Improving the calculation of statistical significance in genome-wide scans

TY - JOUR

T1 - Improving the calculation of statistical significance in genome-wide scans

AU - Ängquist, Lars

AU - Hössjer, Ola

PY - 2005/10/1

Y1 - 2005/10/1

N2 - Calculations of the significance of results from linkage analysis can be performed by simulation or by theoretical approximation, with or without the assumption of perfect marker information. Here we concentrate on theoretical approximation. Our starting point is the asymptotic approximation formula presented by Lander and Kruglyak (1995, Nature Genetics, 11, 241-247), incorporating the effect of finite marker spacing as suggested by Feingold et al. (1993, American Journal of Human Genetics, 53, 234-251). We consider two distinct ways in which this formula can be improved. Firstly, we present a formula for calculating the crossover rate ρ for a pedigree of general structure. For a pedigree set, these values may then be weighted into an overall crossover rate which can be used as input to the original approximation formula. Secondly, the unadjusted p -value formula is based on the assumption of a Normally distributed nonparametric linkage (NPL) score. This leads to conservative or anticonservative p -values of varying magnitude depending on the pedigree set structure. We adjust for non-Normality by calculating the marginal distribution of the NPL score under the null hypothesis of no linkage with an arbitrarily small error. The NPL score is then transformed to have a marginal standard Normal distribution and the transformed maximal NPL score, together with a slightly corrected value of the overall crossover rate, is inserted into the original formula in order to calculate the p -value. We use pedigrees of seven different structures to compare the performance of our suggested approximation formula to the original approximation formula, with and without skewness correction, and to results found by simulation. We also apply the suggested formula to two real pedigree set structure examples. Our method generally seems to provide improved behavior, especially for pedigree sets which show clear departure from Normality, in relation to the competing approximations.

AB - Calculations of the significance of results from linkage analysis can be performed by simulation or by theoretical approximation, with or without the assumption of perfect marker information. Here we concentrate on theoretical approximation. Our starting point is the asymptotic approximation formula presented by Lander and Kruglyak (1995, Nature Genetics, 11, 241-247), incorporating the effect of finite marker spacing as suggested by Feingold et al. (1993, American Journal of Human Genetics, 53, 234-251). We consider two distinct ways in which this formula can be improved. Firstly, we present a formula for calculating the crossover rate ρ for a pedigree of general structure. For a pedigree set, these values may then be weighted into an overall crossover rate which can be used as input to the original approximation formula. Secondly, the unadjusted p -value formula is based on the assumption of a Normally distributed nonparametric linkage (NPL) score. This leads to conservative or anticonservative p -values of varying magnitude depending on the pedigree set structure. We adjust for non-Normality by calculating the marginal distribution of the NPL score under the null hypothesis of no linkage with an arbitrarily small error. The NPL score is then transformed to have a marginal standard Normal distribution and the transformed maximal NPL score, together with a slightly corrected value of the overall crossover rate, is inserted into the original formula in order to calculate the p -value. We use pedigrees of seven different structures to compare the performance of our suggested approximation formula to the original approximation formula, with and without skewness correction, and to results found by simulation. We also apply the suggested formula to two real pedigree set structure examples. Our method generally seems to provide improved behavior, especially for pedigree sets which show clear departure from Normality, in relation to the competing approximations.

KW - Adjusted approximation formula

KW - Allele sharing

KW - Approximation of distributions

KW - Crossover rate

KW - Deviation from Normality

KW - Extreme value formulas

KW - Genome-wide significance

KW - Hermite polynomials

KW - Marker density

KW - Nonparametric linkage analysis

UR - http://www.scopus.com/inward/record.url?scp=18944363606&partnerID=8YFLogxK

U2 - 10.1093/biostatistics/kxi025

DO - 10.1093/biostatistics/kxi025

M3 - Journal article

C2 - 15831574

AN - SCOPUS:18944363606

SN - 1465-4644

VL - 6

SP - 520

EP - 538

JO - Biostatistics

JF - Biostatistics

IS - 4

ER -