Cell segmentation for high-resolution spatial transcriptomics
01:42 Share

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2023.01.11.523658v1?rss=1

Authors: Chen, H., Li, D., Bar-Joseph, Z.

Abstract: Spatial transcriptomics promises to greatly improve our ability to understand tissue organization and cell-cell interactions. While most current platforms for spatial transcriptomics only provide multi-cellular resolution (10-15 cells per spot), recent technologies provide a much denser spot placement leading to sub-cellular resolution. A key challenge for these newer methods is cell segmentation and the assignment of spots to cells. Traditional, image based, segmentation methods face several drawbacks and do not take full advantage of the information profiled by spatial transcriptomics. Here, we present SCS, which integrates imaging data with sequencing data to improve cell segmentation accuracy. SCS combines information from neighboring spots and employs a transformer model to adaptively learn the relevance of different spots and the relative position of each spot to the center of its cell. We tested SCS on two new sub-cellular spatial transcriptomics technologies and compared its performance to traditional image based segmentation methods. As we show, SCS achieves better accuracy, identifies more cells and leads to more realistic cell size estimation. Analysis of RNAs enriched in different sub-cellular regions based on SCS spot assignments provides information on RNA localization and further supports the segmentation results.

Copy rights belong to original authors. Visit the link for more info

Podcast created by Paper Player, LLC