Spatial transcriptomics (ST) facilitates the exploration of biological tissue structures and functions within spatial context. Slice alignment and integration are prevalent for analyzing ST data, and current algorithms either focus on adjacent slices or require prior information to guide alignment, limits their applications for downstream analysis. Here, we present AlignDG, an information theory-based graph model that jointly aligns and integrates ST slices across diverse diseases, platforms and conditions without prior information. Experimental results demonstrate that AlignDG outperforms existing baselines in terms of precision, robustness, and efficiency with approximate 50% of slices, providing an effective alternative for analyzing ST data (code: https://github.com/xkmaxidian/AlignDG).
Wang et al. (Thu,) studied this question.