A point cloud segmentation framework for image-based spatial transcriptomics.

Area of Science:

• Spatial transcriptomics
• Computational biology
• Bioinformatics

Background:

• Image-based spatial RNA profiling offers high-resolution insights into tissue organization.
• Assigning RNA transcripts to individual cells is vital for downstream analyses like cell type identification.
• Accurate cell segmentation is challenging in tissue data, especially without reliable membrane markers.

Purpose of the Study:

• To introduce ComSeg, a novel segmentation algorithm for spatial RNA profiling data.
• To develop a method that operates directly on RNA positions, independent of cell shape priors.
• To enable accurate cell segmentation in complex tissues with diverse cell morphologies.

Main Methods:

• ComSeg algorithm operates directly on single RNA positions.
• Does not rely on implicit or explicit cell shape priors.
• Applicable to complex tissues with arbitrary cell shapes.

Main Results:

• ComSeg demonstrates superior performance compared to existing state-of-the-art methods.
• Outperforms current methods in both in-situ single-cell RNA profiling and cell type calling.
• Validated on simulated and experimental datasets.

Conclusions:

• ComSeg provides an effective solution for cell segmentation in spatial RNA profiling.
• Enables more accurate in-situ cell type calling in complex biological samples.
• The algorithm is available as an open-source package for broader research use.