293t cells Search Results


99
ATCC hek293t
Transient splicing reporter assay using the genomic locus sequence reproduces the endogenous TREM2 and APOE isoform patterns (A) Schematic representation of the genomic TREM2 and APOE sequences cloned into the pcDNA3.1(+) vector. (B) Human cell lines expressing endogenous TREM2 (THP-1, HMC3), non-expressing <t>HEK293T</t> cells, and mouse microglial BV2 cells were transfected with the full-length (FL) TREM2 reporter. (C) HMC3 and HEK293T cells expressing endogenous APOE were transfected with the FL APOE reporter. HB, endogenous TREM2 (B) and APOE (C) transcripts from human brain (frontal cortex). cDNA synthesis was primed with either oligo(dT) or a plasmid-specific RT primer (P) positioned downstream of the gene insert. Arrows mark RT-PCR primer positions. Amplified products were resolved by agarose gel electrophoresis.
Hek293t, supplied by ATCC, used in various techniques. Bioz Stars score: 99/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pmc12990334-41-0-2?v=ATCC
Average 99 stars, based on 1 article reviews
hek293t - by Bioz Stars, 2026-07
99/100 stars
  Buy from Supplier

94
Genecopoeia human embryonic kidney 293t 293t cells
Transient splicing reporter assay using the genomic locus sequence reproduces the endogenous TREM2 and APOE isoform patterns (A) Schematic representation of the genomic TREM2 and APOE sequences cloned into the pcDNA3.1(+) vector. (B) Human cell lines expressing endogenous TREM2 (THP-1, HMC3), non-expressing <t>HEK293T</t> cells, and mouse microglial BV2 cells were transfected with the full-length (FL) TREM2 reporter. (C) HMC3 and HEK293T cells expressing endogenous APOE were transfected with the FL APOE reporter. HB, endogenous TREM2 (B) and APOE (C) transcripts from human brain (frontal cortex). cDNA synthesis was primed with either oligo(dT) or a plasmid-specific RT primer (P) positioned downstream of the gene insert. Arrows mark RT-PCR primer positions. Amplified products were resolved by agarose gel electrophoresis.
Human Embryonic Kidney 293t 293t Cells, supplied by Genecopoeia, used in various techniques. Bioz Stars score: 94/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pm36610792-46-0-31?v=Genecopoeia
Average 94 stars, based on 1 article reviews
human embryonic kidney 293t 293t cells - by Bioz Stars, 2026-07
94/100 stars
  Buy from Supplier

93
Beijing Solarbio Science 293t cells
Transient splicing reporter assay using the genomic locus sequence reproduces the endogenous TREM2 and APOE isoform patterns (A) Schematic representation of the genomic TREM2 and APOE sequences cloned into the pcDNA3.1(+) vector. (B) Human cell lines expressing endogenous TREM2 (THP-1, HMC3), non-expressing <t>HEK293T</t> cells, and mouse microglial BV2 cells were transfected with the full-length (FL) TREM2 reporter. (C) HMC3 and HEK293T cells expressing endogenous APOE were transfected with the FL APOE reporter. HB, endogenous TREM2 (B) and APOE (C) transcripts from human brain (frontal cortex). cDNA synthesis was primed with either oligo(dT) or a plasmid-specific RT primer (P) positioned downstream of the gene insert. Arrows mark RT-PCR primer positions. Amplified products were resolved by agarose gel electrophoresis.
293t Cells, supplied by Beijing Solarbio Science, used in various techniques. Bioz Stars score: 93/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pm40191097-54-22-56?v=Beijing+Solarbio+Science
Average 93 stars, based on 1 article reviews
293t cells - by Bioz Stars, 2026-07
93/100 stars
  Buy from Supplier

