In using the justRMA affy function to generate expression levels from hgu133a chips I noticed that the probeset annotations all had an "X" in front of them, so for example "1007_s_at" became "X1007_s_at". I've looked in quite a few different places in various type of documentation, but haven't been able to shed any more light on what this is about.

If someone could just point me to more detailed info I would greatly appreciate it! BTW, I read somewhere that many researchers prefer RMA over all the other expression level algorithms. Certainly the ability to also process large numbers of samples in a single batch is another significant advantage.

Alan J. Robinson

robin073@umn.edu 

The "X" in front of the affy ID is likely due to R creating a "valid" R name at some point during processing.  R names may not begin with numbers, so when R needs to convert something to a valid name (column names, for example), it adds an "X" in front.  I am not sure what processing was done that resulted in that occurring, though.

Following up on a suggestion from Ben Bolstad, it appears that the problem is not in the RMA routines, but in subsequent R processing of the results. Apparently many probe names violate the R restriction that a name can't start with a number, so some R routines such as data.frame( ) add an "X" prefix, which may then be needed to be stripped off.

I'm working on novel statistical methods for analyzing the normalized expression values, thus the need to further process RMA output. Despite the enormous literature surrounding gene expression, this field is by no means exhausted because of the numerous deficiencies in the traditional statistics literature. I've seen problems in the literature produced by some of the world's leading universities such as Oxford, Stanford, MIT, and Chicago, and by leading statisticians such as heads of departments, full professors, and even by such luminaries such as RA Fisher.

Alan J. Robinson

robin073@umn.edu