Comparison of several sequence-based association methods in pedigrees
© Mathew et al.; licensee BioMed Central Ltd. 2014
Published: 17 June 2014
Genome-wide association studies are very powerful in determining the genetic variants affecting complex diseases. Most of the available methods are very useful in detecting association between common variants and complex diseases. Recently, methods to detect rare variants in association with complex diseases have been developed with the increasingly available sequencing data from next-generation sequencing. In this paper, we evaluate and compare several of these recent methods for performing statistical association using whole genome sequencing data in pedigrees. Specifically, functional principal component analysis (FPCA), extended combined multivariate and collapsing (CMC) method for families, a generalized T2 method, and chi-square minimum approach were compared by analyzing all the genetic variants, common and rare, of both the real data set and the simulated data set provided as part of Genetic Analysis Workshop 18.
With advances in genotyping technologies, genome-wide association studies (GWAS) became a very popular procedure to identify disease genes and other traits by conducting statistical tests on many thousands of single-nucleotide polymorphisms (SNPs). The procedure has great potential for discovering genetic variants influencing complex diseases. However, these procedures have discovered loci that account only for a small percentage of phenotypic variance . One of the reasons for this difficulty may be that rare variants might explain disease susceptibility [2–4]. Recently, several methods have been developed to determine the influence of rare variants on complex diseases. These methods differ from the traditional methods of testing where the focus has been on individual common variants. It is understood that those variants with a population frequency greater than 5% are considered to be common variants, those with less than 1% population frequency as rare variants, and the rest as low-frequency variants . The common variants are believed to be from distant ancestors, whereas rare variants are from recent ancestors . Most of these methods assume the individuals are independent and are designed for population-based data. Only recently have several methods been developed that can perform statistical association of sequence data in pedigrees. In this paper, we used functional principal component analysis (FPCA) , the generalized T2 approach , the combined multivariate and collapsing (CMC) test for family data [2, 4], and the chi-square minimum approach for family data  to analyze association of the dichotomous hypertension trait with all genetic variants, common and rare, of the real data set and all replicates of chromosome 3 of the simulated data set provided by Genetic Analysis Workshop 18 (GAW 18) . We compared the results to assess the merits of these methods.
An extension of the generalized T2 test  for family-based association studies is provided by Zhu and Xiong . The test statistic is given by , where T2 is the generalized T2 statistic , and P corr [4, p. 1030] is the correction factor to account for the familial correlation in the pedigree data. A similar extension of the CMC test is also developed and is provided by equation (15) in Zhu and Xiong . The test statistic is given by , where T CMC is the CMC statistic for the population-based association test, and P corr is the correction factor to adjust T CMC statistic so that it is valid for pedigree data. The FPCA statistic for the population-based association test in Luo et al  also has a similar extension for family data , and is given by , where T FPCA and P corr are defined as in the previous test statistics for pedigree data. Also, , T CMCF , and T FPCAF have chi-square distributions . The chi-square minimum statistic chooses the minimum of the p-values from the individual chi-square tests for each genetic variant from a genomic region. The chi-square minimum statistic (Chi_min) adjusts for relatedness of pedigree members using P corr .
We applied the above 4 methods to analyze the real data set from all odd-numbered chromosomes using hypertension status at exam 1 as the phenotype. The genotypes at each variant are coded as 0, 1, or 2 for aa, Aa, and AA, where allele A is the minor allele. The start and end boundary of all the human genes were obtained from hg19 genome assembly at NCBI. The genetic variants within 1 gene were analyzed together as each gene is 1 natural functional unit. For the FPCA method, if there are too few genetic variants, that is, less than 3, the estimate of the functional relation of the allele counts across the genetic variants will be far off. Consequently, genes with fewer than 3 genetic variants were not analyzed. The significant genes from the 4 methods were then compared with the findings from previous GWAS for genes associated with blood pressure.
To examine the type I error rate and power of the 4methods, we applied these methods to the 200 replicates of the simulated data set and analyzed the data from chromosome 3. As in the real data analysis, we chose the hypertension status at exam 1 as the phenotype.
Number of significant genes out of 10,580 genes in the odd-numbered human chromosomes of the real data set at various significance levels
4.7 × 10−6
Number of overlapped genes associated with blood pressure from GWAS findings at various significance levels
4.7 × 10−6
Type I error probability estimates by FPCA, Chi_min, T2, and CMC methods from all 200 replicates of chromosome 3 of the simulated data set
Estimates of power by FPCA, Chi_min, T2, and CMC methods from all 200 replicates of chromosome 3 of thesimulated data set
With the increasingly available sequence data from the next-generation sequencing technologies, it is important for a statistical association method to handle both common and rare genetic variants. It is also important for these methods to handle data from pedigrees because rare genetic variants are enriched in families with multiple affected individuals, which could confer more statistical power. From our analysis of the real data, T2 seems to be a better method than the other 3 methods because it finds more significant genes at low significance levels. At the Bonferroni corrected p-value of 4.7 × 10−6 , T2 identified the genes CASZ1, ADAMTS8, NUCB2, ABCC8, SLC4A7, MAP4, CASR, EBF1, PLEKHA7, SOX6, ULK4, and MECOM. The last 4 genes were also identified by the CMC method. All the genes mentioned above were found to be associated with blood pressure, in particular ULK4 and PLEKAH7 by Levy et al , and MAP4 by Wain et al .
