This lesson is still being designed and assembled (Pre-Alpha version)

Research Data Management: Instructor Notes

Complete R script

The complete R script is available here and below:

suppressPackageStartupMessages(library(DESeq2))
suppressPackageStartupMessages(library(tidyverse))
suppressPackageStartupMessages(library(EnhancedVolcano))
suppressPackageStartupMessages(library(pheatmap))


###############################
# Episode 05: descriptive plots
###############################

## Data import
counts <- read.delim("counts.txt", header = T, stringsAsFactors = F)
genes <- counts[,1]
counts <- counts[,-1]
row.names(counts) <- genes
xp_design <- read.delim("experimental_design_modified.txt", header = T, stringsAsFactors = F, colClasses = rep("character",4))

# change col names
colnames(xp_design) <- c("sample", "seed", "infected", "dpi")

# reorder counts columns according to the experimental design file
counts <- counts[,xp_design$sample]

## Creation of the DESeqDataSet object
dds <- DESeqDataSetFromMatrix(countData = counts, 
                              colData = xp_design, 
                              design = ~ seed + infected + dpi)


#################################
## Size factors and normalisation
#################################
# normalised counts
dds <- estimateSizeFactors(dds)

size_factors_df <- data.frame(sample = names(sizeFactors(dds)), 
                              size = sizeFactors(dds))

# add the experimental condition of interest for plot labelling
size_factors_df <- left_join(size_factors_df, xp_design, by = "sample")

# sort by seed condition and by infected condition
size_factors_df <- size_factors_df %>% 
  arrange(seed, infected)

size_factors_df$sample = factor(size_factors_df$sample, levels = size_factors_df$sample)

# plot
ggplot(size_factors_df, aes(x = sample, y = size, colour = seed)) + 
  geom_segment(aes(x = sample, xend = sample, y = 0, yend = size), color="grey") +
  geom_point(size = 4) + 
  coord_flip() +
  theme_grey() +
  facet_wrap(~ infected) +
  scale_y_continuous(limits = c(0,2))

# extract normalised counts
counts_normalised = counts(dds, normalized = TRUE)

####################
## sample correlations
####################

# self correlation
pairs(x = counts_normalised[,c(1,1)],pch = 19, log = "xy")
cor(counts_normalised[,c(1,1)])

# correlation between biological replicates
pairs(x = counts_normalised[,1:4],pch = 19, log = "xy")
cor(counts_normalised[,c(1:4)])

# weakly correlated samples 
pairs(x = counts_normalised[,c(1,9,17,25)],pch = 19, log = "xy")
cor(counts_normalised[,c(1,9,17,25)])

##########
# PCA plot
##########

# PCA using the plotPCA function
# variance-stabilizing transformation
vst_dds <- vst(dds)

# customised PCA plot
pcaData <- plotPCA(vst_dds, intgroup = c("seed", "infected", "dpi"), returnData = TRUE)

percentVar <- round(100 * attr(pcaData, "percentVar"))

ggplot(pcaData, aes(PC1, PC2, color = seed, shape = infected, size = dpi)) +
  geom_point() +
  xlab(paste0("PC1: ",percentVar[1],"% variance")) +
  ylab(paste0("PC2: ",percentVar[2],"% variance")) + 
  coord_fixed() +
  ggtitle("PCA plot")


#####################################
# Episode 06: differential expression
#####################################

##################
# diff expression
#################
# Filter design file: keep only mock versus infected
xp_design_mock_vs_infected = xp_design %>% filter(seed == "MgCl2" & dpi == "7") 

# Filter count file accordingly (so the rows correspond to the columns of the filtered xp_design file)
counts_filtered = counts[, colnames(counts) %in% xp_design_mock_vs_infected$sample]

## Creation of the DESeqDataSet object
dds2 <- DESeqDataSetFromMatrix(countData = counts_filtered, 
                              colData = xp_design_mock_vs_infected, 
                              design = ~ infected)



dds2 <- DESeq(dds2)
res <- results(dds2, contrast = c("infected",                        # name of the factor
                                  "Pseudomonas_syringae_DC3000",     # name of the numerator level for fold change
                                  "mock"))                           # name of the denominator level 


