Question

reading a *.txt file of expression values into an AffyBatch

0

Entering edit mode

m • 0

@m-8386

Last seen 8.7 years ago

United States

Dear All,

I have a *.txt file which contains microarray expression data only. Each row corresponds to a probe, and each column to a sample. I know the data has not been normalized and would like to normalize it using the rma() function in affy. I've tried simply importing the data using read.table(), but when I then try to apply the rma() function to the data, this gives the following error:

Error in (function (classes, fdef, mtable) : unable to find an inherited method for function ‘probeNames’ for signature ‘"matrix"’

I believe the way to avoid this error would be to somehow read the data into an AffyBatch, since the rma() function needs an AffyBatch as the input. How would you go about reading a *.txt file in to an AffyBatch? Alternatively, is there some way to apply the rma() function to a matrix/data.frame?

Thanks in advance!

M

affy normalization rma • 3.0k views

ADD COMMENT • link updated 9.4 years ago by James W. MacDonald 67k • written 9.4 years ago by m • 0

score 0 · Answer 1 · 2015-07-13

0

Entering edit mode

arfranco ▴ 130

@arfranco-8341

Last seen 6 months ago

European Union

Agilent data are *.txt files that contains Probenames as a column name into the genes section..

kind of suspicious ...

ADD COMMENT • link 9.4 years ago arfranco ▴ 130

0

Entering edit mode

Hi arfranco,

I believe this was originally Illumina BeadChip data, but the file that I have is not the raw output and contains only expression values (though the data definitely has not been normalized).

What would Agilent output look like (in terms of how the data would be organized)? Sorry for the basic question--I am new to the field.

Thanks!

ADD REPLY • link 9.4 years ago m • 0

0

Entering edit mode

<font><font>Es important to know where the data come from. If it is a txt file, you can look inside. My Agilent data is revised by importing the txt file into Excel.</font></font>

ADD REPLY • link 9.4 years ago arfranco ▴ 130

score 0 · Answer 2 · 2015-07-13

0

Entering edit mode

James W. MacDonald 67k

@james-w-macdonald-5106

Last seen 4 days ago

United States

You do NOT want to normalize Illumina data using RMA. Well, I should say you CANNOT normalize using RMA. It just won't work.

Where are you getting these data? How do you know they aren't normalized? There are any number of normalization methods in the limma package - have you read the user's guide?

ADD COMMENT • link 9.4 years ago James W. MacDonald 67k

0

Entering edit mode

Basically, the situation is that I am working as a research assistant and was given this data by the head researcher, who said that the data has not been normalized and that it would be best to use RMA to do so.

Why does RMA not work on Illumina data?--is it that there is no way to get the data into the correct format?

Thanks!

ADD REPLY • link 9.4 years ago m • 0

1

Entering edit mode

RMA won't work on Illumina data for the same reason that a hammer is not really useful for cutting wood. It's just not designed for that task. For some boring detail, Affy arrays have some number of short (25-mer) probes that are all intended to measure one transcript. RMA is intended to summarize those probes into a single value (based on first mapping probes to a 'probeset'), given the fact that the hybridization of a 25-mer will be highly dependent on the GC content, which is a technical bias that one would like to eliminate.

Illumina probes are much longer (60-mers, IIRC), and there is only one probe per transcript. So the whole rationale behind RMA just doesn't apply. There is a quantile normalization step that is part of the RMA summarization, so your head researcher might have meant for you to quantile normalize the data, rather than explicitly meaning for you to run RMA.

So, as I mentioned before, the limma package is more likely the tool of choice for your project. There is a worked example in the limma User's Guide, starting on page 105 that would be particularly relevant. That example uses negqc() which will background correct, quantile normalize, and log transform, which is what RMA does, without the summarization of probes into a probeset.