CellMixS: quantifying and visualizing batch effects in single-cell RNA-seq data

Integrating single-cell RNA-seq datasets with substantial batch effects

Integration of single-cell RNA-sequencing (scRNA-seq) datasets has become a standard part of the analysis, with conditional variational autoencoders (cVAE) being among the most popular approaches. Increasingly, researchers are asking to map cells across challenging cases such as cross-organs, species, or organoids and primary tissue, as well as different scRNA-seq protocols, including single-cell and single-nuclei. Current computational methods struggle to harmonize datasets with such substantial differences, driven by technical or biological variation. Here, we propose to address these challenges for the popular cVAE-based approaches by introducing and comparing a series of regularization constraints. The two commonly used strategies for increasing batch correction in cVAEs, that is Kullback-Leibler divergence (KL) regularization strength tuning and adversarial learning, suffer from substantial loss of biological information. Therefore, we adapt, implement, and assess alternative regularization strategies for cVAEs and investigate how they improve batch effect removal or better preserve biological variation, enabling us to propose an optimal cVAE-based integration strategy for complex systems. We show that using a VampPrior instead of the commonly used Gaussian prior not only improves the preservation of biological variation but also unexpectedly batch correction. Moreover, we show that our implementation of cycle-consistency loss leads to significantly better biological preservation than adversarial learning implemented in the previously proposed GLUE model. Additionally, we do not recommend relying only on the KL regularization strength tuning for increasing batch correction, as it removes both biological and batch information without discriminating between the two. Based on our findings, we propose a new model that combines VampPrior and cycle-consistency loss. We show that using it for datasets with substantial batch effects improves downstream interpretation of cell states and biological conditions. To ease the use of the newly proposed model, we make it available in the scvi-tools package as an external model named sysVI. Moreover, in the future, these regularization techniques could be added to other established cVAE-based models to improve the integration of datasets with substantial batch effects.

A model of human neural networks reveals NPTX2 pathology in ALS and FTLD

Human cellular models of neurodegeneration require reproducibility and longevity, which is necessary for simulating age-dependent diseases. Such systems are particularly needed for TDP-43 proteinopathies1, which involve human-specific mechanisms2-5 that cannot be directly studied in animal models. Here, to explore the emergence and consequences of TDP-43 pathologies, we generated induced pluripotent stem cell-derived, colony morphology neural stem cells (iCoMoNSCs) via manual selection of neural precursors6. Single-cell transcriptomics and comparison to independent neural stem cells7 showed that iCoMoNSCs are uniquely homogenous and self-renewing. Differentiated iCoMoNSCs formed a self-organized multicellular system consisting of synaptically connected and electrophysiologically active neurons, which matured into long-lived functional networks (which we designate iNets). Neuronal and glial maturation in iNets was similar to that of cortical organoids8. Overexpression of wild-type TDP-43 in a minority of neurons within iNets led to progressive fragmentation and aggregation of the protein, resulting in a partial loss of function and neurotoxicity. Single-cell transcriptomics revealed a novel set of misregulated RNA targets in TDP-43-overexpressing neurons and in patients with TDP-43 proteinopathies exhibiting a loss of nuclear TDP-43. The strongest misregulated target encoded the synaptic protein NPTX2, the levels of which are controlled by TDP-43 binding on its 3' untranslated region. When NPTX2 was overexpressed in iNets, it exhibited neurotoxicity, whereas correcting NPTX2 misregulation partially rescued neurons from TDP-43-induced neurodegeneration. Notably, NPTX2 was consistently misaccumulated in neurons from patients with amyotrophic lateral sclerosis and frontotemporal lobar degeneration with TDP-43 pathology. Our work directly links TDP-43 misregulation and NPTX2 accumulation, thereby revealing a TDP-43-dependent pathway of neurotoxicity.

scDesign3 generates realistic in silico data for multimodal single-cell and spatial omics

We present a statistical simulator, scDesign3, to generate realistic single-cell and spatial omics data, including various cell states, experimental designs and feature modalities, by learning interpretable parameters from real data. Using a unified probabilistic model for single-cell and spatial omics data, scDesign3 infers biologically meaningful parameters; assesses the goodness-of-fit of inferred cell clusters, trajectories and spatial locations; and generates in silico negative and positive controls for benchmarking computational tools.

