Journal: Nucleic Acids Research

Article Title: Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

doi: 10.1093/nar/gkab1147

Figure Lengend Snippet: Integration of eight scRNA-seq datasets on human pancreatic islets. The examined datasets span 27 samples, five technologies, and four laboratories. The examined integration methods include DESC, Harmony, LIGER, MNN, scAlign, Scanorama, scVI, Seurat V3 and VIPCCA. UMAPs are used for visualizing integration results by either technologies ( A ) or cell types ( B ). The overall integration quality is measured by four metrics: mixing metric ( C ), kBET acceptance rate ( D ), adjusted rand index (ARI) ( E ) and differential expression analysis ( F ).

Article Snippet: We obtained 13 cell types in the scRNA-seq data using the standard workflow in Seurat ( https://www.dropbox.com/s/3f3p5nxrn5b3y4y/pbmc_10k_v3.rds?dl=1 ).

Techniques: Quantitative Proteomics

Journal: Nucleic Acids Research

Article Title: Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

doi: 10.1093/nar/gkab1147

Figure Lengend Snippet: Integration of three scRNA-seq datasets with partially overlapped cell types. The examined datasets include one with 293T cells, one with Jurkat cells, and one with a 50:50 mixture of the two cell types. The examined integration methods including DESC, Harmony, LIGER, MNN, scAlign, Scanorama, scVI, Seurat V3, and VIPCCA. UMAPs are used for visualizing integration results by either batches ( A ) or cell types ( B ). Overall integration quality is measured by four metrics: mixing metric ( C ), kBET acceptance rate ( D ), adjusted rand index (ARI) ( E ) and detection of differentially expressed genes ( F ).

Article Snippet: We obtained 13 cell types in the scRNA-seq data using the standard workflow in Seurat ( https://www.dropbox.com/s/3f3p5nxrn5b3y4y/pbmc_10k_v3.rds?dl=1 ).

Techniques:

Journal: Nucleic Acids Research

Article Title: Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

doi: 10.1093/nar/gkab1147

Figure Lengend Snippet: Integration of scRNA-seq and scATAC-seq datasets on PBMCs. ( A ) and ( B ) show UMAP visualizations of scRNA-seq and scATAC-seq data based on the cell embeddings obtained from Seurat and VIPCCA, respectively. Each dot represents a cell/nucleus colored by either datasets (left) or cell types (middle and right). We are unable to color cell types predicted by Seurat V3 because it relies on a completely different strategy to infer cell types that do not look well on the low dimensional space. ( C ) compares results between Seurat V3 and VIPCCA by visualizing the number of overlapped cells and the Jaccard index for each pair of predicted cell types. ( D ) shows the distribution of cells on the UMAP space that are enriched with unknown nuclei by VIPCCA, unknown nuclei by Seurat V3, duplicate mapped read-pairs, chimerically mapped read-pairs, read-pairs with at least one end not mapped, and fragments overlapping with TSS regions. ( E ) shows the mean duplicate mapped read-pairs and mean TSS fragments in the unassigned cells by the two methods. ( F ) shows the mean chimerically mapped read-pairs and unmapped read-pairs in the unassigned cells by the two methods.

Article Snippet: We obtained 13 cell types in the scRNA-seq data using the standard workflow in Seurat ( https://www.dropbox.com/s/3f3p5nxrn5b3y4y/pbmc_10k_v3.rds?dl=1 ).

Techniques:

Journal: Nucleic Acids Research

Article Title: Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

doi: 10.1093/nar/gkab1147

Figure Lengend Snippet: Integration of two datasets of human male germline cells that were collected on a series of time points from 4 weeks to 25 weeks. The examined integration methods include DESC, Harmony, LIGER, MNN, Scanorama, scVI, Seurat V3 and VIPCCA. Data are visualized by either batches ( A ) or collection time ( B ) on the UMAP space before integration and after integration using different integration methods. Slingshot was applied to perform trajectory inference based on the cell embeddings in reduced dimensional space inferred by each alignment method. Integration quality is further evaluated by the mixing metric ( C ) and kBET acceptance rate ( D ).

Article Snippet: We obtained 13 cell types in the scRNA-seq data using the standard workflow in Seurat ( https://www.dropbox.com/s/3f3p5nxrn5b3y4y/pbmc_10k_v3.rds?dl=1 ).

Techniques: