A comprehensive benchmarking with practical guidelines for cellular deconvolution of spatial transcriptomics

Abstract

Spatial transcriptomics technologies are used to profile transcriptomes while preserving spatial information, which enables high-resolution characterization of transcriptional patterns and reconstruction of tissue architecture. Due to the existence of low-resolution spots in recent spatial transcriptomics technologies, uncovering cellular heterogeneity is crucial for disentangling the spatial patterns of cell types, and many related methods have been proposed. Here, we benchmark 18 existing methods resolving a cellular deconvolution task with 50 real-world and simulated datasets by evaluating the accuracy, robustness, and usability of the methods. We compare these methods comprehensively using different metrics, resolutions, spatial transcriptomics technologies, spot numbers, and gene numbers. In terms of performance, CARD, Cell2location, and Tangram are the best methods for conducting the cellular deconvolution task. To refine our comparative results, we provide decision-tree-style guidelines and recommendations for method selection and their additional features, which will help users easily choose the best method for fulfilling their concerns.

Evaluating the Utilities of Foundation Models in Single-cell Data Analysis

“…The evaluation metrics for this task based on clustering and correlation are provided in Appendix C. The imputation results for scRNA-seq are summarized in Figure 5 (a), which suggest that the imputation function of scGPT for scRNA-seq data introduced more noise into the original sequencing data, suggesting the unreliability of the decoder’s output. According to Figure 5 (b), scGPT performed well in the spatial transcriptomic data imputation task compared to the SOTA spatial imputation method, Tangram [55, 56]. Based on the evaluation of correlation and significance proportion, the imputation results of scGPT are better than the results of Tangram.…”

“…2. Perform imputation for spatial transcriptomic data because of unseen or unmeasured genes [55,56]. According to [90], current spatial imputation methods do not show strong performance across different datasets.…”

SpatialDDLS: An R package to deconvolute spatial transcriptomics data using neural networks

“…Next, we benchmarked SpatialDDLS against two state-of-the-art methods in the spatial transcriptomics field: cell2location (Kleshchevnikov et al, 2022) and RCTD (Cable et al, 2022). We chose these tools because of their high performance in different recently published benchmarks (Li et al, 2022, 2023; Yan and Sun, 2023). In both experiments, SpatialDDLS generated similar predictions to those of cell2location and RCTD.…”

VistoSeg: Processing utilities for high-resolution images for spatially resolved transcriptomics data

“…Given that a single Visium spot contains multiple cells with several cell types, spot deconvolution algorithms are rapidly being developed to predict the proportion of different cell types in each spot. ( 37 , 38 ) Spot deconvolution algorithms generally require Visium gene expression data (with or without cell counts) and single cell/nucleus RNA-seq gene expression data from the same tissue type (). Some spot deconvolution software, including Tangram ( 20 ) and Cell2Location, ( 19 ) require the user to input the number of cells per spot, while others do not require this information.…”

“…Given that a single Visium spot contains multiple cells with several cell types, spot deconvolution algorithms are rapidly being developed to predict the proportion of different cell types in each spot. (37,38) Spot deconvolution algorithms generally require Visium gene expression data (with or without cell counts) and single cell/nucleus RNA-seq gene expression data from the same tissue type (Supplementary Nuclei segmentation to identify fluorescent signal for the nucleus (DAPI) and each labeled protein (GFAP, NEUN, OLIG2, TMEM119) was performed by integrating functions from our previously published software, dotdotdot. (28) (e) Using the split images from (c), Space Ranger (10× Genomics) was used to align multiplex fluorescent imaging and gene expression data and obtain extracted spot metrics (Visium spot diameter, spot spacing and spot coordinates) from each image in the "tissue_positions_list.csv" and "scalefactors_json.json" files.…”