DESpace

DESpace identifies spatially variable genes and differential spatial patterns in spatially resolved transcriptomics data by performing differential expression testing between spatial clusters using negative binomial models.

Key Features:

• Spatially Variable Gene Detection: DESpace identifies spatially variable genes by testing the significance of spatial clusters as covariates in a negative binomial model fitted via edgeR.
• Differential Spatial Pattern Detection: DESpace detects genes exhibiting changes in spatial expression patterns across experimental conditions by testing the interaction between spatial clusters and conditions in a negative binomial model.
• Multi-sample Joint Modeling: DESpace enables joint modeling of multiple samples and biological replicates, increasing statistical power and targeting genes with consistent spatial patterns across replicates.
• Cluster-specific Variability Identification: DESpace identifies and tests the main tissue clusters affected by spatial variability, allowing investigation of spatial expression changes in specific areas of interest.
• Multi-condition Comparisons: DESpace detects spatial gene expression pattern changes across more than two experimental conditions using flexible models.
• Universal SRT Data Input: DESpace processes all types of spatially resolved transcriptomics data by summarizing spatial information via pre-annotated spatial clusters.

Scientific Applications:

• Spatially Resolved Transcriptomics Analysis: DESpace enables investigation of spatial context of mRNA abundance in tissue samples.
• Condition Comparison Studies: DESpace supports comparison of spatial gene expression patterns across experimental conditions such as healthy versus diseased states.
• Tissue Region-specific Expression Analysis: DESpace identifies tissue regions where gene expression changes across conditions, enabling focused investigation of spatially localized effects.

Methodology:

DESpace models gene expression using a negative binomial distribution via edgeR, with spatial clusters as covariates. For spatially variable gene detection, significance of spatial clusters is tested. For differential spatial pattern detection across conditions, a negative binomial model incorporating spatial clusters, conditions, and their interactions as covariates is fitted, and the interaction between spatial clusters and conditions is tested.