93
Elabscience Biotechnology 293t cells
Transient splicing reporter assay using the genomic locus sequence reproduces the endogenous TREM2 and APOE isoform patterns (A) Schematic representation of the genomic TREM2 and APOE sequences cloned into the pcDNA3.1(+) vector. (B) Human cell lines expressing endogenous TREM2 (THP-1, HMC3), non-expressing <t>HEK293T</t> cells, and mouse microglial BV2 cells were transfected with the full-length (FL) TREM2 reporter. (C) HMC3 and HEK293T cells expressing endogenous APOE were transfected with the FL APOE reporter. HB, endogenous TREM2 (B) and APOE (C) transcripts from human brain (frontal cortex). cDNA synthesis was primed with either oligo(dT) or a plasmid-specific RT primer (P) positioned downstream of the gene insert. Arrows mark RT-PCR primer positions. Amplified products were resolved by agarose gel electrophoresis.
293t Cells, supplied by Elabscience Biotechnology, used in various techniques. Bioz Stars score: 93/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pm40476564-50-0-31?v=Elabscience+Biotechnology
Average 93 stars, based on 1 article reviews
293t cells - by Bioz Stars, 2026-07
93/100 stars
  Buy from Supplier

96
Proteintech hek293t cells
Transient splicing reporter assay using the genomic locus sequence reproduces the endogenous TREM2 and APOE isoform patterns (A) Schematic representation of the genomic TREM2 and APOE sequences cloned into the pcDNA3.1(+) vector. (B) Human cell lines expressing endogenous TREM2 (THP-1, HMC3), non-expressing <t>HEK293T</t> cells, and mouse microglial BV2 cells were transfected with the full-length (FL) TREM2 reporter. (C) HMC3 and HEK293T cells expressing endogenous APOE were transfected with the FL APOE reporter. HB, endogenous TREM2 (B) and APOE (C) transcripts from human brain (frontal cortex). cDNA synthesis was primed with either oligo(dT) or a plasmid-specific RT primer (P) positioned downstream of the gene insert. Arrows mark RT-PCR primer positions. Amplified products were resolved by agarose gel electrophoresis.
Hek293t Cells, supplied by Proteintech, used in various techniques. Bioz Stars score: 96/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pmc12373839__41467_2025_62751_MOESM1_ESM-2-3-12?v=Proteintech
Average 96 stars, based on 1 article reviews
hek293t cells - by Bioz Stars, 2026-07
96/100 stars
  Buy from Supplier

93
Novus Biologicals h00085377
Reagents and tools table
H00085377, supplied by Novus Biologicals, used in various techniques. Bioz Stars score: 93/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pmc11790967-37-7-5?v=Novus+Biologicals
Average 93 stars, based on 1 article reviews
h00085377 - by Bioz Stars, 2026-07
93/100 stars
  Buy from Supplier

95
Elabscience Biotechnology hek 293t cell line
Reagents and tools table
Hek 293t Cell Line, supplied by Elabscience Biotechnology, used in various techniques. Bioz Stars score: 95/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pmc12960772-127-1-7?v=Elabscience+Biotechnology
Average 95 stars, based on 1 article reviews
hek 293t cell line - by Bioz Stars, 2026-07
95/100 stars
  Buy from Supplier

90
R&D Systems caspase 2
Reagents and tools table
Caspase 2, supplied by R&D Systems, used in various techniques. Bioz Stars score: 90/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pm25188024-112-18-20?v=R%26D+Systems
Average 90 stars, based on 1 article reviews
caspase 2 - by Bioz Stars, 2026-07
90/100 stars
  Buy from Supplier

90
R&D Systems thrombospondin
Reagents and tools table
Thrombospondin, supplied by R&D Systems, used in various techniques. Bioz Stars score: 90/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/10__1074_slash_jbc__m112__444810-110-17-21?v=R%26D+Systems
Average 90 stars, based on 1 article reviews
thrombospondin - by Bioz Stars, 2026-07
90/100 stars
  Buy from Supplier

