I have Illumina data and am using genome studio for the analysis. When normalizaing the micro-array data through Genome studio, how do we decide which normalization to use?
It offers four normalization method, rank invariant, cubic-spline, quantile and average. Also I want to know, does correlation value have any role in deciding the normalization method?
Using quantile-normalization for CNV analysis of complex disease with few and short CNV-aberrations is fine, but for cancer genomes it can be tricky. Cancer genomes often show severe CNV-aberrations. This can violate quantile-normalization's main assumption: all arrays intensities should have roughly the same distribution. Breaking this assumption and squeezing all arrays into same intensity distribution could insert spurious CNVs.
Therefore, I would suggest to use "sparse overcomplete respresentation (SOR)" or "allelic crosstalk calibration (ACC)" to normalize cancer CNV data. (For SOR check our publication at Nucleic Acid Research - http://nar.oxfordjournals.org/content/early/2011/04/12/nar.gkr197.abstract )
Let me know, if you need code for Illumina arrrays.
Cheers,
Okko
I am interested in doing gene expression analysis for cancer data not CNV analysis. Then which normalization method I should use.
Thanks Djork-Arne Clevert and Rohan for the valuable suggestions.
Neha:)
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology