Statistical and Computational Methods in Single-cell Analysis - - PowerPoint PPT Presentation

Statistical and Computational Methods in Single-cell Analysis

Zhicheng Ji

Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 1 / 18

Overview

1

Introductions to Single-cell Sequencing

2

Methods for scRNA-seq

3

Methods for scATAC-seq

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 2 / 18

Introductions

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 3 / 18

Introductions

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 4 / 18

Technologies

single-cell DNA/CNV sequencing: DNA sequence information in single cells. single-cell RNA-seq (scRNA-seq): Gene expressions in single cells. single-cell ATAC-seq (scATAC-seq): Open chromatin activities in single cells. single-cell Bisulfite sequencing: DNA methylation in single cells. single-cell VDJ-seq: TCR/Ig sequence information in single T/B cells. CITE-seq: cell surface protein expressions in single cells. Others: scDNase-seq, scChIP-seq, scHi-C

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 5 / 18

Sparsity

Jason D. Buenrostro, et al. Single-cell chromatin accessibility reveals principles of regulatory

• variation. (2015) Nature.

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 6 / 18

Analysis Tasks

scRNA-seq Imputation Cell Clustering Dimension Reduction Differential Gene Sample alignment/batch effect removal Pseudotime Analysis scATAC-seq Signal Aggregation/Enhancing Cell Clustering Inferring Regulatory Programs Prediction/Aligning with scRNA-seq

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 7 / 18

TSCAN: Pseudotime Reconstruction

Cole Trapnell, et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. (2014) Nat Biotechnol.

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 8 / 18

TSCAN: Pseudotime Reconstruction

• 1

2 3 4 5

−2.5 0.0 2.5 5.0 5 10 15

PC1 PC2

Group

• 1

2 3 4 5

Zhicheng Ji, et al. TSCAN: Pseudo-time reconstruction and evaluation in single-cell RNA-seq

• analysis. (2016) Nucleic Acids Res.

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 9 / 18

TSCAN: Pseudotime Reconstruction

−30 −20 −10 10 20 −30 −20 −10 10 20 −20 −10 10 20 30 PCA_dimension_1 PCA_dimension_3 PCA_dimension_2

• ●
• ●
• ●
• ●
• 1

2 3 4 5

Zhicheng Ji, et al. TSCAN: Pseudo-time reconstruction and evaluation in single-cell RNA-seq

• analysis. (2016) Nucleic Acids Res.

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 10 / 18

TSCAN: Pseudotime Reconstruction

• Monocle

TSCAN nocluTSCAN Marker

3 6 9 3 6 9 3 6 9 3 6 9 10 20 30 40 50 50 100 150 50 100 150 200 50 100 150

Pseudotime Expression

Gene

• MEF2C

MYH2 ENO3

Zhicheng Ji, et al. TSCAN: Pseudo-time reconstruction and evaluation in single-cell RNA-seq

• analysis. (2016) Nucleic Acids Res.

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 11 / 18

STIP: State Transition Inference Prediction

Zeyu Chen*, Zhicheng Ji*, et al. A TCF-1-centered transcriptional network drives an early binary effector versus exhausted CD8+ T cell fate decision . In revision Immunity.

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 12 / 18

Raisin: Differential Gene

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 13 / 18

Raisin: Differential Gene

5 cell 50 cell 0.00 0.25 0.50 0.75 0.00 0.25 0.50 0.75 0.00 0.25 0.50 0.75 1.00

Real_FDR Reported_FDR Method

BPSC_count DESeq2_count DEsingle_count edgeR_count limma_count limma_saver limmacell_count limmacell_saver MAST_count MAST_saver raisin_saver scDD_count scDD_saver scde_count t_count t_saver wilcoxon_count wilcoxon_saver

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 14 / 18

SCRAT: scATAC-seq Signal Aggregation

Zhicheng Ji*, Weiqiang Zhou*, et al. Single-cell regulome data analysis by SCRAT. (2017) Bioinformatics.

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 15 / 18

SCATE: scATAC-seq Signal Enhancement

Baseline Activity True Activity Transformed Activity Observed scATAC-seq counts Cell Type Specific Variation DHS Cluster 1 DHS Cluster 2 DHS Cluster 3 DHS Cluster 1 DHS Cluster 2 DHS Cluster 3

Poisson Distribution Transform Function

Transformation Function

Bulk ATAC−seq scATAC−seq

scATAC-seq Bulk DNase-seq

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 16 / 18

SCATE: scATAC-seq Signal Enhancement

yi,j ∼ Poisson(Ljµsc

i,j)

log µsc

i,j = h(log(µi,j))

log(µi,j) = ai + bi ∗ δc(i),j Objective: Learn δc(i),j with MLE

yi,j: observed scATAC-seq read count for DHS i and cell j µsc

i,j: Transformed activity

µi,j: True activity h: transformation function ai, bi: Baseline mean and standard deviation δc(i),j: Cell type specific bias

Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 17 / 18

SCATE: scATAC-seq Signal Enhancement

3 6 9 0.0 2.5 5.0 7.5 10.0

Bulk DNase−seq scATAC−seq Correlation: 0.49 Raw scATAC−seq

2 4 6 0.0 2.5 5.0 7.5 10.0

Bulk DNase−seq scATAC−seq Correlation: 0.667 500 CRE Cluster

2 4 6 8 0.0 2.5 5.0 7.5 10.0

Bulk DNase−seq scATAC−seq Correlation: 0.678 1000 CRE Cluster

2 4 6 8 0.0 2.5 5.0 7.5 10.0

Bulk DNase−seq scATAC−seq Correlation: 0.686 2000 CRE Cluster

2.5 5.0 7.5 10.0 0.0 2.5 5.0 7.5 10.0

Bulk DNase−seq scATAC−seq Correlation: 0.725 Average DNase−seq

2 4 6 0.0 2.5 5.0 7.5 10.0

Bulk DNase−seq scATAC−seq Correlation: 0.751 SCATE

3 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 A A A A A A A A A R R R R R R R R R S S S S S S S S S 3 3 3 3 3 3 3 2 2 2 2 2 2 2 1 1 1 1 1 1 1 A A A A A A A R R R R R R R S S S S S S S 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 A A A A A A A A R R R R R R R R S S S S S S S S 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 A A A A A A A A R R R R R R R R S S S S S S S S 3 3 3 3 3 3 3 2 2 2 2 2 2 2 1 1 1 1 1 1 1 A A A A A A A R R R R R R R S S S S S S S 3 3 3 3 3 3 3 2 2 2 2 2 2 2 1 1 1 1 1 1 1 A A A A A A A R R R R R R R S S S S S S S

Monocyte K562 Brain CMP GM12878 Thymus

20 40 60 80 50 100 150 25 50 75 100 200 400 600 50 100 150 200 25 50 75 100 0.25 0.50 0.75 0.2 0.4 0.6 0.8 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.3 0.4 0.5 0.6 0.7

Number of Cells Pooled Correlation Method

S R A 1 2 3 SCATE Raw scATAC−seq Average DNase−seq 500 CRE Cluster 1000 CRE Cluster 2000 CRE Cluster

A B Zhicheng Ji (Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health) Single-cell Analysis 18 / 18