Comments on Genetic Future: The challenges of psychiatric genetics

Hi Sarah,

That's a really good question, an...

2008-07-18T21:09:00.000+10:00

Hi Sarah,

That's a really good question, and one I've been thinking about quite a bit recently in the context of massive cohort studies (e.g. the 11 million subjects in the planned P3G consortium's database). I should really do some formal power analysis, but I gather from talking to various people that such enormous cohorts will be sufficient to detect most genetic variants with a "useful" effect on disease risk (useful in the sense that there is some genuine predictive value). However, as soon as you start trying to look at the set of all possible gene-gene and gene-environment interactions, the multiple comparison problem is insurpassable - even if you could use the entire human population of the Earth, there's simply no way to correct for false positive errors without obliterating your ability to detect true positives.

The solution will have to be constraining the comparison space - for instance, rather than examining all possible gene-gene pairs, we instead focus only on those pairs where both genes fall into a similar functional category. This approach means that many, many real effects will be missed, but at least we will be able to filter some useful interactions out of the noise. And of course, our ability to define "likely" interactions between genes will depend on our level of understanding of gene function, which is pretty atrocious right now but will steadily improve over time.

I hope that helps!

I find point three very interesting:

"Stati...

2008-07-17T11:10:00.000+10:00

I find point three very interesting:

"Statistical issues associated with examining so many data-points from each patient and control would greatly increase the required sample sizes, driving costs and computational requirements up even higher"

It seems like the complexity of the questions asked will eventually reach a point where the world population is not a sufficient sample size, particularly if you are also considering combinatorial effects between mutations.

I don't have a good feel for how complicated the questions have to be to reach this point though. (How many hypotheses can you test if your sample size is everyone on the planet?) If you know anything about this I would love to hear more.

Heterogeneity in exposure to environmental factors...

2008-07-13T11:27:00.000+10:00

Heterogeneity in exposure to environmental factors may also explain why psychiatric genetics haven't been as successful, especially if a condition arises due to (as yet unidentified) gene-environment interactions. Childhood environmental factors like abuse are associated with certain disorders, and these may (or may not) manifest more strongly in the presence of certain genetic backgrounds.

Heritability measured through twin studies is commonly thought of as the "genetic component" of risk, but it actually also contains gene-environment interactions, since environment is assumed to be identical for identical twins and for fraternal twins. So when variance in risk for a condition is found to be "80% heritable" a lot of that variance may still be due to environmental variability.

Comments on Genetic Future: The challenges of psychiatric genetics

Hi Sarah,That's a really good question, an...

I find point three very interesting:"Stati...

Heterogeneity in exposure to environmental factors...

Hi Sarah,

That's a really good question, an...

I find point three very interesting:

"Stati...