Background: Single-cell transcriptomics is rapidly advancing our understanding of the cellular composition of complex tissues and organisms. A major limitation in most analysis pipelines is the reliance on manual annotations to determine cell identities, which are time-consuming and irreproducible. The exponential growth in the number of cells and samples has prompted the adaptation and development of supervised classification methods for automatic cell identification. Results: Here, we benchmarked 22 classification methods that automatically assign cell identities including single-cell-specific and general-purpose classifiers. The performance of the methods is evaluated using 27 publicly available single-cell RNA sequencing datasets of different sizes, technologies, species, and levels of complexity. We use 2 experimental setups to evaluate the performance of each method for within dataset predictions (intra-dataset) and across datasets (inter-dataset) based on accuracy, percentage of unclassified cells, and computation time. We further evaluate the methods’ sensitivity to the input features, number of cells per population, and their performance across different annotation levels and datasets. We find that most classifiers perform well on a variety of datasets with decreased accuracy for complex datasets with overlapping classes or deep annotations. The general-purpose support vector machine classifier has overall the best performance across the different experiments. Conclusions: We present a comprehensive evaluation of automatic cell identification methods for single-cell RNA sequencing data. All the code used for the evaluation is available on GitHub (https://github.com/tabdelaal/scRNAseq_Benchmark). Additionally, we provide a Snakemake workflow to facilitate the benchmarking and to support the extension of new methods and new datasets.

Single-cell technologies are emerging fast due to their ability to unravel the heterogeneity of biological systems. While scRNA-seq is a powerful tool that measures whole-transcriptome expression of single cells, it lacks their spatial localization. Novel spatial transcriptomics methods do retain cells spatial information but some methods can only measure tens to hundreds of transcripts. To resolve this discrepancy, we developed SpaGE, a method that integrates spatial and scRNA-seq datasets to predict whole-transcriptome expressions in their spatial configuration. Using five dataset-pairs, SpaGE outperformed previously published methods and showed scalability to large datasets. Moreover, SpaGE predicted new spatial gene patterns that are confirmed independently using in situ hybridization data from the Allen Mouse Brain Atlas.

Understanding the mechanisms and impact of booster vaccinations are essential in the design and delivery of vaccination programs. Here we show that a three dose regimen of a synthetic peptide vaccine elicits an accruing CD8⁺ T cell response against one SARS-CoV-2 Spike epitope. We see protection against lethal SARS-CoV-2 infection in the K18-hACE2 transgenic mouse model in the absence of neutralizing antibodies, but two dose approaches are insufficient to confer protection. The third vaccine dose of the single T cell epitope peptide results in superior generation of effector-memory T cells and tissue-resident memory T cells, and these tertiary vaccine-specific CD8⁺ T cells are characterized by enhanced polyfunctional cytokine production. Moreover, fate mapping shows that a substantial fraction of the tertiary CD8⁺ effector-memory T cells develop from re-migrated tissue-resident memory T cells. Thus, repeated booster vaccinations quantitatively and qualitatively improve the CD8⁺ T cell response leading to protection against otherwise lethal SARS-CoV-2 infection.

Duchenne muscular dystrophy is caused by mutations in the DMD gene, leading to lack of dystrophin. Chronic muscle damage eventually leads to histological alterations in skeletal muscles. The identification of genes and cell types driving tissue remodeling is a key step to developing effective therapies. Here we use spatial transcriptomics in two Duchenne muscular dystrophy mouse models differing in disease severity to identify gene expression signatures underlying skeletal muscle pathology and to directly link gene expression to muscle histology. We perform deconvolution analysis to identify cell types contributing to histological alterations. We show increased expression of specific genes in areas of muscle regeneration (Myl4, Sparc, Hspg2), fibrosis (Vim, Fn1, Thbs4) and calcification (Bgn, Ctsk, Spp1). These findings are confirmed by smFISH. Finally, we use differentiation dynamic analysis in the D2-mdx muscle to identify muscle fibers in the present state that are predicted to become affected in the future state.

RNA Velocity allows the inference of cellular differentiation trajectories from single-cell RNA sequencing (scRNA-seq) data. It would be highly interesting to study these differentiation dynamics in the spatial context of tissues. Estimating spatial RNA velocities is, however, limited by the inability to spatially capture spliced and unspliced mRNA molecules in high-resolution spatial transcriptomics. We present SIRV, a method to spatially infer RNA velocities at the single-cell resolution by enriching spatial transcriptomics data with the expression of spliced and unspliced mRNA from reference scRNA-seq data. We used SIRV to infer spatial differentiation trajectories in the developing mouse brain, including the differentiation of midbrain-hindbrain boundary cells and marking the forebrain origin of the cortical hem and diencephalon cells. Our results show that SIRV reveals spatial differentiation patterns not identifiable with scRNA-seq data alone. Additionally, we applied SIRV to mouse organogenesis data and obtained robust spatial differentiation trajectories. Finally, we verified the spatial RNA velocities obtained by SIRV using 10x Visium data of the developing chicken heart and MERFISH data from human osteosarcoma cells. Altogether, SIRV allows the inference of spatial RNA velocities at the single-cell resolution to facilitate studying tissue development.

The human fetal immune system must protect the infant against the sudden exposure to a large variety of pathogens upon birth. While it is known that the fetal immune system develops in sequential waves, relatively little is known about the composition of the innate and adaptive immune system in the tissues. Here, we applied high-dimensional mass cytometry to profile the immune system in human fetal liver, spleen, and intestine. With Hierarchical Stochastic Neighbor Embedding (HSNE) we distinguished 177 distinct immune cell clusters, including both previously identified and novel cell clusters. PCA analysis indicated substantial differences between the compositions of the immune system in the different organs. Through dual t-SNE we identified tissue-specific cell clusters, which were found both in the innate and adaptive compartment. To determine the spatial location of tissue-specific subsets we developed a 31-antibody panel to reveal both the immune compartment and surrounding stromal elements through analysis of snap-frozen tissue samples with imaging mass cytometry. Imaging mass cytometry reconstructed the tissue architecture and allowed both the characterization and determination of the location of the various immune cell clusters within the tissue context. Moreover, it further underpinned the distinctness of the immune system in the tissues. Thus, our results provide evidence for early compartmentalization of the adaptive and innate immune compartment in fetal spleen, liver, and intestine. Together, our data provide a unique and comprehensive overview of the composition and organization of the human fetal immune system in several tissues.