| Users Online |
 Guests Online: 3
No Members Online
Registered Members: 152
Newest Member: lenalou3
|
| Login |
Not a member yet? Click here to register.
Forgotten your password? Request a new one here.
|
|
| DICE in Genetic Association Studies of Complex Traits |
 This multidimensional approach requires to develop statistical methods able to handle multiple variable loci, possibly in several genes, and to detect, among all measured polymorphisms, those which, alone or in combination, may influence the phenotype. Indeed, there is increasing evidence that, even in the absence of significant marginal effects, polymorphisms may exhibit epistatic effects on complex traits that are detectable only by a multilocus approach.
In this context, we propose an automated method for exploring the effects of several polymorphisms (and other non-genetic covariates) in the framework of association studies involving any kind of phenotype (quantitative or binary). This method, called DICE ([u]D[/u]etection of [u]I[/u]nformative [u]C[/u]ombined [u]E[/u]ffects), combines the advantages of the regressive approaches in terms of modeling and interpretation of effects, with those of data exploration tools. Importantly, the approach considers the problem of interaction between polymorphisms as an effect of interest and not as a nuisance effect. It is therefore well suited to the exploration of the spectrum of polymorphisms within candidate genes and more generally, within biological systems.
The forward selection approach is based on the principle of parsimony, the principle of marginality and the information theory paradigm. The algorithm compares at each step a wide variety of models and chooses the one(s) that provide(s) the best approximation to the data, while having the least number of parameters. To avoid difficulties related to the null-hypothesis testing theory, such as the choice of a significance level (especially for non-independent tests), the selection for the “best” approximating model(s) is based on an information criterion to be minimized. The application of the method to several real data samples revealed that the approach was able to recover results already found using other techniques, but could also detect effects not previously described which should deserve further detailed investigation.
Among the applications, there were the exploration of associations between polymorphisms of :
1 - the P-selectin ([i]SELP[/i]) gene and myocardial infarction (MI) (Genome Research 2003 13:1952-60);
2 - the Cholesteryl Ester Transfer Protein ([i]CETP[/i]) gene and plasma high-density lipoprotein (HDL)-cholesterol concentration, taking into account alcohol consumption (Genome Research 2003 13:1952-60);
3 - several genes belonging to the Renin-Angiotensin-Aldosterone (RAA) system and MI (Genome Research 2003 13:1952-60);
4 - the highly polymorphic apolipoprotein ([i]APOB[/i]) gene and plasma apoB levels (article submitted);
5 - a system of genes coding for cellular adhesion molecules and 2 phenotypes : the presence of femoral plaque and the carotid intima-media thickness (article in progress).
[url=http://ecgene.net/genecanvas/modules/icontent/index.php?page=66]Additional results[/url] are available at our web site, section Articles Supplements, Genome Research 2003 13:1952-60
[url=http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12902385&dopt=Abstract]Link to publication[/url]
Note that DICE has been implemented in the R language for the present applications. We are currently developing an optimized version of the algorithm in the C language, which will be more efficient to accommodate a large number of variables. This program will be made available at our web site.
|
| Ratings |
|
Rating is available to Members only.
Please login or register to vote.
No Ratings have been Posted.
|
|
