BidCell and Baysor on Large-Scale Single-Cell and Spatial Transcriptomics Datasets

Abstract

The purpose of this research is to examine the different types of packages that can be used for cell segmentation, specifically on much larger datasets. The key packages studied were called “Baysor” and “Bidcell”. Previous research discussed the different methods of each package, which served as a foundation for this study to further understand how well each package handles large datasets. In this study, the Xenium mouse brain dataset with 62 million transcripts was used with default parameters used in each package. Furthermore, based on the cell segmentation results, Bidcell appears to have under cell segmentation results, with many cells not being represented when drawn. On the other hand, Baysor appears to have over cell segmentation, with excessive outlining of the cell boundaries. However, Bidcell or Baysor’s results may have accurate results to the real segmentation boundaries. Future studies would benefit from testing by altering parameters and comparing them to the default parameters’ segmentation results.