Comments on: David Goldstein Proves Himself Wrong

By: Rare variants versus common variants in complex disease is a political, not a scientific, debate -Gene Expression

Sun, 14 Mar 2010 20:33:40 +0000

[...] to many of the recently identified signals reported in genome-wide association studies”, is almost certainly wrong (depending on what you mean by “likely” or “contribute to” or [...]

By: Alejandro Montenegro-Montero

Alejandro Montenegro-Montero — Wed, 10 Feb 2010 16:07:36 +0000

Your post was selected as part of my Picks of the Week, of molbio-related blog posts aggregated to RB.

http://amontenegro.blogspot.com/2010/02/gwas-under-attack-historical.html

By: david

david — Mon, 01 Feb 2010 16:42:18 +0000

I agree with Patrick.Of course we know that some common SNPs found in GWAS will be in LD with high penetrant variants,some common and some rare.see AMD;but in advance of sequencing data we have no idea how much of the SNP associations, in which disorders can be attributed to these phenomena. D Goldstein’s graph/sums suggest a minority.

By: patrick

patrick — Sun, 31 Jan 2010 16:06:25 +0000

D. Goldstein, has in the present paper constructed a logical sophism. In brief he says, if a causative rare variants is creating a signal at a given locus then the common variant at this locus with which the rare hitchikes will also be associated. This first point is true but have very rarely happened in the case of well confirmed replicated variants. The second step he does in its logic and that the media coverage is conveying is when you have common variant associated it is because of rare variants. This second step is unfounded and at least is unlikely to be a general scenario.
Yes sequencing will bring data, but in advance one cannot assume he or she knows what will be the results.

By: Luke

Luke — Thu, 28 Jan 2010 10:48:10 +0000

Yeah, the authors mention this. We can work through the numbers (all values extracted from the graph using GIMP), to see how it changes things.

The ‘rare’ peak at 0.085 on the green line (the common synthetic associations) is at 2.8, and at a frequency of 0.5 it is 0.3 – the rare peak is 9.2 times higher than the mid point. For the known associations, the value at 0.085 is 0.8, and at 0.5 is 1.2, giving a ratio of 0.68.

The question is, will correcting for sampling bias bring the predicted rare:common ratio (9.2) into line with the actual rare:common ratio (0.68)?

We can calculate how much the predicted ratio will change once we take into account bias, by noting that a) we can take the ratio of 8.5% to 50% SNPs for the Illumina 1M chip from the graph (it is 1.37), and b) we can predict the actual ratio of 8.5% to 50% SNPs using Kimura’s neutral allele frequency ratio (see here):

(f2/f1)*((1-f1)/(1-f2))^(M-1)

We use M=0.1, as you generally do, but changing this value doesn’t significantly change our results. The ratio is 3.4, giving an overall sampling bias of 3.4/1.37 = 2.5.

This sampling bias changes the predicted rare:common ratio from 9.2 to 3.7; this is still significantly higher than the observed ratio of 0.68, and thus we can still conclude that there are very few, if any, synthetic associations amoungst the known GWAS hits.

Quite why the authors didn’t do this simple analysis, given that they already had all the data available, is an open question.

By: smitty

smitty — Thu, 28 Jan 2010 05:20:14 +0000

Interesting point, but it looks like they used the distribution of all simulated variants rather than the distribution that you would see on the types of chips that are used in GWAS, which have an increasing bias against lower frequency SNPs. I bet if they ran the analysis to take into account that bias that it would pull that green line towards the center and make it look more like GWAS.

By: Tweets that mention David Goldstein Proves Himself Wrong « Genetic Inference -- Topsy.com

Thu, 28 Jan 2010 04:34:44 +0000

[...] This post was mentioned on Twitter by Dan Vorhaus, ResearchBlogging.org, Luke Jostins, jose bras, keith grimaldi and others. keith grimaldi said: Hope ur right! RT@lukejostins My take on the 'syntheic assoc' paper and press: David Goldstein Proves Himself Wrong http://bit.ly/dzD8O4 [...]