93
Sino Biological hek293t hace2 tmprss2 cells
Vaccine design and expression of the mRNA vaccine encoding COVID-19 spike. (A) Structure of HC009 RNA. UTR, untranslated region; CDS, coding domain sequence. (B) Lipid nanoparticle–mRNA formulations used as COVID-19 vaccines. (C) Spike protein expression by flow cytometry using biotinylated <t>hACE2</t> protein. (D) Spike protein expression analyzed by Western blot using anti-spike as the primary antibody. The asterisk indicates a non-specific band. All data are representative of three independent experiments.
Hek293t Hace2 Tmprss2 Cells, supplied by Sino Biological, used in various techniques. Bioz Stars score: 93/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pmc11310568-26-10-12?v=Sino+Biological
Average 93 stars, based on 1 article reviews
hek293t hace2 tmprss2 cells - by Bioz Stars, 2026-07
93/100 stars
  Buy from Supplier

94
Sino Biological 293t ace2 cell line
Vaccine design and expression of the mRNA vaccine encoding COVID-19 spike. (A) Structure of HC009 RNA. UTR, untranslated region; CDS, coding domain sequence. (B) Lipid nanoparticle–mRNA formulations used as COVID-19 vaccines. (C) Spike protein expression by flow cytometry using biotinylated <t>hACE2</t> protein. (D) Spike protein expression analyzed by Western blot using anti-spike as the primary antibody. The asterisk indicates a non-specific band. All data are representative of three independent experiments.
293t Ace2 Cell Line, supplied by Sino Biological, used in various techniques. Bioz Stars score: 94/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pm38001915-62-0-13?v=Sino+Biological
Average 94 stars, based on 1 article reviews
293t ace2 cell line - by Bioz Stars, 2026-07
94/100 stars
  Buy from Supplier

86
10X Genomics jurkat cells
(A) Geometric sketching yields more even coverage of the transcriptomic space. In our experiments, the Hausdorff distance measures the maximum distance from any point in the dataset to its closest point in the sketch; a lower Hausdorff distance indicates that the points represented by a sketch are in general closer to all of the points in the remainder of the dataset. Geometric sketching results in consistently lower Hausdorff distances than other sampling methods across a large number of sketch sizes and datasets. We use a robust Hausdorff distance that is less sensitive to small numbers of outlier observations (Method Details). Solid lines indicate means and shaded areas indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). (B) Geometric sketches contain more balanced summaries of the transcriptional landscape. t-SNE visualizations of sketches containing 2% of the cells from the adult mouse brain (Saunders et al., 2018) and from the developing and adolescent mouse CNS (Zeisel et al., 2018) using uniform random sampling and geometric sketching, with increased representation of rare cell types in the geometric sketch. Numbers of cells from each cell type are given in Tables S3–S4. Uniform sampling, which does not evenly consider the transcriptional space, produces visualizations that are poor at capturing transcriptional heterogeneity. Geometric sketching substantially underrepresents oligodendrocytes in both datasets compared to uniform sampling, which is expected given the low transcriptional heterogeneity among oligodendrocytes as quantified by differential entropy (Method Details; Tables S3–S4). Visualizations based on other sampling approaches as well as a different visualization method are provided in Figure S1. (C) Geometric sketches preserve rare cell types in the subsampled data. In sketches containing 2% of the total dataset, we counted the number of cells that belong to the rarest cell type in each <t>dataset:</t> <t>293T</t> cells (0.66% of total cells) in a <t>293T/Jurkat</t> mixture, dendritic cells (0.38% of total) in a dataset of 68k PBMCs, macrophages (0.25% of total) in a dataset of adult mouse brain cells, and ependymal cells (0.60% of total) in a dataset of developing and adolescent mouse CNS cells. Higher count indicates increased representation of the rare cell type in the sketch. Bar height indicates means and error bars indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). Comparison of rare cell type representation over different sketch sizes is shown in Figure S2B. (D) Geometric sketching is consistently effective at distinguishing biological cell types via clustering. Louvain clustering was applied to a subsample of the dataset, cluster labels were transferred to the full dataset using a k-nearest-neighbor classifier fit to the sketch, and the balanced adjusted mutual information (BAMI) was measured between the unsupervised cluster labels and the labels corresponding to biological clusters provided by each previous study (Method Details). Higher score indicates greater agreement between unsupervised clustering and biological cell type labels. Solid lines indicate means and shaded areas indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). Unsupervised clustering of geometric sketches consistently recapitulates biological cell types better than clustering results obtained by uniform sampling. Other non-uniform sampling methods, k-means++ and SRS, show performance comparable to ours in a few cases, but only geometric sketching obtains competitive performance across all settings. Because samples are drawn without replacement, clustering accuracy may approach that of uniform sampling as the sketch size increases, as is the case in the 293T/Jurkat mixture experiments.
Jurkat Cells, supplied by 10X Genomics, used in various techniques. Bioz Stars score: 86/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/293t+cells/pmc06597305-384-8-11?v=10X+Genomics
Average 86 stars, based on 1 article reviews
jurkat cells - by Bioz Stars, 2026-07
86/100 stars
  Buy from Supplier

