Please find below details of two talks which may be of interest. The meeting is free but please register with Kath Forrest [log in to unmask]<mailto:[log in to unmask]> if you plan to attend
University of Liverpool
Room 5.07 Cedar House
Friday 18th June 2010 at 2pm
Genome-wide association analysis of melanoma and related phenotypes
Jenny Barrett, University of Leeds
Genome-wide association (GWA) studies are used to investigate genetic susceptibility to disease by comparing the frequencies of hundreds of thousands of genetic variants, spread across the whole genome, between cases and controls. On behalf of the GenoMEL consortium, we have carried out a GWA study of melanoma (Bishop et al, Nature Genetics, 2009) based on ~1650 cases and ~4300 controls from numerous mainly European countries which identified several disease-susceptibility variants, and we are in the process of analysing a second phase of this study with similar sample size. Most risk variants identified so far are in genes related to nevus phenotype or pigmentation, phenotypes that have been shown in epidemiological studies to be strongly related to melanoma. GWA studies have also been carried out to investigate the genetic basis of these phenotypes. I will discuss some of the statistical issues encountered in GWA studies, especially when combining samples from different countries, and also present the results of a joint analysis of genes and phenotype on risk in a study of melanoma cases and controls from Leeds.
Replication of GWA findings - problems and possible solutions
Cosetta Minelli, European Academy of Bozen/Bolzano (EURAC), Italy
Genome-wide association (GWA) studies, which examine variation across the whole genome, represent hypothesis free investigations aimed at identifying novel associations of genetic variants with a given trait, be it a quantitative phenotype or a binary disease outcome. GWA studies imply testing a huge number of variants, which poses the issues of multiple testing and how to best adjust for it. A fundamental challenge is to develop statistical methods that can minimise the possibility of false positives while conserving enough power to detect true associations. The current way of dealing with multiple testing is through definition of strict GWA significance thresholds (p-value < 10-7 - 10-8), although the gold standard remains the replication of the best GWA findings (associations with lowest p-values) in independent samples. However, empirical and theoretical arguments have led researchers to question the value of such an approach, given that lack of replication often cannot exclude the presence of true association. The problem of false negative findings can be reduced by increasing statistical power, and this has stimulated the creation of international consortia performing meta-analysis to pool data across GWA studies. However, even consortia may not be able to reach adequate power, for example in the case of rare disease outcomes. I will discuss some of the issues surrounding lack of replication and possible solutions, including Bayesian approaches to incorporate prior biological evidence.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|