R: Pearson correlation rcorr(x,y) [x=matrix, y=vector] ignores y

Question

I have a matrix x (30x2000) of 2000 gene expressions in different cell lines and a vector y (30x1) of a continuous variable outcome. I want to calculate Pearson correlation between each gene and the outcome, so, I expect a 2000x1 vector of r-values. I've used rcorr(x,y) but the result is a 2000x2000 matrix, so I guess it's ignoring the y and calculating all genes against all (the manual says:

x = a numeric matrix with at least 5 rows and at least 2 columns (if y is absent)

But can I have more than one column and have y too? Do I have to use a different function?

Answer 1

Using the function cor will work. In general, if x is MxN andy y is MxP, then cor(x,y) will be an NxP matrix where the entry (i,j) is the correlation between x[,i] and y[,j].

Building on SimonO101's reproducible example:

> set.seed(1)
> x <- matrix( runif(12) , nrow = 3 )
> y <- runif(3)
> cor(x,y)
           [,1]
[1,]  0.3712437
[2,]  0.9764443
[3,]  0.2249998
[4,] -0.4903723

If you want just a vector and not a matrix:

> array(cor(x,y))
[1]  0.3712437  0.9764443  0.2249998 -0.4903723

Answer 2

You need to apply the cor function across the columns of your x matrix...

apply( x , 2 , cor , y = y )

A reproducible example

#  For reproducible data
set.seed(1)

#  3 x 4 matrix
x <- matrix( runif(12) , nrow = 3 )
#          [,1]      [,2]      [,3]       [,4]
#[1,] 0.2655087 0.9082078 0.9446753 0.06178627
#[2,] 0.3721239 0.2016819 0.6607978 0.20597457
#[3,] 0.5728534 0.8983897 0.6291140 0.17655675

# Length 3 vector
y <- runif(3)
#[1] 0.6870228 0.3841037 0.7698414

# Length 4 otuput vector
apply( x , 2 , cor , y = y )
#[1]  0.3712437  0.9764443  0.2249998 -0.4903723