Image Search Results


Transient splicing reporter assay using the genomic locus sequence reproduces the endogenous TREM2 and APOE isoform patterns (A) Schematic representation of the genomic TREM2 and APOE sequences cloned into the pcDNA3.1(+) vector. (B) Human cell lines expressing endogenous TREM2 (THP-1, HMC3), non-expressing HEK293T cells, and mouse microglial BV2 cells were transfected with the full-length (FL) TREM2 reporter. (C) HMC3 and HEK293T cells expressing endogenous APOE were transfected with the FL APOE reporter. HB, endogenous TREM2 (B) and APOE (C) transcripts from human brain (frontal cortex). cDNA synthesis was primed with either oligo(dT) or a plasmid-specific RT primer (P) positioned downstream of the gene insert. Arrows mark RT-PCR primer positions. Amplified products were resolved by agarose gel electrophoresis.

Journal: STAR Protocols

Article Title: Protocol for validating computationally predicted splice-altering variants using full-length gene reporter assays

doi: 10.1016/j.xpro.2026.104433

Figure Lengend Snippet: Transient splicing reporter assay using the genomic locus sequence reproduces the endogenous TREM2 and APOE isoform patterns (A) Schematic representation of the genomic TREM2 and APOE sequences cloned into the pcDNA3.1(+) vector. (B) Human cell lines expressing endogenous TREM2 (THP-1, HMC3), non-expressing HEK293T cells, and mouse microglial BV2 cells were transfected with the full-length (FL) TREM2 reporter. (C) HMC3 and HEK293T cells expressing endogenous APOE were transfected with the FL APOE reporter. HB, endogenous TREM2 (B) and APOE (C) transcripts from human brain (frontal cortex). cDNA synthesis was primed with either oligo(dT) or a plasmid-specific RT primer (P) positioned downstream of the gene insert. Arrows mark RT-PCR primer positions. Amplified products were resolved by agarose gel electrophoresis.

Article Snippet: HEK293T , ATCC , CRL-3216.

Techniques: Reporter Assay, Sequencing, Clone Assay, Plasmid Preparation, Expressing, Transfection, cDNA Synthesis, Reverse Transcription Polymerase Chain Reaction, Amplification, Agarose Gel Electrophoresis

Reagents and tools table

Journal: The EMBO Journal

Article Title: A novel human fetal lung-derived alveolar organoid model reveals mechanisms of surfactant protein C maturation relevant to interstitial lung disease

doi: 10.1038/s44318-024-00328-6

Figure Lengend Snippet: Reagents and tools table

Article Snippet: Mouse anti-Mical-L1 , 1:100 , Novus , H00085377.

Techniques: Recombinant, Transduction, CRISPR, Gene Knockout, Expressing, Concentration Assay, Immunofluorescence, Western Blot, Flow Cytometry, Sequencing, Red Blood Cell Lysis, Cell Recovery, Plasmid Preparation, SYBR Green Assay, Reverse Transcription, Software, Microscopy, Magnetic Beads, Transmission Assay

Vaccine design and expression of the mRNA vaccine encoding COVID-19 spike. (A) Structure of HC009 RNA. UTR, untranslated region; CDS, coding domain sequence. (B) Lipid nanoparticle–mRNA formulations used as COVID-19 vaccines. (C) Spike protein expression by flow cytometry using biotinylated hACE2 protein. (D) Spike protein expression analyzed by Western blot using anti-spike as the primary antibody. The asterisk indicates a non-specific band. All data are representative of three independent experiments.

