a manual, slow process based on morphology. However, he investigated the script.
Their new AI program, Neuronal Type Assignment from Connectivity (NTAC), proves that the “wiring diagram” of a neuron—which it communicates with—is a more accurate fingerprint than its physical appearance. NTAC can classify thousands of neurons in minutes on a standard laptop with over 90% accuracy, even in brain regions where neurons look similar.
Important Facts
- Communication as DNA: NTAC shows that synaptic connections alone contain enough information to identify the type of neuron, making manual morphological classification (often misleading) obsolete.
- High Accuracy: In the optic lobe of the fruit fly—where neurons look very similar—NTAC has been achieved 90% accuracy.while traditional alignment-based methods (such as NBLAST) struggle to reach 50%.
- Speed and efficiency: What once took months of expert manual work can now be done minutes on a standard laptop.
- Two ways of working:
- Less supervision: It uses a subset of previously labeled neurons to “teach” the AI to identify the rest.
- Unsupervised: It groups neurons into categories based on wiring patterns without any previous labels (achieving ~70% accuracy in complex regions).
- Genomic Parallel: Researchers compare this success to the rise of genomics; Just as genetic mapping revolutionized medicine, mapping the “Connectome” could unlock how circuits fail in brain disorders.
Source: JAIST
Recent technological advances facilitate the reconstruction of complete brain connections in small animals and incomplete connections in mammals, including the mapping of networks of neurons and synaptic connections. Accurate cell typing of these connections helps define regional functions and compare brain organization across species.
Traditionally, cell typing relied on manual morphological classification by experts—a slow process that required detailed anatomical knowledge. However, morphology can be misleading or inadequate in many brain regions, especially in circuits with repeated cell types, where neurons can share a very similar morphology despite differences in connectivity.
In a recently published study, researchers have developed Neuronal Type Assignment from Connectivity (NTAC), an automated system that delivers high-precision results and works well even on conventional computers, showing that synaptic connectivity alone contains enough information to identify neuronal types without relying on morphological characteristics.
This article is the result of an international collaboration between the Japan Advanced Institute of Science and Technology (JAIST), the Princeton Neuroscience Institute, the University of Edinburgh, and the Technical University of Catalonia. The study was led by Dr. Gregory Schwartzman, Associate Professor at JAIST, and included Dr. Ben Jourdan from the University of Edinburgh, Dr. David García-Soriano from the Technical University of Catalonia, and Dr. Arie Matsliah from Princeton University.
The article was published in Volume 17 of the Natural Communication on January 6, 2026. Posted on the Editors' Highlights page, which aims to showcase the top papers recently published in the area.
Explaining their research, Dr. Schwartzman says, “Our research comes in the context of expanding connectomes and the growing need for automated and scalable tools.
“NTAC is able to assign neuronal types based exclusively on synaptic connections, with very high accuracy. It shows that the wiring diagram itself carries enough signal to quickly identify the types of neurons, even if only a small fraction of neurons are labeled.”
Researchers have developed two methods of NTAC operation. Another version is less supervised, where a small fraction of neurons are pre-labeled, and the algorithm uses connectivity patterns to understand the types of remaining neurons. In the unsupervised version, no pre-labels are required; the algorithm groups neurons based solely on the similarity of their wires.
The algorithm was applied to multiple high-resolution fruit fly brain connections, and the accuracy of NTAC was compared to morphology-based methods that rely on NBLAST, a widely used method for comparing neuronal states. In the optic lobe, a region where neurons are tiled and difficult to distinguish morphologically, NTAC performed better than NBLAST-based classifiers.
While morphology-based methods require many labeled specimens and still struggle to reach 50% accuracy in some settings, NTAC exceeded 90% for a portion of the labeled data and in minutes on a laptop.
In fully unsupervised mode, NTAC achieved an accuracy of about 70%, which far exceeds morphology-based clustering methods, which often remain below 10%. For the full brain, which contains thousands of different cell types, the unsupervised accuracy reached 52%, an encouraging result given the scale and complexity of the data.
“The long-term goal of connectomics is to map the entire human brain and derive scientific and medical information from it, similar to how biology and medicine were revolutionized by genomics. Currently, complete connectivity is designed only for very small organisms such as fruit flies.
“NTAC can speed up the development and analysis of connectors, which could accelerate scientific discovery, and, in the future, could contribute to efforts to isolate neuronal cell types in a large number of mammals and eventually connect them to humans.
“This algorithm has already been successfully used to label thousands of neurons in the brain-and-cord connectome (BANC) dataset. The next frontier in connectomics is mapping the mouse brain, and our algorithm can play a major role in this effort,” explained Dr. Schwartzman.
Further improving the algorithm by combining multimodal data can improve classification performance and yield a more comprehensive understanding of neuronal cell types.
Funding information
Arie Matsliah was supported by grants to Murthy and Seung from the NIH BRAIN Initiative (RF1 MH117815, RF1 MH129268, U24 NS126935). Gregory Schwartzman was supported by the following research grants: KAKENHI 25K00370, JST ASPIRE JPMJAP2302, and JST CRONOS JPMJCS24K2. Ben Jourdan was supported by an EPSRC Early Career Fellowship (EP/T00729). David García-Soriano was supported by the Spanish Agencia Estatal de Investigación AEI/10.13039/501100011033 (Project PID2020-112581GB-C21 MOTION).
Important Questions Answered:
A: Imagine two electrical wires that look the same in your house. One connects to the light switch; the other connects to the doorbell. Looking at the cable itself won't tell you its function—you have to see where it's connected. In the “optic lobe” of the brain, neurons “tile” the space and look like individual clones, but NTAC shows their unique “plugs” (synapses) that reveal their true nature.
A: We are getting closer. Currently, we have mapped small organisms such as flies. However, NTAC is designed to be scalable. As we move closer to mapping the mouse brain and eventually the human brain (which has 86 billion neurons), automated tools like NTAC are the only way to process that large amount of data.
A: Surprisingly, no. One of the highlights of the study is that NTAC works very well. It's built to last standard computers (laptops)making high-level neuroscience accessible to many researchers around the world without requiring large server farms.
Editor's Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper is fully revised.
- More content has been added by our staff.
About this neuroscience research news
Author: Gregory Schwartzman
Source: JAIST
Contact person: Gregory Schwartzman – JAIST
Image: Image posted in Neuroscience News
Actual research: Open access.
“NTAC: Neuronal type assignment from connectivity” by Gregory Schwartzman, Ben Jourdan, David García-Soriano & Arie Matsliah. Natural Communication
DOI:10.1038/s41467-025-68044-1
Abstract
NTAC: Neuronal type assignment from connectivity
Recent advances in electron microscopy and computer vision now allow the reconstruction of complete wiring diagrams, or connections, of animal brains. This creates an urgent need for methods that can automatically identify neuronal cell types directly from these large connectivity datasets.
Here we show that synaptic connections alone can be used to assign neurons to cell types with high precision. We present NTAC (Neuronal Type Assignment from Connectivity), which groups neurons based solely on connectivity.
NTAC has two types: semi-supervised that uses a small subset of labeled neurons to infer the types of all others, and unsupervised that requires no labels at all.
Applied to high-throughput fruit fly brain connectivity, NTAC achieves high accuracy within minutes on a laptop, demonstrating that connectivity provides a powerful and scalable basis for distinguishing neuronal cell types in the brain.
