Integrated analysis of multimodal single-cell data

Abstract

Summary The simultaneous measurement of multiple modalities represents an exciting frontier for single-cell genomics and necessitates computational methods that can define cellular states based on multimodal data. Here, we introduce “weighted-nearest neighbor” analysis, an unsupervised framework to learn the relative utility of each data type in each cell, enabling an integrative analysis of multiple modalities. We apply our procedure to a CITE-seq dataset of 211,000 human peripheral blood mononuclear cells (PBMCs) with panels extending to 228 antibodies to construct a multimodal reference atlas of the circulating immune system. Multimodal analysis substantially improves our ability to resolve cell states, allowing us to identify and validate previously unreported lymphoid subpopulations. Moreover, we demonstrate how to leverage this reference to rapidly map new datasets and to interpret immune responses to vaccination and coronavirus disease 2019 (COVID-19). Our approach represents a broadly applicable strategy to analyze single-cell multimodal datasets and to look beyond the transcriptome toward a unified and multimodal definition of cellular identity.

Multi-omic profiling reveals widespread dysregulation of innate immunity and hematopoiesis in COVID-19

“…To address this, we mapped our transcriptomic dataset to a large multimodal reference dataset introduced by Seurat v4, which incorporated extensive surface marker information to improve cell type calls (Fig. 2 A; Hao et al, 2021).…”

“…To prioritize downstream analysis of cell subsets most affected by COVID-19, we calculated a perturbation score (Hao et al, 2021;Papalexi et al, 2021) for each cell type from each COVID-19 sample relative to healthy control subject samples (see Materials and methods). The perturbation score for each cell type is calculated by first identifying genes that display evidence of differential expression between COVID-19 samples and healthy control samples, calculating the difference of pseudobulk expression vectors of these genes between COVID-19 samples and healthy control samples, and finally projecting the whole transcriptome of each donor onto this vector.…”

“…Seurat v4 introduced weighted nearest neighbors (WNNs) analysis as a strategy to integrate multimodal single-cell sequencing data and released a large multimodal dataset of 228 cell surface markers with simultaneous whole-transcriptome measurements (Hao et al, 2021). By using WNN to learn the relative utility of each data modality in each cell, this reference dataset contains highly robust and granular cell type annotations.…”

“…By using WNN to learn the relative utility of each data modality in each cell, this reference dataset contains highly robust and granular cell type annotations. The web application provided with the release of Seurat v4 (http://azimuth.satijalab.org/) was used to map our transcriptomic dataset to the annotated multimodal reference (Hao et al, 2021). Briefly, anchors between the query and reference dataset were identified using a precomputed supervised PCA on the reference dataset that maximally captures the structure of the WNN graph.…”

“…To prioritize analysis of cell types of interest, we calculated a perturbation score for each cell type of each sample as previously described (Hao et al, 2021;Papalexi et al, 2021). This perturbation score is motivated by the observation that the statistical significance of per-gene differential expression tests is strongly influenced by the number of cells in each cluster or cell type.…”