Mitigating autocorrelation during spatially resolved transcriptomics data analysis

Abstract

Several computational methods have recently been developed for characterizing molecular tissue regions in spatially resolved transcriptomics (SRT) data. However, each method fundamentally relies on spatially smoothing transcriptomic features across neighboring cells. Here, we demonstrate that smoothing increases autocorrelation between neighboring cells, causing latent space to encode physical adjacency rather than spatial transcriptomic patterns. We find that randomly subsampling neighbors before smoothing mitigates autocorrelation, improving the performance of existing methods and further enabling a simpler, more efficient approach that we call spatial integration (SPIN). SPIN leverages the conventional single-cell toolkit, yielding spatial analogies to each tool: clustering identifies molecular tissue regions; differentially expressed gene analysis calculates region marker genes; trajectory inference reveals continuous, molecularly defined anatomical axes; and integration allows joint analysis across multiple SRT datasets, regardless of tissue morphology, spatial resolution, or experimental technology. We apply SPIN to SRT datasets from mouse and marmoset brains to calculate shared and species-specific region marker genes as well as a molecularly defined neocortical depth axis along which several genes and cell types differ across species.

Quantitative characterization of cell niches in spatial atlases

“…This encoder jointly encodes the omics features of the subgraphs composed of nodes and their immediate neighbors, thus capturing cellular microenvironments. During model training, neighbors in this subgraph are subsampled ( Methods ), reducing the computational footprint of NicheCompass and mitigating the common autocorrelation problem in spatial transcriptomics GNNs 78 . A separate encoding module generates embedding vectors for the node-specific associated covariates used for batch-effect removal 32 ( Methods ).…”

Ablation of Vitamin D Signaling in Cardiomyocytes Leads to Functional Impairment and Stimulation of Pro-Inflammatory and Pro-Fibrotic Gene Regulatory Networks in a Left Ventricular Hypertrophy Model in Mice

“…In all spots with VDR expression, we also detected cardiomyocyte-specific Myl2 expression in the ventricle of the VDR fl/fl mouse. Using the SPIN algorithm [ 29 ], we identified two distinct region clusters across TAC VDR fl/fl and VDR CM-KO samples, based on unique gene marker patterns that define specific tissue regions ( Figure 4 A). This spatial integration revealed two primary clusters corresponding to inner and outer tissue regions (cluster 0 representing the outer region and cluster 1 the inner region), with an even distribution of spots among them (0_fl (1649), 0_KO (1641), 1_fl (1414), and 1_KO (1327) ( Figure 4 B).…”

“…In our spatial transcriptomics analysis, we first applied the SPIN algorithm [ 29 ] to address autocorrelation issues between neighboring spots, enhancing the resolution of spatial patterns and improving cluster identification between hypertrophic and healthy heart tissues. This precision allowed for a detailed scrutiny of gene regulatory network (GRN).…”

“…The SPIN algorithm [ 29 ] integrated TAC VDR CM-KO and VDR fl/fl datasets, addressing batch effects while retaining spatial information. Dimensionality reduction via PCA and spatial pattern visualization through UMAP facilitated the identification of unique expression domains through Leiden clustering.…”