Unable to run ComBat script from R's sva library

Question

I am trying to run ComBat script on a dataset with 2 batches, but I am getting errors and I do not know how to inspect code since I am an R newbie.

I am running ComBat method in this way:

# Load sva
library(sva)
# Read expression values
dat = read.table('dataset.xls', header=TRUE, sep='\t')
# Read sample information file about batches
sif = read.delim('sif.tsv', header=TRUE, sep='\t')
# Call ComBat
ComBat(dat=dat,batch=sif$Batch, mod=NULL)

Anyway my output is:

Found 2 batches
Found 0  categorical covariate(s)
Found 54675 Missing Data Values
Standardizing Data across genes
Error in solve(t(des) %*% des) %*% t(des) %*% y1 : 
  requires numeric/complex matrix/vector arguments

Data format for dat is:

probe set    <sample1>   ...    <sampleN>
<gene_name>  <value1>    ...    <valueN>
...

Data format for sif is:

Array name   Sample name   Batch
<Array1>     <Sample1>     <Batch1>
...

Any hint is appreciated. I'll provide more info if needed.

Thanks

Answer 1

I have figured out how to do it, adding:

# Remove NA from end of lines
l_dat = length(dat)
dat[l_dat] <- NULL
# Remove probe set from beginning of lines
dat[1] <- NULL

just before the ComBat call. This because last column contains NA values (next warning goes away:

Found 54675 Missing Data Values

), and first column contains probe set (non numeric values) that raise next error:

Error in solve(t(des) %*% des) %*% t(des) %*% y1 : 
  requires numeric/complex matrix/vector arguments

Answer 2

It looks like when you're reading in the information it's not storing it in a format you expect.

From the error, ComBat is expecting a matrix with numeric or complex values. read.table from memory will give you a data.frame.

So try running:

dat <- as.matrix(dat)
ComBat(dat=dat,batch=sif$Batch, mod=NULL)