Best Practice Approach In R To Processing Illumina Microarrays For Rna Profiling

Question

[Brian Tsai]

Hi,

I'm tasked with analyzing rna expression data collected using various illumina microarrays. I was wondering what the current best approaches (preferably using R) to pre-processing the raw data and collapsing to gene expression levels (as opposed to probe-based or transcript-based).

I found a paper online that discusses this topic, but it's from 2008/early 2009, and was wondering if best practices have been updated since. Thanks!

http://bioinformatics.oxfordjournals.org/content/25/6/751.full

Answer 1

[Ying W]

To add a little bit of detail to UnivStudent's answer:

Personally I find the lumi package to have less flexibility than beadarray. If you are new to microarray analysis in R, the limma user guide is a must read. How you should think of microarray analysis is that there are certain platform specific methods used (covered by beadarray/lumi/limma) and the underlying analysis is typically the same.

One of the platform specific methods is background correction. I've used neqc() in limma to great success. ref

Now your question of probe-based vs transcript-based gets to one of the differences between Affymetrix and Illumina arrays. Typically with Illumina arrays you stick with probes since each probe 'usually' corresponds to a gene (1 probe per gene) rather than summarizing across probes (several probes per gene). However, some tools (such as IPA and partek) will gene-summarize (combine multiple probes -> gene using mean/median/etc) but this will not change the vast majority of gene expression values.

Lastly, be aware of terminology, with Illumina microarrays you will sometimes find the term bead-summarized. This has to do with the technology (multiple identical probes are put onto different beads) and this bead-summarized value will usually be value you get from the facility running the microarray and the one you use for differential expression analysis. More advanced analysis using raw bead values can be done using beadarray.

Answer 2

[UnivStudent]

I would recommend taking a look at the Bioconductor project. There are two reccomended packages for Illumina microarrays lumi and beadarray that are part of their standard workflow. Check the documentation on those pages for detailed workflows for typical analyses. I'd also reccomend looking for well-cited papers using the platform that you plan to analyze and seeing how they did their analysis.