TransST: Transfer Learning Embedded Spatial Factor Modeling of Spatial Transcriptomics Data

Abstract

2## Methods We use “spot” to denote a location that corresponds to either a cell or a cell mixture in spatial transcriptomics data, depending on the specific technology. Let $\text{X}_{src}\inℝn_{1}\timesp$ and $\text{X}_{tgt}\inℝn_{0}\timesp$ be the $p$-dimensional gene expression matrices from the source data with $n_{1}$ cells/spots and the target data with $n_{0}$ cells/spots, respectively. Denote $\text{U}$ and $\text{V}$ as the corresponding low-dimensional representations of $\text{X}_{src}$ and $\text{X}_{tgt}$. We use the lowercase letters with subscript $i$ to denote the $p$-dimensio