Robust integration of multiple single-cell RNA sequencing datasets using a single reference space

Liu, Yang; Wang, Tao; Zhou, Bin; Zheng, Deyou

doi:10.1038/s41587-021-00859-x

Cited by 38 publications

(43 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…4a-c and Supplementary Data 2). Data integration provides consistency in defining common cell types among different samples, in addition to removing batch effects 37 . We found that the CPM can be identified in both populations (Fig.…”

Section: Resultsmentioning

confidence: 99%

“…The Mesp1 Cre Ctrl data from different stages are comprised of homogeneous cell types (transcriptomes of each cell type are concordant across datasets), but the Mesp1 Cre Ctrl with cKO and then separately, the Tbx1 Cre Ctrl versus cKO, are comprised of heterogeneous cell types (the same cell types from different conditions, Ctrl vs cKO, with dissimilar gene expression). Thus, we aligned gene expression values for Ctrl vs cKO data using the RPCI (reference principal component integration) method in RISC, which utilizes the global gene reference to calibrate the gene expression changes of heterogeneous cell types 37 . In detail, we combined individual datasets by the scMultiIntegrate function of the RISC package and outputted the corrected gene expression values after the clustering by Seurat 3.1.5 82 .…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Single cell multi-omic analysis identifies a Tbx1-dependent multilineage primed population in murine cardiopharyngeal mesoderm

et al. 2021

Self Cite

View full text Add to dashboard Cite

The poles of the heart and branchiomeric muscles of the face and neck are formed from the cardiopharyngeal mesoderm within the pharyngeal apparatus. They are disrupted in patients with 22q11.2 deletion syndrome, due to haploinsufficiency of TBX1, encoding a T-box transcription factor. Here, using single cell RNA-sequencing, we now identify a multilineage primed population within the cardiopharyngeal mesoderm, marked by Tbx1, which has bipotent properties to form cardiac and branchiomeric muscle cells. The multilineage primed cells are localized within the nascent mesoderm of the caudal lateral pharyngeal apparatus and provide a continuous source of cardiopharyngeal mesoderm progenitors. Tbx1 regulates the maturation of multilineage primed progenitor cells to cardiopharyngeal mesoderm derivatives while restricting ectopic non-mesodermal gene expression. We further show that TBX1 confers this balance of gene expression by direct and indirect regulation of enriched genes in multilineage primed progenitors and downstream pathways, partly through altering chromatin accessibility, the perturbation of which can lead to congenital defects in individuals with 22q11.2 deletion syndrome.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Single cell multi-omic analysis identifies a Tbx1-dependent multilineage primed population in murine cardiopharyngeal mesoderm

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…After individual samples were analyzed by Seurat for clustering, the data were integrated by the RISC software (v1.5) using the Reference Principal Component Integration (RPCI) algorithm for removing batch effects and aligning gene expression values between the control and Tbx1 null samples at E9.5 and E10.5 29 . The integrated data were re-clustered by RISC, using parameters adjusted to match the cell type clusters in the Seurat results.…”

Section: Methodsmentioning

confidence: 99%

Single cell transcriptomics uncovers a non-autonomous Tbx1-dependent genetic program controlling cardiac neural crest cell deployment and progression

Bono

Liu

Ferrena

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Disruption of cardiac neural crest cells (CNCCs) results in congenital heart disease, yet we do not understand the cell fate dynamics as these cells differentiate to vascular smooth muscle cells. Here we utilized single-cell RNA-sequencing of NCCs from the pharyngeal apparatus with heart in control mouse embryos and when Tbx1, the gene for 22q11.2 deletion syndrome, is inactivated. We uncovered three dynamic transitions of pharyngeal NCCs expressing Tbx2 and Tbx3 through differentiated CNCCs expressing cardiac transcription factors with smooth muscle genes, and that these transitions are altered non-autonomously by loss of Tbx1. Further, inactivation of Tbx2 and Tbx3 in early CNCCs resulted in aortic arch branching defects due to failed smooth muscle differentiation. Loss of Tbx1 interrupted mesoderm to CNCC cell-cell communication with upregulation of BMP signaling with reduced MAPK signaling and failed dynamic transitions of CNCCs leading to disruption of aortic arch artery formation and cardiac outflow tract septation.

show abstract

“…Whereas, the simpler removal of batch effects with the covariate naive zerocentering approach retained sufficient biological signal between disease labels for non-linear ML models to correctly classify samples across batches. Batch effect correction methods that do not require input of a covariate have been developed, such as frozen surrogate variable analysis or reference principal component integration (RPCI) (Parker et al, 2014;Liu et al, 2021), although their applicability to microbiome data has not been assessed.…”

Section: Discussionmentioning

confidence: 99%

Benchmark of Data Processing Methods and Machine Learning Models for Gut Microbiome-Based Diagnosis of Inflammatory Bowel Disease

Kubinski¹,

Djamen-Kepaou²,

Zhanabaev³

et al. 2022

Front. Genet.

View full text Add to dashboard Cite

Patients with inflammatory bowel disease (IBD) wait months and undergo numerous invasive procedures between the initial appearance of symptoms and receiving a diagnosis. In order to reduce time until diagnosis and improve patient wellbeing, machine learning algorithms capable of diagnosing IBD from the gut microbiome’s composition are currently being explored. To date, these models have had limited clinical application due to decreased performance when applied to a new cohort of patient samples. Various methods have been developed to analyze microbiome data which may improve the generalizability of machine learning IBD diagnostic tests. With an abundance of methods, there is a need to benchmark the performance and generalizability of various machine learning pipelines (from data processing to training a machine learning model) for microbiome-based IBD diagnostic tools. We collected fifteen 16S rRNA microbiome datasets (7,707 samples) from North America to benchmark combinations of gut microbiome features, data normalization and transformation methods, batch effect correction methods, and machine learning models. Pipeline generalizability to new cohorts of patients was evaluated with two binary classification metrics following leave-one-dataset-out cross (LODO) validation, where all samples from one study were left out of the training set and tested upon. We demonstrate that taxonomic features processed with a compositional transformation method and batch effect correction with the naive zero-centering method attain the best classification performance. In addition, machine learning models that identify non-linear decision boundaries between labels are more generalizable than those that are linearly constrained. Lastly, we illustrate the importance of generating a curated training dataset to ensure similar performance across patient demographics. These findings will help improve the generalizability of machine learning models as we move towards non-invasive diagnostic and disease management tools for patients with IBD.

show abstract

Robust integration of multiple single-cell RNA sequencing datasets using a single reference space

Cited by 38 publications

References 66 publications

Single cell multi-omic analysis identifies a Tbx1-dependent multilineage primed population in murine cardiopharyngeal mesoderm

Single cell multi-omic analysis identifies a Tbx1-dependent multilineage primed population in murine cardiopharyngeal mesoderm

Single cell transcriptomics uncovers a non-autonomous Tbx1-dependent genetic program controlling cardiac neural crest cell deployment and progression

Benchmark of Data Processing Methods and Machine Learning Models for Gut Microbiome-Based Diagnosis of Inflammatory Bowel Disease

Contact Info

Product

Resources

About