Journal: Frontiers in Immunology

Article Title: Immunogenicity and protective efficacy of the HC009 mRNA vaccine against SARS-CoV-2

doi: 10.3389/fimmu.2024.1416375

Figure Lengend Snippet: Vaccine design and expression of the mRNA vaccine encoding COVID-19 spike. (A) Structure of HC009 RNA. UTR, untranslated region; CDS, coding domain sequence. (B) Lipid nanoparticle–mRNA formulations used as COVID-19 vaccines. (C) Spike protein expression by flow cytometry using biotinylated hACE2 protein. (D) Spike protein expression analyzed by Western blot using anti-spike as the primary antibody. The asterisk indicates a non-specific band. All data are representative of three independent experiments.

Article Snippet: Human embryonic kidney 293 cells (HEK293 cells) (ATCC CRL-3216) and HEK293T/hACE2-TMPRSS2 cells (SinoBiological, OEC003, China) were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, 11965092, USA) supplemented with 10% fetal bovine serum (FBS, Gibco, 11965092, USA) and penicillin (100 U/mL)–streptomycin (100 mg/mL) (Gibco, 15140148, USA).

Techniques: Expressing, Sequencing, Vaccines, Flow Cytometry, Western Blot

Evaluation of the immune protection provided by HC009 during in vivo challenge. (A) Immunization and challenge procedures for the 0.5-, 2-, and 10-μg dose of HC009 in mice. Six- to 8-week-old female hACE2 transgenic mice were immunized with two doses of the vaccines via the intramuscular route at 3-week intervals ( n = 12). Subsequently, they were challenged with live SARS-CoV-2 at 50 days post-vaccination, and the lung and nasal turbinate tissues were collected at the indicated time points after immunization. (B) The body weights of the mice were monitored and recorded for six consecutive days after the challenge ( n = 6). The mice were euthanized after observation. (C) Average clinical scores for disease signs, including lethargy, ruffled fur, hunched back posture, and rapid breathing. A score of 1 was given to each of these clinical signs ( n = 12). (D) Viral RNA in the lungs and nasal turbinate tissues of challenged mice were measured with qRT-PCR at 1, 3, 5, and 6 dpi, respectively ( n = 6). (E) H&E staining was performed to assess pathological changes in the lungs of mice at 1, 3, 5, and 6 dpi ( n = 3). The data are shown as the mean ± SEM. Horizontal dashed line indicates the lower limit of quantification. All the data are representative of three independent experiments. A two-way ANOVA with Tukey’s multiple comparisons test was performed, * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

Journal: Frontiers in Immunology

Article Title: Immunogenicity and protective efficacy of the HC009 mRNA vaccine against SARS-CoV-2

doi: 10.3389/fimmu.2024.1416375

Figure Lengend Snippet: Evaluation of the immune protection provided by HC009 during in vivo challenge. (A) Immunization and challenge procedures for the 0.5-, 2-, and 10-μg dose of HC009 in mice. Six- to 8-week-old female hACE2 transgenic mice were immunized with two doses of the vaccines via the intramuscular route at 3-week intervals ( n = 12). Subsequently, they were challenged with live SARS-CoV-2 at 50 days post-vaccination, and the lung and nasal turbinate tissues were collected at the indicated time points after immunization. (B) The body weights of the mice were monitored and recorded for six consecutive days after the challenge ( n = 6). The mice were euthanized after observation. (C) Average clinical scores for disease signs, including lethargy, ruffled fur, hunched back posture, and rapid breathing. A score of 1 was given to each of these clinical signs ( n = 12). (D) Viral RNA in the lungs and nasal turbinate tissues of challenged mice were measured with qRT-PCR at 1, 3, 5, and 6 dpi, respectively ( n = 6). (E) H&E staining was performed to assess pathological changes in the lungs of mice at 1, 3, 5, and 6 dpi ( n = 3). The data are shown as the mean ± SEM. Horizontal dashed line indicates the lower limit of quantification. All the data are representative of three independent experiments. A two-way ANOVA with Tukey’s multiple comparisons test was performed, * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

