SpaTM: Topic Models for Inferring Spatially Informed Transcriptional Programs

Abstract

BackgroundSpatial transcriptomics has revolutionized our ability to characterize tissues and diseases by contextualizing gene expression with spatial organization. Current spatial transcriptomics pipelines require researchers to use a variety of models and tools to explore spatial domains. However, few methods provide researchers with a way to jointly analyze spatial data from both annotation-free and annotation-guided perspectives using consistent inductive biases and levels of interpretability. A single framework with consistent inductive biases ensures coherence and transferability across tasks, reducing the risks of conflicting assumptions. ResultsWe propose the Spatial Topic Model (SpaTM), a topic-modelling framework capable of annotation-guided and annotation-free analysis of spatial transcriptomics data. SpaTM can be used to learn gene programs that represent histology-based annotations while providing researchers with the ability to infer spatial domains with an annotation-free approach if manual annotations are limited or noisy. Our benchmarking experiments reveal SpaTMs competitiveness at spatial label prediction and clustering when compared to state-of-the-art methods. We also demonstrate SpaTMs interpretability with its use of topic mixtures to represent cell states and transcriptional programs in dorsolateral prefrontal cortex and ductal carcinoma samples and how its intuitive framework facilitates the integration of annotation-guided and annotation-free analyses of spatial data with downstream analyses. Finally, we demonstrate how SpaTM can be used to extend the analysis of large-scale snRNA-seq atlases with the inference of cell proximity and spatial annotations in human brains with Major Depressive Disorder. ConclusionsSpaTM provides researchers with a unified analysis framework for spatial transcriptomics data. By enabling competitive performance in a variety of tasks, SpaTM helps researchers undertake biologically-driven analyses through the identification of interpretable and biologically informed gene programs.