Benchmarking Computational Integration Methods for Spatial Transcriptomics Data

Abstract

The increasing popularity of spatial transcriptomics has allowed researchers to analyze transcriptome data in its tissue sample's spatial context. Various methods have been developed for detecting SV (spatially variable) genes, with distinct spatial expression patterns. However, the accuracy of using such SV genes in clustering cell types has not been thoroughly studied. On the other hand, in single cell resolution sequencing data, clustering analysis is usually done on highly variable (HV) genes. Here we investigate if integrating SV genes and HV genes from spatial transcriptomics data can improve clustering performance beyond using SV genes alone. We evaluated six methods that integrate different features measured from the same samples including MOFA+, scVI, Seurat v4 , CIMLR, SNF, and the straightforward concatenation approach. We applied these methods on 19 real datasets from three different spatial transcriptomics technologies (merFISH, SeqFISH+, and Visium) as well as 20 simulated datasets of varying spatial expression conditions. Our evaluations show that the performances of these integration methods are largely dependent on spatial transcriptomics platforms. Despite the variations among the results, in general MOFA+ and simple concatenation have good performances across different types of spatial transcriptomics platforms. This work shows that integrating quantitative and spatial marker genes in the spatial transcriptomics data can improve clustering. It also provides practical guides on the choices of computational methods to accomplish this goal.

nnSVG for the scalable identification of spatially variable genes using nearest-neighbor Gaussian processes

“…In spatial data, we can modify this workflow by replacing the set of top HVGs with the set of top SVGs from nnSVG, and then perform unsupervised clustering on the set of top SVGs. Since the set of top SVGs has been generated by methodology that takes spatial information into account, this can give a spatially-aware clustering of cell populations [8, 40].…”

Computational Methods for Single-Cell Multi-Omics Integration and Alignment

“…Oftentimes, the integration methods were developed with one specific application/assay in mind, generalization of these methods with the emergence of new technologies needs to be demonstrated. Fortunately, some benchmarking studies have been conducted in other sub-fields of single-cell computational biology for references, such as those focused on the integration of data from different cells and atlas study [72] , cell-type annotation [73] , and integration algorithms for spatial transcriptomics [74] . Creating standardized high-quality benchmarking datasets would aid such efforts, as proposed in [75] for scRNA-seq data.…”

The state of the art for artificial intelligence in lung digital pathology

“…Intratumoral heterogeneity in NSCLC has been well studied [160]. Recent studies have mapped spatial transcriptomics onto histopathology feature maps to analyze the role of different cell populations in cancers [161,162]. The underlying molecular mechanisms for several morphologic features in NSCLC have largely remained unexplained [163].…”