Please use this identifier to cite or link to this item:
|Title:||Choosing between per-genotype, per-allele, and trend approaches for initial detection of gene-disease association|
John R. Thompson
University of Newcastle Faculty of Medicine and Health Sciences
University of Leicester
|Citation:||Journal of Applied Statistics. Vol.36, No.6 (2009), 633-646|
|Abstract:||There are a number of approaches to detect candidate gene-disease associations including: (i) 'pergenotype', which looks for any difference across the genotype groups without making any assumptions about the direction of the effect or the genetic model; (ii) 'per-allele', which assumes an additive genetic model, i.e. an effect for each allele copy; and (iii) linear trend, which looks for an incremental effect across the genotype groups. We simulated a number of gene-disease associations, varying odds ratios, allele frequency, genetic model, and deviation from Hardy-Weinberg equilibrium (HWE) and tested the performance of each of the three methods to detect the associations, where performance was judged by looking at critical values, power, coverage, bias, and root mean square error. Results indicate that the per-allele method is very susceptible to false positives and false negatives when deviations from HWE occur. The linear trend test appears to have the best power under most simulated scenarios, but can sometimes be biased and have poor coverage. These results indicate that of these strategies a linear trend test may be best for initially testing an association, and the per-genotype approach may be best for estimating the magnitude of the association. © 2009 Taylor & Francis.|
|Appears in Collections:||Scopus 2006-2010|
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.