per-gene differential expression testing; and the "results" func-
tion returns the results of the differential expression testing as a
data frame. The R command to compute log 2 expression ratios
andp-values for the second species (in this vignette, human)
would be identical to the above command, but with "human"
replacing "dog."- Select ortholog pairs of genes for which the genes are differen-
tially expressed (padj <0.05 for each of the genes in the
ortholog pair), and organize log 2 ratios for each of the ortho-
log pairs (in the two respective species) in a two-column data
frame.
deg_dog_human_comb_df <- merge(as.data.frame(deg_dog_df),
merge(as.data.frame(deg_human_df),
human_dog_ensg, by.x=0, by.y=1), by.x=0, by.y=8)
deg_dog_human_comb_df_filt <- deg_dog_human_comb_df[which(deg_dog_human_comb_df$padj.x <
0.05 &
deg_dog_human_comb_df$padj.y < 0.05),]
>dim(deg_dog_human_comb_df_filt)
[1] 1788 14
Thus, 1788 ortholog pairs of genes are differentially
expressed in both dog and human bladder cancer (vs. normal
bladder) with a false discovery rate of 0.05.- Scatter plot of log 2 ratios for ortholog gene pairs, across
mRNA-seq studies in two species.
scatter_plot <- ggplot(deg_dog_human_comb_df_filt) +
geom_point(aes(x=log2FoldChange.x, y=log2FoldChange.y)) +
theme(text=element_text(size=20)) +
xlab(expression(log[2](cancer/normal)~"in dog")) +
ylab(expression(log[2](cancer/normal)~"in human"))
>print(scatter_plot)
This analysis shows that there is a significant correlation
between upregulation or downregulation of a gene in human
bladder cancer, and upregulation or downregulation of its dog
ortholog in canine bladder cancer (Fig.3).4 Notes
- Not all non-human genes will have a known human ortholog
with a HGNC symbol. - A single-human gene can have multiple dog orthologs (or vice-
versa), and thus, a given Ensembl human or dog gene identifier
can appear multiple times in the above data frame (thus
explaining the large number of rows in the data frame). - Even in the unsupervised analysis shown in Fig.2, a general
separation of normal from cancer samples is evident across the
Cross-Species RNA-Seq Analysis 303