Semi-supervised integration of single-cell transcriptomics data

Batch effects in single-cell RNA-seq data pose a significant challenge for comparative analyses across samples, individuals, and conditions. Although batch effect correction methods are routinely applied, data integration often leads to overcorrection and can result in the loss of biological variability. In this work we present STACAS, a batch correction method for scRNA-seq that leverages prior knowledge on cell types to preserve biological variability upon integration. Through an open-source benchmark, we show that semi-supervised STACAS outperforms state-of-the-art unsupervised methods, as well as supervised methods such as scANVI and scGen. STACAS scales well to large datasets and is robust to incomplete and imprecise input cell type labels, which are commonly encountered in real-life integration tasks. We argue that the incorporation of prior cell type information should be a common practice in single-cell data integration, and we provide a flexible framework for semi-supervised batch effect correction.

Single-cell characterization of human GBM reveals regional differences in tumor-infiltrating leukocyte activation

Glioblastoma (GBM) harbors a highly immunosuppressive tumor microenvironment (TME) which influences glioma growth. Major efforts have been undertaken to describe the TME on a single-cell level. However, human data on regional differences within the TME remain scarce. Here, we performed high-depth single-cell RNA sequencing (scRNAseq) on paired biopsies from the tumor center, peripheral infiltration zone and blood of five primary GBM patients. Through analysis of >45,000 cells, we revealed a regionally distinct transcription profile of microglia (MG) and monocyte-derived macrophages (MdMs) and an impaired activation signature in the tumor-peripheral cytotoxic-cell compartment. Comparing tumor-infiltrating CD8+ T cells with circulating cells identified CX3CR1high and CX3CR1int CD8+ T cells with effector and memory phenotype, respectively, enriched in blood but absent in the TME. Tumor CD8+ T cells displayed a tissue-resident memory phenotype with dysfunctional features. Our analysis provides a regionally resolved mapping of transcriptional states in GBM-associated leukocytes, serving as an additional asset in the effort towards novel therapeutic strategies to combat this fatal disease.

The Simple prEservatioN of Single cElls method for cryopreservation enables the generation of single-cell immune profiles from whole blood

Introduction

Current multistep methods utilized for preparing and cryopreserving single-cell suspensions from blood samples for single-cell RNA sequencing (scRNA-seq) are time-consuming, requiring trained personnel and special equipment, so limiting their clinical adoption. We developed a method, Simple prEservatioN of Single cElls (SENSE), for single-step cryopreservation of whole blood (WB) along with granulocyte depletion during single-cell assay, to generate high quality single-cell profiles (SCP).

Methods

WB was cryopreserved using the SENSE method and peripheral blood mononuclear cells (PBMCs) were isolated and cryopreserved using the traditional density-gradient method (PBMC method) from the same blood sample (n=6). The SCPs obtained from both methods were processed using a similar pipeline and quality control parameters. Further, entropy calculation, differential gene expression, and cellular communication analysis were performed to compare cell types and subtypes from both methods.

Results

Highly viable (86.3 ± 1.51%) single-cell suspensions (22,353 cells) were obtained from the six WB samples cryopreserved using the SENSE method. In-depth characterization of the scRNA-seq datasets from the samples processed with the SENSE method yielded high-quality profiles of lymphoid and myeloid cell types which were in concordance with the profiles obtained with classical multistep PBMC method processed samples. Additionally, the SENSE method cryopreserved samples exhibited significantly higher T-cell enrichment, enabling deeper characterization of T-cell subtypes. Overall, the SENSE and PBMC methods processed samples exhibited transcriptional, and cellular communication network level similarities across cell types with no batch effect except in myeloid lineage cells.

