do anybody know is there a package in R which can replace the empty value in microarray data based on the normalized distribution

Question

[sw.arker]

hello, everybody,

I am wondering if there is a package or function in R which can replace the empty value in microarray data based on the normalized abundance distribution. because otherwise I have to remove them along with other non-empty replicates since the log transformation will return error. I know some software can do that. I am wondering which function/package can do the same job in R. there must be something available there in R. please let me know, thanks!

[Devon Ryan]

What format is your data in? If you've already read in the values into a matrix or eSet, then just substitute NA for the 0s.

Answer 1

[pld]

The proper thing to do is to remove them. How do you know that the true value isn't empty, and the problem is coming from replicates showing signal? You can't know that so it would be incorrect to replace null values with values based on the other probes.

You might want to double check that there aren't problems that occurred when performing the array experiment. This might mean that some of the arrays were performed incorrectly while others were fine, there's no way to tell. You should see if there are any correlations between the samples with null and those without for other probes/genes. You can also biologically validate the expression for that gene, if it comes up the same across all samples then you know there's a problem with you arrays.

On the other hand, this just might be the result of sample to sample variation, in which case you should definitely remove that probe for all samples. If the non-null values are close to zero, you might want to add an offset to your background subtraction, that might help, or explore some of the other methods for background correction. The best practice is to throw them out.