Volume 6 Supplement 1
Genome-wide linkage and association mapping of disease genes with the GAW14 simulated datasets
© Carter et al; licensee BioMed Central Ltd 2005
Published: 30 December 2005
We combined the results of whole-genome linkage and association analyses to determine which markers were most strongly associated with Kofendrerd Personality Disorder. Using replicate 1 from the Genetic Analysis Workshop 14 Aipotu, Karangar, Danacaa, and New York City simulated populations, we determined that several markers showed significant linkage and association with disease status. We used both SNP and microsatellite markers to determine patterns and chromosomal regions of markers. Three consistently associated markers were C01R0050, C03R0280, and C10R0882. Using generalized linear mixed models, we modelled the effect of the three predefined phenotypic categories on disease status and concluded that the phenotypes defining the "anxiety-related" category best predicted the outcome.
Whole-genome linkage analyses involve looking for coinheritance of chromosomal regions with disease in families. Association studies seek to determine differences in the frequency of genetic variants between individuals exhibiting or not exhibiting a phenotype of interest (commonly case-control status). Family-based association studies utilize the available pedigree genetic variations to determine whether the transmission of particular genetic variants is associated with disease status. The results of linkage and association studies have been successfully combined in many analyses to refine the location of disease genes and to test the involvement of candidate genes in disease. The aim of this contribution was to perform linkage analyses, in combination with association analyses, on replicate 1 of the simulated Genetic Analysis Workshop 14 (GAW14) data to determine which markers or regions of markers are associated with Kofendrerd Personality Disorder (KPD).
Recoding the data
The GAW14 problem 2 description states because of the "varied phenotypes" for KPD, the "nosology for KPD falls into three different classifications", and that all three are used in diagnosis. The three main groups of phenotypes are indicative of three different methods used by each population for disease ascertainment. The different ascertainment methods and phenotypic categories suggest that complex interactions may be a key factor in identifying the causes and genetic determinants of KPD. Because we were blind to the simulated dataset answers, we chose to recode the data into these three additional grouped phenotypes to determine if complex combinations of phenotypes are of importance, in addition to examining the relationship between individual phenotypes and affection status. We chose replicate 1 as a representative data set for each of the four simulated populations. The first category, consisting of phenotypes a through e, is referred to as "communally shared emotions" (CSE). This was constructed in the data by assigning a positive affection status to an individual if they possessed at least one of phenotypes a through e, and assigning an individual a negative affection status otherwise. This procedure was similarly performed for the second category, consisting of phenotypes f through i, termed "behavioral-related" (BR) and for the third category, comprising phenotypes j through l, referred to as "anxiety related" (AR). This recoding procedure allowed us to assess affection status not only in terms of an overall status, but also in terms of the three different methods adopted by the different populations for deciding disease ascertainment.
To perform linkage analysis on the simulated datasets, we used the MERLIN  pedigree analysis software package. We performed a nonparametric linkage analysis using primary affection status and CSE, BR, and AR as binary outcomes. JLGRAPH  was used to generate linkage graphs for each chromosome for each population from the MERLIN results.
To perform association analysis on the binary traits for the simulated datasets, we used the computer program QTDT  to perform family-based tests. We performed an association analysis using affected individuals, including producing empirical p-values. The QTDT result files were input to JLGRAPH to produce association graphs for each chromosome for each population.
Regions of interest
The results from the linkage and association analyses were collated to provide a list of potential regions of interest for further study. Each of the 917 SNPs and 416 microsatellites markers were examined to determine their significance (in terms of both linkage and association) for affection, CSE, BR, and AR, for each of the four populations. Marker regions that appeared to be significant for both linkage and association were closely examined. Candidate packets of markers consisting of potentially important SNPs were "purchased" in order to analyze chromosomal regions of interest in fine detail. The procedures outlined above for association and linkage analysis were then repeated, this time incorporating the new marker sets.