As with GWAS, we need to keep a low significance level to account for multiple testing. We note from the analysis of the simulated data that FPCA has empirical type I error rate much less than the nominal value, making it very conservative. The Chi_min method has inflated type I error rate. The type I error rates by T2 and CMC are close to the nominal value. Also, T2 has better power than CMC, which is consistent with the result from the real data.
From the analysis of the data sets we find that T2 is a better method, which is different from the findings of Zhu and Xiong , which suggest that FPCA is a better procedure. There are 2 possible reasons why FPCA performs less well here. One reason may be that the SNPs in the genes are sparse. If there are too few SNPs in 1 gene, the FPCA may not perform well because the number of SNPs is not enough to estimate the function describing the allele counts across the SNPs in the gene. A second reason may be that the assumption of a smooth function of the allele counts across the SNPs for the FPCA may not hold for the GAW 18data sets. We observed a large overlap between the results of CMC and T2. This mainly comes from the fact that CMC uses the T2 approach with common variants. There is also a tendency to pick up more genes with more variants for both CMC and T2 methods.
From the analysis results of both real and simulated data, T2 is a preferable method for pedigree-based association studies with whole-genome sequencing data because it controls the false positive rate and is more powerful than the other two methods with similar type I error rates.
The GAW18 whole genome sequence data were provided by the T2D-GENES Consortium, which is supported by NIH grants U01 DK085524, U01 DK085584, U01 DK085501, U01 DK085526, and U01 DK085545. The other genetic and phenotypic data for GAW18 were provided by the San Antonio Family Heart Study and San Antonio Family Diabetes/Gallbladder Study, which are supported by NIH grants P01 HL045222, R01 DK047482, and R01 DK053889. The Genetic Analysis Workshop is supported by NIH grant R01 GM031575.
This article has been published as part of BMC Proceedings Volume 8 Supplement 1, 2014: Genetic Analysis Workshop 18. The full contents of the supplement are available online at http://www.biomedcentral.com/bmcproc/supplements/8/S1. Publication charges for this supplement were funded by the Texas Biomedical Research Institute.
- Schork NJ, Murray SS, Frazer KA, Topol EJ: Common vs. rare allele hypotheses for complex diseases. Curr Opin Genet Dev. 2009, 19: 212-219. 10.1016/j.gde.2009.04.010.PubMed CentralView ArticlePubMedGoogle Scholar
- Li B, Leal SM: Methods for detecting associations with rare variants for common diseases: application to analysis of sequence data. Am J Hum Genet. 2008, 83: 311-321. 10.1016/j.ajhg.2008.06.024.PubMed CentralView ArticlePubMedGoogle Scholar
- Gorlov IP, Gorlova OY, Sunyaev SR, Spitz MR, Amos CI: Shifting paradigm of association studies: value of rare single-nucleotide polymorphisms. Am J Hum Genet. 2008, 82: 100-112. 10.1016/j.ajhg.2007.09.006.PubMed CentralView ArticlePubMedGoogle Scholar
- Zhu Y, Xiong M: Family based association studies for next generation sequencing. Am J Hum Genet. 2012, 90: 1028-1045. 10.1016/j.ajhg.2012.04.022.PubMed CentralView ArticlePubMedGoogle Scholar
- Lupski JR, Belmont JW, Boerwinkle E, Gibbs RA: Clan genomics and the complex architecture of human disease. Cell. 2011, 147: 32-43. 10.1016/j.cell.2011.09.008.PubMed CentralView ArticlePubMedGoogle Scholar
- Almasy L, Dyer TD, Peralta JM, Jun G, Wood AR, Fuchsberger C, Almeida MA, Kent JW, Fowler S, Blackwell TW, Puppala S, Kumar S, Curran JE, Lehman D, Abecasis G, Duggirala R, Blangero J, The T2D-GENES Consortium: The T2D-GENES Consortium: Data for Genetic Analysis Workshop 18: human whole genome sequence, blood pressure, and simulated phenotypes in extended pedigrees. BMC Proc. 2014, 8 (Suppl 1): S2-PubMed CentralView ArticlePubMedGoogle Scholar
- Xiong M, Zhao J, Boerwinkle E: Generalized T2 for genome association studies. Am J Hum Genet. 2002, 70: 1257-1268. 10.1086/340392.PubMed CentralView ArticlePubMedGoogle Scholar
- Luo L, Boerwinkle E, Xiong M: Association studies for next-generation sequencing. Genome Res. 2011, 21: 1099-1108. 10.1101/gr.115998.110.PubMed CentralView ArticlePubMedGoogle Scholar
- Barrett JC, Fry B, Maller J, Daly MJ: Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics. 2005, 263-265. 21Google Scholar
- Levy D, Ehret GB, Rice K, Verwoert C, Launer LJ, Dehghan A, Glazer NL, Morrison AC, Johnson AD, Aspelund T, et al: Genome-wide association study of blood pressure and hypertension. Nat Genet. 2009, 41: 677-687. 10.1038/ng.384.PubMed CentralView ArticlePubMedGoogle Scholar
- Wain LV, Verwoett GC, O'Reilly PF, Shi G, Johnson T, Johnson AD, Bochud M, Rice KM, Henneman P, Smith AV, et al: Genome-wide association study identifies six new loci influencing pulse pressure and mean arterial pressure. Nat Genet. 2011, 43: 1005-1011. 10.1038/ng.922.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. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.