# how many genes differentially regulated ?
# threshold of p = 0.01
res %>% 
  as.data.frame() %>% 
  filter(padj < 0.01) %>% 
  dim()

# threshold of p = 0.001
res %>% 
  as.data.frame() %>% 
  filter(padj < 0.001) %>% 
  dim()


# one can also see the impact of false discovery rate method
hist(res$pvalue, col="grey", main = "Non-adjusted p-value distribution")
hist(res$padj, col="lightblue", main = "Adjusted p-value distribution")



#########
# MA plot
#########
# MA plot
plotMA(object = dds2, alpha = 0.01)

# shrink effect size
resLFC <- lfcShrink(dds = dds2, 
                  res = res,
                  type = "normal",
                  coef = 2) # corresponds to "infected_Pseudomonas_syringae_DC3000_vs_mock" comparison

# MA plot shrinked
plotMA(object = resLFC, alpha = 0.01)


############## 
# Volcano plot
#############

EnhancedVolcano(toptable = resLFC,
                x = "log2FoldChange",
                y = "padj",
                lab = rownames(resLFC)
                )

EnhancedVolcano(toptable = resLFC,
                x = "log2FoldChange",
                y = "padj",
                lab = rownames(resLFC),
                xlim = c(-10, +10),
                ylim = c(0,100),
                pCutoff = 1e-06,
                transcriptPointSize = 2.0,
                FCcutoff = 2, 
                title = "Pseudomonas syringae DC3000 versus mock \n (fold change cutoff = 2, p-value cutoff = 1e-06)",
                legend=c(
                  'Not significant',
                  'Log2 fold-change (but do not pass p-value cutoff)',
                  'Pass p-value cutoff',
                  'Pass both p-value & Log2 fold change')
                )



######### 
# Heatmap
#########
#counts_normalised_scaled = t(scale(t(counts_normalised), center = T, scale = F))

# not scaled
pheatmap(counts_normalised[1:10,], cluster_rows = F, cluster_cols = F)
pheatmap(counts_normalised[1:20,], cluster_rows = F, cluster_cols = F)
pheatmap(counts_normalised[1:50,], cluster_rows = F, cluster_cols = F)

# scaled using a log2 transformation
counts_normalised[counts_normalised == 0] <- 1

pheatmap(log2(counts_normalised[1:50,]), cluster_rows = F, cluster_cols = F)

# clearer heatmaps by filtering out genes not differentially expressed
genes_differential = 
  res %>%
  as.data.frame() %>%
  mutate(gene = row.names(res)) %>% 
  filter(padj < 0.001) %>% 
  select(gene) 
  
# use the genes names in res_only_diff to filter the counts_normalised matrix
dim(counts_normalised) # contains all gene info = 33,768 genes

# filter
counts_normalised_only_diff = counts_normalised[row.names(counts_normalised) %in% genes_differential$gene, ]
dim(counts_normalised_only_diff)

# no clustering
pheatmap(log2(counts_normalised_only_diff + 1), 
         cluster_rows = F, 
         cluster_cols = F, 
         show_rownames = F, 
         show_colnames = F,
         main = "No clustering")

# cluster genes
pheatmap(log2(counts_normalised_only_diff + 1), 
         cluster_rows = T, 
         cluster_cols = F,
         show_rownames = F, 
         show_colnames = F, 
         main = "Clustering of genes")

# cluster genes and samples
pheatmap(log2(counts_normalised_only_diff + 1), 
         cluster_rows = T, 
         cluster_cols = T,
         show_rownames = F, 
         show_colnames = F, 
         main = "Clustering of genes and samples")

# even clearer heatmaps by filtering out genes not differentially expressed
# and genes with a high fold change
genes_differential_fold = 
  res %>%
  as.data.frame() %>%
  mutate(gene = row.names(res)) %>% 
  filter(padj < 0.001) %>% 
  filter(log2FoldChange > 4) %>% 
  select(gene) 

counts_normalised_only_diff_fold = counts_normalised[row.names(counts_normalised) %in% genes_differential_fold$gene, ]

# cluster genes and samples
pheatmap(log2(counts_normalised_only_diff_fold + 1), 
         cluster_rows = T, 
         cluster_cols = T,
         show_rownames = F, 
         show_colnames = F, 
         main = "Clustering of genes and samples")