1.3.Differential Expression

Differential Expression for bulk RNAs

Pipeline

Data Structure

Inputs

File format

Information contained in file

File description

Notes

txt

miRNA.homer.ct.mx

Raw counts matrix

Used for edgeR and DESeq2 analysis

txt

miRNA.homer.rpm.mx

RPM matrix

Used for wilcoxon test

txt

miRNA.NCvsHCC.DESeq2.ids.rpkm.mx

RPKM matrix

Used for clustering plot

txt

design.txt

Sample info.

Classification information

Outputs

File format

Information contained in file

File description

Notes

tsv

miRNA.NCvsHCC.edgeR.tsv

EdgeR result

LogFC, PValue and FDR

txt

miRNA.homer.DESeq2.rlog.mx

DESeq2 result

Rlog normalized counts

tsv

miRNA.NCvsHCC.DESeq2.tsv

DESeq2 result

LogFC, PValue and padj

tsv

miRNA.NCvsHCC.wilcox.tsv

Wilcox result

LogF, PVaule and FDR

pdf

miRNA.NCvsHCC.DESeq2.norm.pdf

Heatmap result

Rlog normalized count

pdf

miRNA.NCvsHCC.DESeq2.pca.pdf

Scatter plot

PCA analysis

pdf

miRNA.NCvsHCC.DESeq2.MAplot.pdf

Scatter plot

MAplot analysis

pdf

miRNA.NCvsHCC.DESeq2.Dist.pdf

Heatmap result

Sample distance

pdf

miRNA.NCvsHCC.DESeq2.ids.DE.pdf

Heatmap result

DE genes

Running Scripts

Software/Tools

Assumption for most normalization and differential expression analysis tools: The expression levels of most genes are similar, i.e., not differentially expressed.

a) DEseq: defines scaling factor (also known as size factor) estimates based on a pseudoreferencesample, which is built with the geometric mean of gene counts across all cells (samples).

b) EdgeR (TMM): trimmed mean of M values

c) Wilcox Test using RPM: Read counts Per Million of total mapped reads; alternatives: RPKM, TPM

Performance:

Normalizing single-cell RNA sequencing data: challenges and opportunities, Nature Methods, 2017

Example of single case

# experimential design
design <- read.table("design.txt",sep="\t",header=T)

# expression matrix
mx <- read.table("miRNA.homer.ct.mx",sep="\t",header=T)

# filter genes
filter_genes <- apply(
    mx[,2:ncol(mx)],
    1,
    function(x) length(x[x > 2]) >= 2
)
mx_filterGenes <- mx[filter_genes,]

Draw Plots

1. Heatmap for DESeq2 normalized count matrix

2. PCA analysis

3. MA plot

4. Distance between samples

5. Hierarchical clustering for differential expressed genes

Tips/Utilities

edgeR User Guide

DESeq2 User Guide

Homework and more

  1. Identify differential expressed genes for other RNA types. between differential conditions, i.e. Normal Control (NC) V.S. HCC using three methods: edgeR, DESeq2 and Wilcox/Mann-Whitney-U Test.

  2. Draw Venn plot to show the difference among the above three methods.

More Reading and Practice

Video

1. Differential expression

@Youtube

@Bilibili

Last updated