CellART extracts single-cell data from high-resolution spatial transcriptomics
Researchers have introduced CellART, a unified framework designed to extract single-cell information from high-resolution spatial transcriptomics (ST) platforms. The tool addresses limitations in existing systems, which often capture sparse transcript counts or measure only limited gene sets. By integrating multimodal data, including staining images, ST data, and single-cell RNA sequencing…
Key points
- CellART combines deep learning and probabilistic modeling to segment cells and annotate types from spatial transcriptomics data.
- The framework works across diverse platforms including 10x Genomics Xenium, Visium HD, Vizgen MERFISH, and BGI Stereo-seq.
- CellART is open-source, available in Python under the MIT license, and compatible with existing community analysis tools.
The framework is described as efficient, generalizable, and robust, with outputs compatible with widely used community tools to facilitate downstream analyses. The study demonstrates CellART's performance on datasets from 10x Genomics (Xenium, Visium HD), Vizgen (MERFISH), and BGI (Stereo-seq), including human lung, mouse brain, and cancer tissues. The authors report that CellART improves segmentation accuracy and transcript coverage compared to methods like ProSeg, Baysor, Cellpose, and Bin2Cell.
CellART is implemented in Python and is available under the MIT license on GitHub. All datasets analyzed in the study are publicly available, and processed data along with simulation scripts are deposited on Zenodo. The work was supported by various grants from the National Natural Science Foundation of China and the Hong Kong Research Grants Council.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researchers introduce APDMem hierarchical memory for long-context LLM assistants · 1 src
- Researchers introduce calibrated user embeddings for multi-turn AI benchmarking · 1 src
- SymCE corpus released for Counterexample generation · 1 src
- HakemBench releases 2,346-item Turkish benchmark for typed decisions · 1 src
- Researchers introduce MEA, a Reward-Driven Multi-Agent system for faithful model explanations · 1 src
Comments
via GitHub Discussions