stImage: A versatile framework for optimizing spatial transcriptomic analysis through customizable deep histology and location informed integration
Presenting author: Shilin Zhao, Department of Biostatistics, Vanderbilt University Medical Center
Co-authored by:
- Yu Wang, Department of Biostatistics, Vanderbilt University Medical Center
- Haichun Yang, Department of Pathology, Microbiology and Immunology, Vanderbilt University Medical Center
- Ruining Deng, Department of Computer Science, Vanderbilt University
- Yuankai Huo, Department of Computer Science, Vanderbilt University
- Qi Liu, Department of Biostatistics, Vanderbilt University Medical Center
- Yu Shyr, Department of Biostatistics, Vanderbilt University Medical Center
Abstract:
Spatial transcriptomics maps out organizational structures of cells with their genome-wide transcriptional profiles and histology images. Fully exploiting these three different views of data holds great promise to characterize spatial expression heterogeneity accurately. Although several methods have been developed to perform location- and/or histology-informed integration of spatial transcriptomics, they are not one-size-fits-all approaches, but only perform well for certain data and conditions. Here, we present stImage, a comprehensive and flexible framework for optimizing spatial transcriptomic analysis. stImage provides 54 integrative strategies and allows users to develop customized pipelines freely. We illustrate the benefits of stImage for spatial cell clustering on a variety of datasets, demonstrating its superior performance by selecting optimal integrative strategies in different datasets.