The Utah population has been described as being of Northern European descent, having similar inbreeding levels as other parts of the United States,23, 24
and having characteristics conducive to genetic studies including large family size and an interest in genealogy. High-risk pedigrees were ascertained through a Utah genealogy resource linked to hospital diagnosis data from Intermountain Health Care, the largest health care provider in Utah, and also through local advertisement and physician referral. These patients and their connecting relatives were contacted and sampled by the Genetic Research group at Intermountain Health Care at LDS Hospital between 1996 and 2000. Informed consent was obtained for each individual involved in the study. Institutional Review Board approval was obtained from Intermountain Health Care and the University of Utah.
Asthma affection status was determined by a comprehensive clinical evaluation. Study subjects completed a modified National Heart Lung and Blood Institute Collaborative Studies on the Genetics of Asthma questionnaire specific to asthma, and pulmonary function testing by spirometry was performed according to Intermountain Thoracic Society standards for individuals age 6 or older.25
In some cases, diagnosis also relied on available medical records, pre- and post-bronchodilator testing, or methacholine challenge testing.26
On the basis of the results of the clinical evaluation an experienced pulmonologist conferred a diagnosis of affected, unaffected or unclassifiable. A total of 1451 patients were seen, resulting in 744 affected individuals, 628 unaffected individuals and 79 unclassifiable individuals. The resulting diagnosis provides the basis for the phenotype definition specific to this analysis.
Study subjects provided a blood sample for DNA. Informative individuals in pedigrees with at least 2 sampled asthma cases were genotyped (n
=1314) on a set of 535 fluorescent dye-labeled microsatellite markers by Myriad Genetics. Markers spanned the entire genome, including the pseudoautosomal region of the X chromosome, with an average spacing between markers of 6.4
cM. The genetic map was developed internally using CRIMAP software on 3916 meioses from a set of high-risk Utah pedigrees ascertained for asthma and multiple other disorders. It corresponds closely to published deCODE maps.27, 28, 29
Inheritances were verified using PEDCHECK software.30
Inconsistencies were re-genotyped where possible. Unresolved inconsistencies, including Mendelian errors which occurred at a rate of 0.036%, were set to missing.
Linkage analysis was performed with MCLINK which estimates multipoint inheritance vectors using a Monte Carlo Markov Chain (MCMC) methodology employing a blocked Gibbs sampling method to infer phase.31
The advantage of the MCMC approach is the ability to analyze entire pedigrees without constraint on size or nonunilineal structure, allowing the analysis to take advantage of all inheritance information. MCLINK utilizes the robust multipoint statistic proposed by Goring and Terwilliger (hereafter referred to as TLOD),32
which has been implemented in MCLINK. The TLOD statistic uses multipoint inheritance vectors to estimate inheritance probabilities at specific marker position and is maximized over the recombination fraction (hence, theta-LOD, or TLOD), providing a multipoint linkage statistic that is robust to model misspecification.33
For each marker, the heterogeneity TLOD (het-TLOD) statistic was calculated with HOMOG software to account for any interfamilial heterogeneity.34
Unbiased marker allele frequencies were estimated from thousands of individuals in high-risk Utah pedigrees genotyped for multiple disorders.35
Parametric analyses were performed using general dominant and recessive models. Nonparametric methodologies are generally favored in linkage analysis of complex traits in small family structures because no assumptions about the mode of inheritance are required for these methods. However, it has been shown that in extended pedigree settings, analyses relying on non-parametric (or ‘model-free') methodologies will approach similar power as parametric analyses only when stringent assumptions are in place.36
Parametric linkage analyses utilizing general models have been shown to be effective in detecting evidence for linkage when information about the mode of inheritance is not well established.37, 38
Further, simulation has shown that in a complex disease setting, general parametric analyses are more powerful if they include both a dominant and recessive model as it is more critical to distinguish the mode of inheritance at the linked locus, and not that of the disease in general.37
We assumed a disease allele frequency of 0.005 and 0.05 for the dominant and recessive models, respectively; both models assumed a penetrance of 50% for gene carriers and 0.5% for nongene carriers.
Significance for genome-wide linkage was evaluated according to the thresholds established by Lander and Kruglyak39
additionally corrected for the two models analyzed. Using a previously published method, we established that the two models corresponded to 1.9 independent tests and we applied a Bonferonni correction to correct for this.40
As a result of the correction, LOD scores in excess of 3.6 were considered significant, and LOD scores in excess of 2.1 were considered suggestive.
A 1-LOD support interval based on the TLOD statistic was used to provide general boundaries for a region of interest. However, locus heterogeneity and random noise from unlinked pedigrees in the heterogeneity-TLOD calculation can shift the support interval. To delimit the regions of interest we also used an alternative recombinant-mapping approach based on pedigrees with evidence of linkage to the region of interest.27, 41
Linked pedigrees were considered as those with a classic multipoint LOD score >0.59 (nominal P
=0.05) within 30
cM centered on the TLOD peak. Within this region, recombination events within linked pedigrees can be estimated. A recombination event was identified by a reduction in haplotype sharing among affected individuals in the pedigree, and was defined as the outermost marker position of a 0.5 LOD unit decrease in a linked pedigree.27
We defined a localized region defined by the linked pedigrees as delimited by the outermost of three recombination events in any linked pedigree in either direction. The linked pedigree region of interest has been described as a 99% credible interval, thereby providing greater precision for localization than a 1-LOD support interval centered on the maximal TLOD statistic.42
A more formal description of our localization method is given by Camp et al.42
In linked pedigrees we identified all affected individuals who shared the segregating haplotype(s) that contributed to the LOD score, according to the model calculated. For pedigrees linked under the recessive model this included all affected individuals who clearly shared two of any segregating predisposition haplotypes in the pedigree. For pedigrees linked under the dominant model this included all affected individuals sharing the single segregating predisposition haplotype in the pedigree.