Too much information

I am becoming increasingly convinced that the biggest problem facing biomedical research going forward is how to deal with all the information. One aspect of this is the multiple variations in a single gene and how to sort out which, if any, are important in predisposing to disease. As a review (Holloway et al. in linked article, subscribers only, alas) of the association between variations in the beta-adrenergic receptor (BAR) gene and asthma puts it:

The ever increasing availability of data on single nucleotide polymorphisms in the human genome and the vast range of clinical phenotypes against which they can be tested for association has led to an explosion of publications in the literature reporting associations between polymorphisms in candidate genes and disease-related phenotypes, with allergy and asthma being no exception. It has become readily apparent that few reported genetic associations can be replicated unequivocally, that the first published report is usually a poor guide to the final conclusions regarding a particular association, and that if it is correct, it usually greatly overestimates the contribution of the polymorphism to disease risk.

Basically, when you do small studies you often find correlations which may not represent underlying reality.

The BAR polymorphism field is know in the situation where effects are seen, at least in severity and response to medications, but the effects are in the opposite direction one would expect from functional data.

Another review from the same issue of JACI (Journal of Allergy and Clinical Immunology, article is Bochner and Busse) notes the following study

Raby et al performed a family-based study with 652 nuclear families. Seventeen ADAM33 single nucleotide polymorphisms (SNPs) were genotyped. Because no single SNP had an association with asthma, the possibility exists that the association between ADAM33 and asthma might be found in only selected populations with very specific characteristics

17 polymorphisms (a polymorphism is a change in a single DNA base) to study in one gene. It doesn’t take a statistician to realize that if you study each variant and several outcomes you’ll end up with a lot of spurious correlations.

And things are only made more complex by the fact that the polymorphisms aren't independent but tend to be linked to one another in what are called haplotypes.

How a clinician like me can make sense of all the data published is another aspect of the overall issue.