To determine the effectiveness of each of the three newly defined categories for disease ascertainment, the data were modelled using a generalized linear mixed model (GLMM). Affection status was used as a binary outcome, with CSE, BR, and AR as factored predictor variables and family ID as a clustering variable. A mixed effects model was used so that a random intercept could be fitted, defining individuals within families to be correlated.
Linkage and association
Markers determined to be the most significant from linkage and association analysis
Chromosome 1 0.52 M
Chromosome 1 2.40 M
Chromosome 3 0.35 M
Chromosome 3 2.94 M
Chromosome 5 0.20 M
Chromosome 7 0.54 M
Chromosome 7 2.12 M
Chromosome 9 0.70 M
Chromosome 10 0.75 M
After "purchasing" more packets of markers and analyzing these in conjunction with the marker information already provided, several chromosomal regions showed significant linkage and association with disease status. The additional fine mapped packets available for download contained mostly SNP markers and we consequently determined that SNP markers would be of higher importance than microsatellites. In particular, regions surrounding the SNP markers C01R0050, C03R0280, and C10R0882 were examined. We describe the region surrounding marker C03R0280 as follows.
Disease ascertainment variables as predictors of disease
Phenotypic data were modelled using GLMMs to determine how effective each of the three categories used for disease ascertainment were at predicting overall affection status for each population. For all four populations in the simulated data, AR was the only category that effectively predicted disease status (p-value < 0.05).
Table 1 highlights linkage and association p-values for SNP markers across the ten chromosomes. While both markers from chromosome 1 had the most significant p-values, closer examination of these reveals our reason for choosing chromosome 3. The two chromosome 1 markers from Table 1 were localized to linkage with affection status in the New York City population only. In comparison, marker C03R0280 showed significant linkage in three of the four populations, and C03R0199 in two populations (results not shown). Because the individual p-values for chromosome 1 were not replicated across more than one population, we examined chromosome 3, in particular marker C03R0280, in more detail.
The microsatellite datapoints in Figure 1C and 1D were more sparsely located, as there were lower total numbers of markers, but overall trend pattern followed those of the SNP markers. While Table 1 showed only highlights from SNP markers, chromosomal regions such as from 2.5 M to 2.9 M in chromosome 3 returned significant results for both linkage and association for both SNP and microsatellite markers (other chromosomes not shown).
By defining the three new phenotypic categories we have created an effective method for determining the particular category (or categories) that contributed to significant linkage and association for affection status within markers for each population. This flexibility enabled us to locate markers such as C03R0280 and determine that this particular marker had significant p-values for linkage across all populations for affection status and two of the three additional categories. Creating phenotype categories allowed us to examine groups of phenotypic effects, and to determine the contribution of subsets to the overall disease status. This approach can provide very valuable information for conducting further analysis by narrowing down target phenotypes or phenotypic groups.
Through linkage and association analyses, markers C01R0050, C03R0280, and C10R0882 and markers surrounding these were found to be associated with affection status and the three phenotypic categories we defined (to varying degree within each region). Given time and "budgetary" constraints imposed on GAW14 participants, we successfully identified three gene regions, one within each of the three regions examined. While we did not find all the disease-associated genes contained within the GAW14 simulated datasets, we were successful in locating genes in the regions we focused on, indicating that our linkage and association mapping approach can successfully identify genes.
Anxiety related phenotypes j-l
Behavior related phenotypes f-i
Communally shared emotions phenotype a-e
Genetic Analysis Workshop
Generalized linear mixed model
Kofendrerd Personality Disorder
The work was supported by the National Health and Medical Research Council of Australia (Enabling Grant 303312).
- Abecasis GR, Cherny SS, Cookson WO, Cardon LR: Merlin-rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet. 2002, 30: 97-101. 10.1038/ng786.View ArticlePubMedGoogle Scholar
- Carter KW, Palmer LJ: JLGraph: Java Linkage Graph. 2004, [http://www.genepi.com.au/projects/jlgraph]Google Scholar
- Abecasis GR, Cardon LR, Cookson WO: A general test of association for quantitative traits in nuclear families. Am J Hum Genet. 2000, 66: 279-292. 10.1086/302698.PubMed CentralView ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.