[GenABEL-dev] Interesting idea by Gulnara Svischeva

[Yury Aulchenko]

> I was going through the changelog of GenABEL and saw that Gulya's
> poly_eigen is the default method for polygenic() since version 1.6-8. I
> did a quick comparison on the computation speed of the three methods
> (polygenic_hglm, polygenic with polylik and with poly_eigen) on our
> population and even though I knew what to expect from Yurii's results, I
> am still impressed at the speedup! 

This was indeed a major step -- with multiple improvements after that -- and now 'polygenic' can estimate the model for say 3k people in few minutes (in contrast with few hours / days? in old implementation). This development is thanks to Gulya's brilliant decomposition idea, comments of other people about it, and -- in some part -- my coding. 

> I was wondering, is there a publication on this method/implementation?

Not yet, and it is not very clear (at least to me) how to sell this. There are many issues with moving this idea to the point of a pub -- basically, this type of ideas was around for years (e.g. William Astle's TurboGC is quite similar;  Paolo Bientinesi & Diego Fabregat suggested something related in some parts; looking into literature, I see similar ideas in gene expression analysis, e.g. from Tibshirani; ), and it is difficult to figure out how to serve such pub out. While Gulya developed these ideas independently of others, while for GenABEL and practical GWAS this is a big thing; theoretically, it seems to be not that new... Well, more discussion will follow. 

> Has there been any comparison of poly_eigen vs. the old method (in terms
> of variance, lambda etc.?

It is 100% identical (there is an GenA's RUnit test running 'old' vs 'new', and it passes checkEquals with tolerance of 1e-6 :) ). No approximation, just speed-up and still full ML. Cool, he?

best wishes,
Yurii

-------------------------------------------------------
Yurii Aulchenko, PhD, Dr. Habil.
Independent researcher and consultant
yurii [dot] aulchenko [at] gmail [dot] com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/genabel-devel/attachments/20111011/2f9e80d9/attachment.htm>