Spatial transcriptomics dimensionality reduction using wavelet bases

Background: Spatially resolved transcriptomics (ST) measures gene expression along with the spatial coordinates of measurements. The analysis ST data involves significant computation complexity. In this work, we propose a dimensionality reduction algorithm that retains structure. Methods: We combine wavelet transformation matrix factorization to select spatially-varying genes. extract low-dimensional representation these adopt an Empirical Bayes perspective, imposing regularization through prior distribution factor Additionally, visualize extracted representations, providing overview global patterns. illustrate performance our methods structure recovery and reconstruction using simulation real analysis. Results: experiments, method identifies factors outperforms regular decomposition regarding error. find connection between fluctuation patterns estimates, allows us provide smoother visualizations. develop package share workflow generating reproducible quantitative results visualization. is available at https://github.com/OliverXUZY/waveST. Conclusions: have proposed pipeline for respects structure. Both simulations experiments demonstrate shrinkage techniques show positive in spatially data. highlight idea combining image processing statistical application genomics context