Frma Normalization Via Insilicodb

Question

[Jonathan Crowther]

I have downloaded data from InsilicoDB which is an ExpressionSet. When performing the e<- exprs(eSet) I get values which have been fRMA normalized however I expected that these values would have been Log2Ratio Transformed however they seem to be in a scale of 2.00-16.00

If I were to run these through a log2 Transformation then it would suggest that none of the values would be negative and hence show no down regulation across the 54,000 probes which is not really realistic.

My question is looking for clarification about the format of the data after fRMA I have read the fRMA paper and this seems to hint toward data that is already log2 transformed.

I am quite confused by this so any help or ideas would be appreciated.

cheers, Jonathan

[alaincoletta]

frma data is log2 transformed. Are you using the InSilico DB R package (http://www.bioconductor.org/packages/2.12/bioc/html/inSilicoDb.html) or the web interface (https://insilicodb.org)?

Answer 1

[Sean Davis]

The data are log2 transformed normalized intensities. Affymetrix arrays are single-color arrays that measure fluorescence intensity of the probes. So, there is not a concept of up- or down-regulation measured by a single array. To measure relative expression, one can look at the relationship between two arrays for a simple 2-sample experiment by subtracting the log2 intensities (the equivalent of taking a ratio). More commonly, one would use a statistical test to look for differences between two or more groups; often this is done using Bioconductor tools like limma.

Since you seem to be using R, I would encourage you to read the limma user guide. It is stocked full of useful general information as well as practical examples of working with affymetrix data.

[Jonathan Crowther]

Hi Sean, Thank you for your response, I have gotten the feeling from reading that this was the case. I have used limma previously to compare samples as you have mentioned however I was planning on creating a scoring type matrix that would create a score for multiple factors and I was hoping to convert the log2 gene expression value to a absolute value, as an example; up-regulation (1), no change (0) and down Regulation (-1) This could then be taken into account but I assume that this would need a scale in order to determine the genes/probes that fall into each of these categories?

Thanks again for your help, Jonathan