I have one illumina beadchip expression data and other being the Agilent microarray data from GEO database. I want to do meta-analysis of these two different datasets based on some kind of statistical algorithm so that i can get differentially expressed genes in combined datasets.
So what should be the best approach to be used..
In addition to the previous poster, if you want to control for the biases attributed to two different microarrays, you could use likelihood-ratio test to test for differential expression. The alternative model (group + microarray_type) against a null model (microarray_type). This will give you differentially expressed genes while controlling for microarray_type.
You can also normalize the data separately. Then take common probes (i.e. those that target same genes) and perform batch-effect correction using combat, treating different platforms as batches. You can then carry on with regular differential expression analysis, eg. limma.
We recently published ImaGEO, a webtool to perform meta-analysis with GEO datasets. You only have to introduce the GSE codes and assign samples to groups. All the necessary steps are automatically performed.
http://bioinfo.genyo.es/imageo/
Paper: https://academic.oup.com/bioinformatics/advance-article-abstract/doi/10.1093/bioinformatics/bty721/5076580
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis