Monthly Archives: February 2008

Follow up on meta-analysis

February 28, 2008 – 21:42

Fourteen days since my last post. Quite a while, indeed. Mostly I’ve been stumbled with work and some health related issues. Anyway, I thought I’d follow up on the meta analysis matter I discussed in my last post.

It turns out that it’s a fault of both limma and the data sets, because apparently the raw data found in the Stanford Microarray Database have different length, gene-wise (a result of not all spots on the array being good?) and limma itself does need equal length tables to form a single object (I stumbled upon the same problem when doing my thesis, but I used a hack to work around it), and does not perform any checking.

According to the documentation, the “merge” command should be used to deal with these cases, but here’s what I get:

[sourcecode lang='c']

>> RG1 = read.maimages(file=”file1.txt”,source=”smd”)
Read file1.txt
>> RG2 = read.maimages(file=”file2.txt”,source=”smd”)
Read file2.txt
>> merge(RG1,RG2)
Error in merge(RG1,RG2): Need row names to align on
>> rownames(RG1)
NULL
>> rownames(RG2)
NULL [/sourcecode]

I’m going to ask the Bioconductor ML and see what they tell me.

By Einar | Posted in Science | Tagged , , | Comments (0)

Meta analysis difficulty increasing

February 14, 2008 – 22:17

Again in the past days I’ve been banging my head thanks to the fact that doing meta-analysis with microarray data is more difficult than what it seems.

The problem sometimes lies in the data, sometimes lies in the analysis software and sometimes in a combination of factors. When doing work on a public data set (Zhao et al., 2005), I had to start analysis from raw data. Now, I tried using both the limma and marray Bioconductor packages, but both of them bail out with cryptic error messages. From what I’ve learnt by googling around, it seems that R doesn’t like batch loading of tables of different length.

I have 177 samples and I have to normalize them all together. Apparently this is a quirk of marray and limma (or worse, R itself) which is preventing me to work properly. And this is not the first time it happens, either: in the past year I’ve lost a lot of time dealing with software issues rather than performing real analsis. The problem has been posted already on some R mailing lists (and on BioC, too), but judging from the responses I doubt I’ll see a solution.

I guess I’ll have to work around this somehow (and of course, this doesn’t improve the idea I have of R…).

By Einar | Posted in Science | Tagged , , | Comments (0)