SPARK-X: non-parametric modeling enables scalable and robust detection of spatial expression patterns for large spatial transcriptomic studies

Abstract

Spatial transcriptomic studies are becoming increasingly common and large, posing important statistical and computational challenges for many analytic tasks. Here, we present SPARK-X, a non-parametric method for rapid and effective detection of spatially expressed genes in large spatial transcriptomic studies. SPARK-X not only produces effective type I error control and high power but also brings orders of magnitude computational savings. We apply SPARK-X to analyze three large datasets, one of which is only analyzable by SPARK-X. In these data, SPARK-X identifies many spatially expressed genes including those that are spatially expressed within the same cell type, revealing new biological insights.

Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

“…We expect VIPCCA to be broadly used for various integrative analyses such as reference atlas assembly and transfer annotations of reference single-cell data across experiments and modalities. Although we have only demonstrated the ability of VIPCCA in integrating data types through the analysis of scRNA-seq and scATAC-seq data, our integration strategy has the potential for integrating other data types such as the recent spatially resolved transcriptomics datasets with single cell RNAseq data ( 45 , 46 ). Specifically, for single-cell resolution spatial transcriptomics such as MERFISH, SeqFISH, SeqFISH + and STARmap, we can directly treat each location of the spatial transcriptomics data as a single cell and proceed with VIPCCA alignment as if the spatial transcriptomics data is a single cell RNA-seq data.…”

Computational elucidation of spatial gene expression variation from spatially resolved transcriptomics data

“…Based on the non-parametric modeling, SPARK-X effectively reduces memory requirements and computational times while keeping a reliable model effectiveness. 21 Negative binomial distribution refers to a discrete probability distribution, which is suitable for the characteristics of overdispersion and zeroinflated counts in single-cell data. 22,23 GPcounts takes advantage of the Gaussian process regression method, which implements negative binomial likelihood models (sometimes zero-inflated negative binomial [ZINB]) for modeling SRT data, achieving a better fit than the Gaussian likelihood function when dealing with count data.…”

Statistical and machine learning methods for spatially resolved transcriptomics data analysis

“…Although SpatialDE [ 41 ] and SPARK [ 42 ] are more efficient than Trendsceek [ 40 ], the computational complexity of these two methods [ 41 , 42 ] still scales cubically with respect to the number of spatial locations. To reduce computational burden, SPARK-X [ 44 ] proposes a scalable non-parametric model using the following algebraic manipulations. For a given gene, SPARK-X [ 44 ] first builds a covariance matrix for the gene expression and a covariance matrix for the spatial coordinates (Fig.…”

“…To reduce computational burden, SPARK-X [ 44 ] proposes a scalable non-parametric model using the following algebraic manipulations. For a given gene, SPARK-X [ 44 ] first builds a covariance matrix for the gene expression and a covariance matrix for the spatial coordinates (Fig. 2 C).…”

“…Identify SVGs 2. Spatial location reconstruction for scRNA-seq data Jointly identifies SVGs and spatial domains Does not incorporate cell type information and tissue anatomical structure into the computation SPARK-X [ 44 ] Non-parametric covariance test 1. Identify SVGs 2.…”