Article Snippet: Human embryonic kidney 293 cells (HEK293 cells) (ATCC CRL-3216) and HEK293T/hACE2-TMPRSS2 cells (SinoBiological, OEC003, China) were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, 11965092, USA) supplemented with 10% fetal bovine serum (FBS, Gibco, 11965092, USA) and penicillin (100 U/mL)–streptomycin (100 mg/mL) (Gibco, 15140148, USA).

Techniques: In Vivo, Transgenic Assay, Vaccines, Quantitative RT-PCR, Staining

(A) Geometric sketching yields more even coverage of the transcriptomic space. In our experiments, the Hausdorff distance measures the maximum distance from any point in the dataset to its closest point in the sketch; a lower Hausdorff distance indicates that the points represented by a sketch are in general closer to all of the points in the remainder of the dataset. Geometric sketching results in consistently lower Hausdorff distances than other sampling methods across a large number of sketch sizes and datasets. We use a robust Hausdorff distance that is less sensitive to small numbers of outlier observations (Method Details). Solid lines indicate means and shaded areas indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). (B) Geometric sketches contain more balanced summaries of the transcriptional landscape. t-SNE visualizations of sketches containing 2% of the cells from the adult mouse brain (Saunders et al., 2018) and from the developing and adolescent mouse CNS (Zeisel et al., 2018) using uniform random sampling and geometric sketching, with increased representation of rare cell types in the geometric sketch. Numbers of cells from each cell type are given in Tables S3–S4. Uniform sampling, which does not evenly consider the transcriptional space, produces visualizations that are poor at capturing transcriptional heterogeneity. Geometric sketching substantially underrepresents oligodendrocytes in both datasets compared to uniform sampling, which is expected given the low transcriptional heterogeneity among oligodendrocytes as quantified by differential entropy (Method Details; Tables S3–S4). Visualizations based on other sampling approaches as well as a different visualization method are provided in Figure S1. (C) Geometric sketches preserve rare cell types in the subsampled data. In sketches containing 2% of the total dataset, we counted the number of cells that belong to the rarest cell type in each dataset: 293T cells (0.66% of total cells) in a 293T/Jurkat mixture, dendritic cells (0.38% of total) in a dataset of 68k PBMCs, macrophages (0.25% of total) in a dataset of adult mouse brain cells, and ependymal cells (0.60% of total) in a dataset of developing and adolescent mouse CNS cells. Higher count indicates increased representation of the rare cell type in the sketch. Bar height indicates means and error bars indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). Comparison of rare cell type representation over different sketch sizes is shown in Figure S2B. (D) Geometric sketching is consistently effective at distinguishing biological cell types via clustering. Louvain clustering was applied to a subsample of the dataset, cluster labels were transferred to the full dataset using a k-nearest-neighbor classifier fit to the sketch, and the balanced adjusted mutual information (BAMI) was measured between the unsupervised cluster labels and the labels corresponding to biological clusters provided by each previous study (Method Details). Higher score indicates greater agreement between unsupervised clustering and biological cell type labels. Solid lines indicate means and shaded areas indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). Unsupervised clustering of geometric sketches consistently recapitulates biological cell types better than clustering results obtained by uniform sampling. Other non-uniform sampling methods, k-means++ and SRS, show performance comparable to ours in a few cases, but only geometric sketching obtains competitive performance across all settings. Because samples are drawn without replacement, clustering accuracy may approach that of uniform sampling as the sketch size increases, as is the case in the 293T/Jurkat mixture experiments.