Discussion

Comparative analysis of scRNA-seq datasets obtained with the two cryopreservation methods i.e., SENSE and PBMC methods, yielded similar cellular and molecular profiles, confirming the suitability of the former method's incorporation in clinics/labs for cryopreserving and obtaining high-quality single-cells for conducting critical translational research.

Transcriptomes and metabolism define mouse and human MAIT cell populations

Mucosal-associated invariant T (MAIT) cells are a subset of T lymphocytes that respond to microbial metabolites. We defined MAIT cell populations in different organs and characterized the developmental pathway of mouse and human MAIT cells in the thymus using single-cell RNA sequencing and phenotypic and metabolic analyses. We showed that the predominant mouse subset, which produced IL-17 (MAIT17), and the subset that produced IFN-γ (MAIT1) had not only greatly different transcriptomes but also different metabolic states. MAIT17 cells in different organs exhibited increased lipid uptake, lipid storage, and mitochondrial potential compared with MAIT1 cells. All these properties were similar in the thymus and likely acquired there. Human MAIT cells in lung and blood were more homogeneous but still differed between tissues. Human MAIT cells had increased fatty acid uptake and lipid storage in blood and lung, similar to human CD8 T resident memory cells, but unlike mouse MAIT17 cells, they lacked increased mitochondrial potential. Although mouse and human MAIT cell transcriptomes showed similarities for immature cells in the thymus, they diverged more strikingly in the periphery. Analysis of pet store mice demonstrated decreased lung MAIT17 cells in these so-called "dirty" mice, indicative of an environmental influence on MAIT cell subsets and function.

A structured evaluation of cryopreservation in generating single-cell transcriptomes from cerebrospinal fluid

Single-cell transcriptomics allows characterization of cerebrospinal fluid (CSF) cells at an unprecedented level. Here, we report a robust cryopreservation protocol adapted for the characterization of fragile CSF cells by single-cell RNA sequencing (RNA-seq) in moderate- to large-scale studies. Fresh CSF was collected from twenty-one participants at two independent sites. Each CSF sample was split into two fractions: one was processed fresh, while the second was cryopreserved for months and profiled after thawing. B and T cell receptor sequencing was also performed. Our comparison of fresh and cryopreserved data from the same individuals demonstrates highly efficient recovery of all known CSF cell types. We find no significant difference in cell type proportions and cellular transcriptomes between fresh and cryopreserved cells. Results were comparable at both sites and with different single-cell sequencing chemistries. Cryopreservation did not affect recovery of T and B cell clonotype diversity. Our CSF cell cryopreservation protocol provides an important alternative to fresh processing of fragile CSF cells.

Meta-analysis of (single-cell method) benchmarks reveals the need for extensibility and interoperability

Computational methods represent the lifeblood of modern molecular biology. Benchmarking is important for all methods, but with a focus here on computational methods, benchmarking is critical to dissect important steps of analysis pipelines, formally assess performance across common situations as well as edge cases, and ultimately guide users on what tools to use. Benchmarking can also be important for community building and advancing methods in a principled way. We conducted a meta-analysis of recent single-cell benchmarks to summarize the scope, extensibility, and neutrality, as well as technical features and whether best practices in open data and reproducible research were followed. The results highlight that while benchmarks often make code available and are in principle reproducible, they remain difficult to extend, for example, as new methods and new ways to assess methods emerge. In addition, embracing containerization and workflow systems would enhance reusability of intermediate benchmarking results, thus also driving wider adoption.

