How to estimate the inter sample and inter replicate correlation coefficient?
0
0
Entering edit mode
2.3 years ago
ATCG ▴ 370

How to estimate the inter sample and inter replicate correlation coefficient? My dataset consists of 2 RNA-Seq samples treatment and control and each sample has 3 replicates. I would like to show that the replicates have a higher correlation coefficient than that of the samples( treatment and control). Thanks!

RNA-Seq Replicates R2 Correlation coefficients • 1.1k views
1
Entering edit mode

Multiple options. cor on the count matrix, or PCA, showing that replicates cluster together.

0
Entering edit mode

Okay thanks. How is this done in R? and how can I actually print the r2 value? I know how to do and plot PCA but not sure how I can get the r2 value..

Also, is there a way to perform this analysis on the fastq files before alignment?

1
Entering edit mode

The R function is called cor, and it will return a matrix that indicates the pairwise correlation between all samples.

Also, is there a way to perform this analysis on the fastq files before alignment?

No. It works on counts. PCA on normalized log2-tranformed ones. cor can be done on raw counts since it measures linear correlations and common normalization techniques simply compute linear scaling factors, so cor does not change.

0
Entering edit mode

The cor function does what I was expecting to do. For example form a counts matrix

a=cor(counts(dds))
cor.samples=corrplot(a, method="number")


Will produce a plot such as this one: https://cran.r-project.org/web/packages/corrplot/vignettes/corrplot-intro.html

But I would like to show a linear correlation plot such as this one

library("ggpubr")
ggscatter(z, x = "WT", y = "TREATMENT",
add = "reg.line", conf.int = TRUE,
cor.coef = TRUE, cor.method = "pearson",
xlab = "WT", ylab = "TREATMENT")


Is this a good way of accomplishing this?

1
Entering edit mode

Use this code on your count matrix after subsetting it to the two samples of interest or the mean between the groups.