7.6 years ago by
Washington University School of Medicine, St. Louis, USA
This thread describes the: Expected correlation between Exon Array and RNA-Seq.
The paper Alternative expression analysis by RNA sequencing contains a variety of comparisons between Affymetrix Exon arrays, custom NimbleGen arrays, and RNA-seq. The paper RNA-seq: An assessment of technical reproducibility and comparison with gene expression arrays also seems relevant.
As someone who has worked with both microarrays (including custom designed arrays), and RNA-seq I think you hit the nail on the head with your comment about the hypothesis neutral approach of RNA-seq. IMHO, being locked to extant information at design time is a significant limitation of microarrays (one that is impractical to overcome).
Other pros of RNA-seq over microarrays include: theoretically unlimited dynamic range and better signal-to-noise ratio. Probably the clincher though is the diversity of information your can simultaneously obtain (with appropriate analyses of course): gene expression, alternative isoform detection and quantification, mutation detection, allele specific expression, gene fusion discovery and quantification, RNA editing, etc.
One pro of microarrays is that they have arguably more robust strand specific assays currently. Another is that they are less influenced by a wide expression distribution/range (the difference in copy number between the lowest and highest expressed transcripts in the cell). This is both the blessing and the curse of the finite dynamic range in microarrays. In RNA-seq, the random sampling nature of the assay means that you can burn a large percentage of your data sequencing the top N% of expressed genes. I have seen libraries where >75% of sequenced reads corresponded to the top 5 genes. Microarrays do not suffer from this phenomenon, although you can improve signal by enriching for polyA+ RNA or otherwise removing rRNA species.