Next Gen Sequencing Vs Microarray Advanced Applications

Question

[Travis]

Hi all,

I was wondering if anyone could recommend a good review highlighting the state of the NGS field for applications like classifier generation (prognostics and diagnostics), SNP association studies etc. My main area of interest is cancer.

Obviously NGS has been shown to have many strengths over microarrays and has proven itself in areas like transcript discovery but in areas like classifier generation it may not yet be so strong due to inter-individual variation/noise & the number of samples required to correct this etc.

I am new to the field and would just like a good overview of where the strengths and weaknesses are and what remains to be done before these more advanced applications are possible.

ANY input would be greatly appreciated.

Thanks a lot in advance,

tb34

Answer 1

[Jeremy Leipzig]

My understanding is that RNA-Seq NGS has better dynamic range and less noise than expression microarrays. The ERANGE paper from Mortazavi looks at this in detail.

Because microarrays have no units, normalization is a mess. There are still papers coming out dealing with this issue.

You need replicates regardless of which route you choose.

Microarrays are easier to process bioinformatic sense because the probes are essentially digital, there is no alignment step. They are cheaper, but even that advantage is disappearing.

In terms of genotyping, microarrays are convenient means of finding what you expect to see. They are certainly no better at detecting rare variants or SNPs.

[Travis]

Are there any publications where people have attempted e.g. multivariate prognostic classifier generation with NGS?

Obvious examples (Mammaprint) exist for microarrays but I don't know if anything similar has been attempted with NGS...

[Jeremy Leipzig]

That's just a genotyping snp microarray - you won't see anything you aren't already looking for. NGS is sequencing, so that allows you to do actual discovery. You should read "Application of second-generation sequencing to cancer genomics." http://bib.oxfordjournals.org/content/11/5/524.abstract

[David Quigley]

Jeremy, I think you're confusing GWAS-prediction of breast cancer susceptibility with prognostic prediction based on gene expression in tumors. See van 't Veer Nature 2002, van 't Veer NEJM 2002 for details.

[Jeremy Leipzig]

David, ok thx for pointing that out. That should keep the discussion on expression microarray vs RNA-Seq issue.

[Travis]

Thanks for the responses and I'll take a look at the manuscript when I get to the office.

I guess I am primarily more concerened about classifier generation previously performed with expression arrays. Has antone tried it yet? Is expense vs microarrays still perhaps an issue? Or data volumes? I think Mammaprint for example initially was developed with 70 patients...

[Travis]

Thanks for the responses and I'll take a look at the manuscript when I get to the office. I guess I am primarily more concerened about classifier generation previously performed with expression arrays. Has anyone tried it yet? Is expense vs microarrays still perhaps an issue? Or data volumes? I think Mammaprint for example initially was developed with 70 patients..

[Junfeng]

Hi Travis, I am also interested in this application. Do you have any plan for this application now?