The shaky foundations of simulating single-cell RNA sequencing data

BACKGROUND

With the emergence of hundreds of single-cell RNA-sequencing (scRNA-seq) datasets, the number of computational tools to analyze aspects of the generated data has grown rapidly. As a result, there is a recurring need to demonstrate whether newly developed methods are truly performant-on their own as well as in comparison to existing tools. Benchmark studies aim to consolidate the space of available methods for a given task and often use simulated data that provide a ground truth for evaluations, thus demanding a high quality standard results credible and transferable to real data.

RESULTS

Here, we evaluated methods for synthetic scRNA-seq data generation in their ability to mimic experimental data. Besides comparing gene- and cell-level quality control summaries in both one- and two-dimensional settings, we further quantified these at the batch- and cluster-level. Secondly, we investigate the effect of simulators on clustering and batch correction method comparisons, and, thirdly, which and to what extent quality control summaries can capture reference-simulation similarity.

CONCLUSIONS

Our results suggest that most simulators are unable to accommodate complex designs without introducing artificial effects, they yield over-optimistic performance of integration and potentially unreliable ranking of clustering methods, and it is generally unknown which summaries are important to ensure effective simulation-based method comparisons.

Temporal progress of gene expression analysis with RNA-Seq data: A review on the relationship between computational methods

Analysis of differential gene expression from RNA-seq data has become a standard for several research areas. The steps for the computational analysis include many data types and file formats, and a wide variety of computational tools that can be applied alone or together as pipelines. This paper presents a review of the differential expression analysis pipeline, addressing its steps and the respective objectives, the principal methods available in each step, and their properties, therefore introducing an organized overview to this context. This review aims to address mainly the aspects involved in the differentially expressed gene (DEG) analysis from RNA sequencing data (RNA-seq), considering the computational methods. In addition, a timeline of the computational methods for DEG is shown and discussed, and the relationships existing between the most important computational tools are presented by an interaction network. A discussion on the challenges and gaps in DEG analysis is also highlighted in this review. This paper will serve as a tutorial for new entrants into the field and help established users update their analysis pipelines.

Single-cell transcriptional regulation and genetic evolution of neuroendocrine prostate cancer

Neuroendocrine prostate cancer (NEPC) is a lethal subtype of prostate cancer, with a 10% five-year survival rate. However, little is known about its origin and the mechanisms governing its emergence. Our study characterized ADPC and NEPC in prostate tumors from 7 patients using scRNA-seq. First, we identified two NEPC gene expression signatures representing different phases of trans-differentiation. New marker genes we identified may be used for clinical diagnosis. Second, integrative analyses combining expression and subclonal architecture revealed different paths by which NEPC diverges from the original ADPC, either directly from treatment-naïve tumor cells or from specific intermediate states of treatment-resistance. Third, we inferred a hierarchical transcription factor (TF) network underlying the progression, which involves constitutive regulation by ASCL1, FOXA2, and selective regulation by NKX2-2, POU3F2, and SOX2. Together, these results defined the complex expression profiles and advanced our understanding of the genetic and transcriptomic mechanisms leading to NEPC differentiation.

•

Single-cell RNA sequencing revealed two distinct transcriptional programs of NEPC

•

Cell-level clonal evolution analysis extended the divergent model of ADPC to NEPC

•

Screening of NEPC-specific transcription factors through network-based approaches

Single-cell transcriptomics and surface epitope detection in human brain epileptic lesions identifies pro-inflammatory signaling

Epileptogenic triggers are multifactorial and not well understood. Here we aimed to address the hypothesis that inappropriate pro-inflammatory mechanisms contribute to the pathogenesis of refractory epilepsy (non-responsiveness to antiepileptic drugs) in human patients. We used single-cell cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) to reveal the immunotranscriptome of surgically resected epileptic lesion tissues. Our approach uncovered a pro-inflammatory microenvironment, including extensive activation of microglia and infiltration of other pro-inflammatory immune cells. These findings were supported by ligand–receptor (LR) interactome analysis, which demonstrated potential mechanisms of infiltration and evidence of direct physical interactions between microglia and T cells. Together, these data provide insight into the immune microenvironment in epileptic tissue, which may aid the development of new therapeutics.

