Summary: Identifying neuron types has long depended on manual, morphology-based methods—like telling trees apart by their leaves. A new artificial intelligence approach reverses that logic: Neuronal Type Assignment from Connectivity (NTAC) shows that a neuron’s pattern of synaptic connections—the wiring diagram—can serve as a more reliable and scalable fingerprint than shape alone.
NTAC classifies thousands of neurons in minutes on an ordinary laptop, reaching over 90% accuracy in regions where different cell types look nearly identical. The system works both with a small set of labeled examples and entirely without labels, offering a fast, reproducible alternative to expert-driven morphological sorting.
Key Facts
- Connectivity as a signature: NTAC demonstrates that synaptic connectivity patterns contain enough information to identify neuronal types, reducing the reliance on morphology which can be misleading in many brain regions.
- High accuracy where shape fails: In the fruit fly optic lobe—an area where neurons tile space and appear highly similar—NTAC achieved approximately 90% accuracy, while shape-based methods (e.g., NBLAST) often failed to exceed 50%.
- Fast and efficient: Tasks that previously required months of expert labor can now be completed in minutes on a conventional laptop.
- Two operational modes:
- Semi-supervised: Uses a small proportion of pre-labeled neurons to guide classification of the rest.
- Unsupervised: Groups neurons into types solely from wiring patterns, without any labels (achieving roughly 70% accuracy in challenging regions).
- Connectome potential: Researchers compare NTAC’s significance to the rise of genomics—mapping connectivity at scale could transform our understanding of circuit organization and disease mechanisms.
Source: JAIST
Recent technological progress in electron microscopy and computer vision has enabled reconstruction of complete connectomes in small organisms and partial connectomes in larger animals. Accurately assigning cell types within these connectomes is essential for interpreting circuit function and for comparing brain organization across species. Manual, morphology-based cell typing remains slow and error-prone, especially in circuits with repeated or highly similar cell shapes. Connectivity-based approaches like NTAC offer a robust alternative.
In a recently published study, an international team introduced NTAC, an automated system that assigns neuronal types using only synaptic connectivity. NTAC runs efficiently on standard computers and delivers high-precision results without requiring detailed morphological reconstructions. The work was led by Dr. Gregory Schwartzman (Japan Advanced Institute of Science and Technology) with collaborators at Princeton Neuroscience Institute, the University of Edinburgh, and the Technical University of Catalonia.
Published in Volume 17 of Nature Communications on January 6, 2026, the paper was highlighted by the journal’s Editors’ Highlights for its importance to the field. Dr. Schwartzman explains that as connectome datasets grow, labeling cells manually becomes a major bottleneck. “NTAC assigns neuronal types from synaptic connectivity with very high accuracy. The wiring diagram itself carries enough signal to identify neuron types quickly, even when only a small fraction of neurons is labeled,” he said.
The team evaluated NTAC on multiple high-quality fruit fly connectomes and compared its performance to morphology-driven methods such as NBLAST. In regions like the optic lobe, where neurons tile space and morphological differences are subtle, NTAC significantly outperformed shape-based classifiers. Morphology-based approaches often required far more labeled examples and still struggled to exceed 50% accuracy, while NTAC surpassed 90% using only a fraction of labeled data and completed processing in minutes on a laptop.
In fully unsupervised mode—no labels provided—NTAC achieved roughly 70% accuracy in challenging regions, greatly outperforming morphology-based clustering which frequently remained under 10%. When applied across an entire brain containing thousands of distinct cell types, NTAC’s unsupervised accuracy reached 52%, a promising result given the scale and complexity involved.
The researchers note that NTAC has already been used to annotate thousands of neurons in the brain-and-cord connectome (BANC) dataset. Looking forward, NTAC’s scalability positions it as a valuable tool for larger-scale connectomics projects, including mouse brain efforts and, ultimately, the much larger human connectome. Incorporating multimodal data—combining connectivity with gene expression, morphology, or other signals—could further improve classification accuracy and provide a richer view of neuronal diversity.
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 received support from 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 funded by the Spanish Agencia Estatal de Investigación (Project PID2020-112581GB-C21 MOTION).
Key Questions Answered:
A: Shape can be misleading. Two neurons may look nearly identical but connect to different partners and therefore perform different roles. Like electrical wires that appear the same but serve different functions depending on their connections, neurons’ synaptic partners reveal their functional identity—information NTAC extracts from the connectome.
A: NTAC makes an important step toward scalable analysis but mapping the human brain remains an enormous challenge. So far, complete connectomes exist for very small organisms. NTAC is designed to scale, and automated tools like it will be essential for processing the massive datasets needed for larger mammalian and future human connectomes.
A: No. NTAC was built for efficiency and runs on conventional computers, including laptops, making advanced connectome analysis more accessible to research groups without large computing clusters.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full.
- Additional context was added by editorial staff.
About this neuroscience research news
Author: Gregory Schwartzman
Source: JAIST
Contact: Gregory Schwartzman – JAIST
Image: Image credit: Neuroscience News
Original Research: Open access. “NTAC: Neuronal type assignment from connectivity” by Gregory Schwartzman, Ben Jourdan, David García-Soriano & Arie Matsliah. Nature Communications. DOI: 10.1038/s41467-025-68044-1
Abstract
NTAC: Neuronal type assignment from connectivity
Advances in electron microscopy and automated image analysis now permit reconstruction of large-scale wiring diagrams, or connectomes. These datasets create an urgent need for automated methods that can identify neuronal cell types directly from connectivity data. We show that synaptic connectivity alone can assign neurons to cell types with high accuracy. We introduce NTAC, which groups neurons solely by connectivity, and provide two variants: a semi-supervised form that leverages a small fraction of labeled neurons to infer types for the rest, and an unsupervised form that requires no labels. Applied to multiple state-of-the-art fruit fly connectomes, NTAC achieves high accuracy in minutes on a laptop, demonstrating that connectivity is a powerful, scalable basis for classifying neuronal cell types across the brain.