Journal: Nature Communications

Article Title: MetaQ: fast, scalable and accurate metacell inference via single-cell quantization

doi: 10.1038/s41467-025-56424-6

Figure Lengend Snippet: a UMAP visualization of the original 433,495 cells from the human fetal atlas. The red box highlights retina cells. b Classification accuracy of cell classifiers trained with [500, 1000, 2000, 4000] metacells inferred by MetaQ, SEACell, MetaCell V2, SuperCell, and random sub-sampling on five random experiments. Each boxplot ranges from the upper and lower quartiles with the median as the horizontal line and whiskers extend to 1.5 times the interquartile range. Two-sided T-test results: *0.01 < p ≤ 0.05, **0.001 < p ≤ 0.01, ***0.0001 < p ≤ 0.001, **** p ≤ 0.0001. c UMAP visualization of 4000 metacells inferred by the four methods, with cell type colors matching those in b . Retina cells are marked with red boxes. d Agreement between the ground-truth annotations and the labels predicted by classification models trained with 500 metacells. Matrices with a clearer diagonal structure indicate better classification performance. e Compactness of [500, 1000, 2000, 4000] metacells inferred by different methods on five random experiments. Each boxplot ranges from the upper and lower quartiles with the median as the horizontal line and whiskers extend to 1.5 times the interquartile range. Two-sided T-test results: **** p ≤ 0.0001. f Separation of [500, 1000, 2000, 4000] metacells inferred by different methods on five random experiments. Each boxplot ranges from the upper and lower quartiles with the median as the horizontal line and whiskers extend to 1.5 times the interquartile range. g Running times (logged) and memory cost for inferring 1000 metacells from different numbers of original cells. h Running times (logged) and memory cost for inferring different numbers of metacells from 100,000 cells. Source data are provided as a Source Data file.

Article Snippet: For MetaCell V2, we used its Python package ( https://github.com/tanaylab/metacells ), v.0.9.4.

Techniques: Sampling

Journal: Nature Communications

Article Title: MetaQ: fast, scalable and accurate metacell inference via single-cell quantization

doi: 10.1038/s41467-025-56424-6

Figure Lengend Snippet: a UMAP visualization of the original 30,672 cells from human bone marrow in RNA and ADT modalities (CD4 Memory Memory CD4+ T cell, CD4 Naive Naive CD4+ T cell, CD8 Effector Effector CD8+ T cell, CD8 Memory Memory CD8+ T cell, CD8 Naive Naive CD8+ T cell, CD14 Mono CD14 Monocytes, CD16 Mono CD16 Monocytes, CD56 bright NK CD56 bright natural killer, GMP Granulocyte-macrophage progenitors, HSC Hematopoietic stem cell, LMPP Lymphoid-primed multipotent progenitors, MAIT Mucosal-associated invariant T cell, NK Natural killer, Prog_B Progenitors of B cell, Prog_DC Progenitors of dendritic cell lineages, Prog_Mk Progenitor of megakaryocyte, Prog_RBC Progenitors of erythroid, Treg Regulatory T cell, cDC2 Type 2 conventional dendritic cell, gdT Gamma delta T cell, pDC Plasmacytoid dendritic cell). b UMAP visualization of WNN results on original cells and 613 metacells (a 50-fold reduction) inferred by MetaQ, SEACell, MetaCell V2, and SuperCell. c Compactness and separation of 613 metacells inferred by different methods, calculated separately for RNA and ADT modalities. Each boxplot ranges from the upper and lower quartiles with the median as the horizontal line and whiskers extend to 1.5 times the interquartile range. d Purity of 613 metacells inferred by different methods across RNA-informative (top panel) and ADT-informative (bottom panel) cell types. Each boxplot ranges from the upper and lower quartiles with the median as the horizontal line and whiskers extend to 1.5 times the interquartile range. Two-sided T-test results: *0.01 < p ≤ 0.05, **0.001 < p ≤ 0.01, ***0.0001 < p ≤ 0.001, **** p ≤ 0.0001. e PAGA cell embedding of metacells along the plasmablast developmental path inferred by MetaQ, with metacells colored by type and pseudotime, respectively. f Annotation and marker gene expression changes along the plasmablast developmental path. Source data are provided as a Source Data file.

Article Snippet: For MetaCell V2, we used its Python package ( https://github.com/tanaylab/metacells ), v.0.9.4.

Techniques: Marker, Gene Expression

Journal: Nature Communications

Article Title: MetaQ: fast, scalable and accurate metacell inference via single-cell quantization

doi: 10.1038/s41467-025-56424-6

Figure Lengend Snippet: a UMAP visualization of the original 14,767 cells from the human pancreas dataset, with cells colored by types and batches, respectively. b UMAP visualization of integrated cell embeddings obtained by performing Harmony on the original data (MHC class II: Major histocompatibility complex Class II). c UMAP visualization of 590 MetaQ metacells (a 25-fold reduction) integrated by Harmony. d UMAP visualization of the integrated embeddings of original cells recovered by MetaQ. e Sankey plots showing Louvain cluster assignments on cell embeddings integrated by Harmony, recovered by SuperCell, and recovered by MetaQ. f Normalized expression of the marker gene TM4SF4 projected on the UMAP plot of alpha cell embeddings by MetaQ and Harmony, respectively. The subplot in the left plot shows the results for the Baron batch. g AMI, ARI, and Homogeneity scores of Louvain clustering with three different resolutions of [0.5, 1.0, 2.0] on cell embeddings obtained by Harmony and recovered by MetaQ, SEACell, MetaCell V2, and SuperCell. h AMI, ARI, and Homogeneity scores of Louvain clustering with three different resolutions of [1.0, 2.0, 5.0] on 590 metacells inferred by different metacell algorithms. Source data are provided as a Source Data file.

Article Snippet: For MetaCell V2, we used its Python package ( https://github.com/tanaylab/metacells ), v.0.9.4.

Techniques: Immunopeptidomics, Expressing, Marker