Single-cell analysis of immune cells from surgically resected human epileptic brain tissues showed heterogeneity and pro-inflammatory signaling in microglia and evidence for direct interaction of microglia with T cells.

Doublet identification in single-cell sequencing data using scDblFinder

Doublets are prevalent in single-cell sequencing data and can lead to artifactual findings. A number of strategies have therefore been proposed to detect them. Building on the strengths of existing

approaches, we developed scDblFinder, a fast, flexible and accurate Bioconductor-based doublet detection method. Here we present the method, justify its design choices, demonstrate its performance on both single-cell RNA and accessibility (ATAC) sequencing data, and provide some observations on doublet formation, detection, and enrichment analysis. Even in complex datasets, scDblFinder can accurately identify most heterotypic doublets, and was already found by an independent benchmark to outcompete alternatives.

Single-cell eQTL analysis of activated T cell subsets reveals activation and cell type-dependent effects of disease-risk variants

The impact of genetic variants on cells challenged in biologically relevant contexts has not been fully explored. Here, we activated CD4⁺ T cells from 89 healthy donors and performed a single-cell RNA sequencing assay with >1 million cells to examine cell type-specific and activation-dependent effects of genetic variants. Single-cell expression quantitative trait loci (sc-eQTL) analysis of 19 distinct CD4⁺ T cell subsets showed that the expression of over 4000 genes is significantly associated with common genetic polymorphisms and that most of these genes show their most prominent effects in specific cell types. These genes included many that encode for molecules important for activation, differentiation, and effector functions of T cells. We also found new gene associations for disease-risk variants identified from genome-wide association studies and highlighted the cell types in which their effects are most prominent. We found that biological sex has a major influence on activation-dependent gene expression in CD4⁺ T cell subsets. Sex-biased transcripts were significantly enriched in several pathways that are essential for the initiation and execution of effector functions by CD4⁺ T cells like TCR signaling, cytokines, cytokine receptors, costimulatory, apoptosis, and cell-cell adhesion pathways. Overall, this DICE (Database of Immune Cell Expression, eQTLs, and Epigenomics) subproject highlights the power of sc-eQTL studies for simultaneously exploring the activation and cell type-dependent effects of common genetic variants on gene expression (https://dice-database.org).

Doublet identification in single-cell sequencing data using scDblFinder

Doublets are prevalent in single-cell sequencing data and can lead to artifactual findings. A number of strategies have therefore been proposed to detect them. Building on the strengths of existing approaches, we developed scDblFinder, a fast, flexible and accurate Bioconductor-based doublet detection method. Here we present the method, justify its design choices, demonstrate its performance on both single-cell RNA and accessibility (ATAC) sequencing data, and provide some observations on doublet formation, detection, and enrichment analysis. Even in complex datasets, scDblFinder can accurately identify most heterotypic doublets, and was already found by an independent benchmark to outcompete alternatives.

Differentiation and activation of fibroblastic reticular cells

Secondary lymphoid organs (SLO) are underpinned by fibroblastic reticular cells (FRC) that form dedicated microenvironmental niches to secure induction and regulation of innate and adaptive immunity. Distinct FRC subsets are strategically positioned in SLOs to provide niche factors and govern efficient immune cell interaction. In recent years, the use of specialized mouse models in combination with single-cell transcriptomics has facilitated the elaboration of the molecular FRC landscape at an unprecedented resolution. While single-cell RNA-sequencing has advanced the resolution of FRC subset characterization and function, the high dimensionality of the generated data necessitates careful analysis and validation. Here, we reviewed novel findings from high-resolution transcriptomic analyses that refine our understanding of FRC differentiation and activation processes in the context of infection and inflammation. We further discuss concepts, strategies, and limitations for the analysis of single-cell transcriptome data from FRCs and the wide-ranging implications for our understanding of stromal cell biology.