Journal: Cell systems

Article Title: Geometric Sketching Compactly Summarizes the Single-Cell Transcriptomic Landscape

doi: 10.1016/j.cels.2019.05.003

Figure Lengend Snippet: (A) Geometric sketching yields more even coverage of the transcriptomic space. In our experiments, the Hausdorff distance measures the maximum distance from any point in the dataset to its closest point in the sketch; a lower Hausdorff distance indicates that the points represented by a sketch are in general closer to all of the points in the remainder of the dataset. Geometric sketching results in consistently lower Hausdorff distances than other sampling methods across a large number of sketch sizes and datasets. We use a robust Hausdorff distance that is less sensitive to small numbers of outlier observations (Method Details). Solid lines indicate means and shaded areas indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). (B) Geometric sketches contain more balanced summaries of the transcriptional landscape. t-SNE visualizations of sketches containing 2% of the cells from the adult mouse brain (Saunders et al., 2018) and from the developing and adolescent mouse CNS (Zeisel et al., 2018) using uniform random sampling and geometric sketching, with increased representation of rare cell types in the geometric sketch. Numbers of cells from each cell type are given in Tables S3–S4. Uniform sampling, which does not evenly consider the transcriptional space, produces visualizations that are poor at capturing transcriptional heterogeneity. Geometric sketching substantially underrepresents oligodendrocytes in both datasets compared to uniform sampling, which is expected given the low transcriptional heterogeneity among oligodendrocytes as quantified by differential entropy (Method Details; Tables S3–S4). Visualizations based on other sampling approaches as well as a different visualization method are provided in Figure S1. (C) Geometric sketches preserve rare cell types in the subsampled data. In sketches containing 2% of the total dataset, we counted the number of cells that belong to the rarest cell type in each dataset: 293T cells (0.66% of total cells) in a 293T/Jurkat mixture, dendritic cells (0.38% of total) in a dataset of 68k PBMCs, macrophages (0.25% of total) in a dataset of adult mouse brain cells, and ependymal cells (0.60% of total) in a dataset of developing and adolescent mouse CNS cells. Higher count indicates increased representation of the rare cell type in the sketch. Bar height indicates means and error bars indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). Comparison of rare cell type representation over different sketch sizes is shown in Figure S2B. (D) Geometric sketching is consistently effective at distinguishing biological cell types via clustering. Louvain clustering was applied to a subsample of the dataset, cluster labels were transferred to the full dataset using a k-nearest-neighbor classifier fit to the sketch, and the balanced adjusted mutual information (BAMI) was measured between the unsupervised cluster labels and the labels corresponding to biological clusters provided by each previous study (Method Details). Higher score indicates greater agreement between unsupervised clustering and biological cell type labels. Solid lines indicate means and shaded areas indicate standard error across 10 random trials for geometric sketching and uniform sampling and 4 random trials for k-means++ and SRS (due to long runtimes). Unsupervised clustering of geometric sketches consistently recapitulates biological cell types better than clustering results obtained by uniform sampling. Other non-uniform sampling methods, k-means++ and SRS, show performance comparable to ours in a few cases, but only geometric sketching obtains competitive performance across all settings. Because samples are drawn without replacement, clustering accuracy may approach that of uniform sampling as the sketch size increases, as is the case in the 293T/Jurkat mixture experiments.

Article Snippet: We obtained a mixture of 293T cells and Jurkat cells from 10X Genomics ( Zheng et al., 2017 ) containing a much smaller number of 293T cells than Jurkat cells, where cell types are computationally inferred based on consensus clustering and marker genes.

Techniques: Sampling, Comparison

KEY RESOURCES TABLE

Journal: Cell systems

Article Title: Geometric Sketching Compactly Summarizes the Single-Cell Transcriptomic Landscape

doi: 10.1016/j.cels.2019.05.003

Figure Lengend Snippet: KEY RESOURCES TABLE

Article Snippet: We obtained a mixture of 293T cells and Jurkat cells from 10X Genomics ( Zheng et al., 2017 ) containing a much smaller number of 293T cells than Jurkat cells, where cell types are computationally inferred based on consensus clustering and marker genes.

Techniques: Sequencing, Expressing, Software