Limits...
A 2-step strategy for detecting pleiotropic effects on multiple longitudinal traits.

Wang W, Feng Z, Bull SB, Wang Z - Front Genet (2014)

Bottom Line: We focus on estimation of the random effect that accounts for the subject-specific genetic contribution to the trait; fixed effects of other confounding covariates are also estimated.This first step enables separation of the genetic effect from other confounding effects for each subject and for each longitudinal trait.The proposed method can efficiently detect pleiotropic effects on multiple longitudinal traits and can flexibly handle traits of different data types such as quantitative, binary, or count data.

View Article: PubMed Central - PubMed

Affiliation: Department of Mathematics and Statistics, University of Guelph Guelph, ON, Canada.

ABSTRACT
Genetic pleiotropy refers to the situation in which a single gene influences multiple traits and so it is considered as a major factor that underlies genetic correlation among traits. To identify pleiotropy, an important focus in genome-wide association studies (GWAS) is on finding genetic variants that are simultaneously associated with multiple traits. On the other hand, longitudinal designs are often employed in many complex disease studies, such that, traits are measured repeatedly over time within the same subject. Performing genetic association analysis simultaneously on multiple longitudinal traits for detecting pleiotropic effects is interesting but challenging. In this paper, we propose a 2-step method for simultaneously testing the genetic association with multiple longitudinal traits. In the first step, a mixed effects model is used to analyze each longitudinal trait. We focus on estimation of the random effect that accounts for the subject-specific genetic contribution to the trait; fixed effects of other confounding covariates are also estimated. This first step enables separation of the genetic effect from other confounding effects for each subject and for each longitudinal trait. Then in the second step, we perform a simultaneous association test on multiple estimated random effects arising from multiple longitudinal traits. The proposed method can efficiently detect pleiotropic effects on multiple longitudinal traits and can flexibly handle traits of different data types such as quantitative, binary, or count data. We apply this method to analyze the 16th Genetic Analysis Workshop (GAW16) Framingham Heart Study (FHS) data. A simulation study is also conducted to validate this 2-step method and evaluate its performance.

No MeSH data available.


Related in: MedlinePlus

Most significant SNPs with their −log(p-value) from the simultaneous test compared with their significance levels from individual tests.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4202779&req=5

Figure 1: Most significant SNPs with their −log(p-value) from the simultaneous test compared with their significance levels from individual tests.

Mentions: In Step 2, we simultaneously test the association between each SNP and all estimated subject-specific effects corresponding to the four traits. We also test the association between each SNP and the estimated subject-specific effect for each trait. SNPs with p-value < 1.0 × 10−5 from the simultaneous tests (either score test or LRT test) are summarized in Table 8. We also compare their significance levels with those obtained by individual tests in Table 8. Note the p-value associated with each individual trait are adjusted via Bonferroni procedure for multiple testing. For easy comparison, results are also presented in Figure 1. In Figure 1, SNPs that are significantly associated with more than one traits generally have a higher −log(p-values) or equivalently a lower p-value. SNPs that are significantly associated with only one trait have a comparative −log(p-value) or equivalently a similar level of significance in p-value. On chromosome 8, nine SNPs are found by the simultaneous test to have a strong and significant association with at least one of the four traits. These SNPs are in the LPL gene or very close to this gene. The LPL gene encodes lipoprotein lipase, a triglyceride hydrolase that acts as a ligand factor for receptor-mediated lipoprotein uptake. According to the individual tests, these nine SNPs are significantly associated with HDL and TG but their p-values based on the union of the individual tests are consistently larger than the p-values based on the simultaneous test. Note that a larger p-value means a less significant level. These significant findings are consistent with other FHS analyses reported by Piccolo et al. (2009) and Ma et al. (2010). SNP RS1800775, less than 0.6 kb from the CEPT gene on chromosome 16, is significant in both the simultaneous test (p-value = 6.51 × 10−12) and the union of the individual tests (p-value = 2.32 × 10−10). The CEPT gene mediates the transfer of cholesterol ester from HDL to other lipoproteins. So, not surprising, this SNP is strongly associated with HDL in the individual test (p-value = 5.81 × 10−11). This result is also confirmed by Sull et al. (2012) and Sarzynski et al. (2011) in the analyses of independent data sets.


A 2-step strategy for detecting pleiotropic effects on multiple longitudinal traits.

Wang W, Feng Z, Bull SB, Wang Z - Front Genet (2014)

Most significant SNPs with their −log(p-value) from the simultaneous test compared with their significance levels from individual tests.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4202779&req=5

Figure 1: Most significant SNPs with their −log(p-value) from the simultaneous test compared with their significance levels from individual tests.
Mentions: In Step 2, we simultaneously test the association between each SNP and all estimated subject-specific effects corresponding to the four traits. We also test the association between each SNP and the estimated subject-specific effect for each trait. SNPs with p-value < 1.0 × 10−5 from the simultaneous tests (either score test or LRT test) are summarized in Table 8. We also compare their significance levels with those obtained by individual tests in Table 8. Note the p-value associated with each individual trait are adjusted via Bonferroni procedure for multiple testing. For easy comparison, results are also presented in Figure 1. In Figure 1, SNPs that are significantly associated with more than one traits generally have a higher −log(p-values) or equivalently a lower p-value. SNPs that are significantly associated with only one trait have a comparative −log(p-value) or equivalently a similar level of significance in p-value. On chromosome 8, nine SNPs are found by the simultaneous test to have a strong and significant association with at least one of the four traits. These SNPs are in the LPL gene or very close to this gene. The LPL gene encodes lipoprotein lipase, a triglyceride hydrolase that acts as a ligand factor for receptor-mediated lipoprotein uptake. According to the individual tests, these nine SNPs are significantly associated with HDL and TG but their p-values based on the union of the individual tests are consistently larger than the p-values based on the simultaneous test. Note that a larger p-value means a less significant level. These significant findings are consistent with other FHS analyses reported by Piccolo et al. (2009) and Ma et al. (2010). SNP RS1800775, less than 0.6 kb from the CEPT gene on chromosome 16, is significant in both the simultaneous test (p-value = 6.51 × 10−12) and the union of the individual tests (p-value = 2.32 × 10−10). The CEPT gene mediates the transfer of cholesterol ester from HDL to other lipoproteins. So, not surprising, this SNP is strongly associated with HDL in the individual test (p-value = 5.81 × 10−11). This result is also confirmed by Sull et al. (2012) and Sarzynski et al. (2011) in the analyses of independent data sets.

Bottom Line: We focus on estimation of the random effect that accounts for the subject-specific genetic contribution to the trait; fixed effects of other confounding covariates are also estimated.This first step enables separation of the genetic effect from other confounding effects for each subject and for each longitudinal trait.The proposed method can efficiently detect pleiotropic effects on multiple longitudinal traits and can flexibly handle traits of different data types such as quantitative, binary, or count data.

View Article: PubMed Central - PubMed

Affiliation: Department of Mathematics and Statistics, University of Guelph Guelph, ON, Canada.

ABSTRACT
Genetic pleiotropy refers to the situation in which a single gene influences multiple traits and so it is considered as a major factor that underlies genetic correlation among traits. To identify pleiotropy, an important focus in genome-wide association studies (GWAS) is on finding genetic variants that are simultaneously associated with multiple traits. On the other hand, longitudinal designs are often employed in many complex disease studies, such that, traits are measured repeatedly over time within the same subject. Performing genetic association analysis simultaneously on multiple longitudinal traits for detecting pleiotropic effects is interesting but challenging. In this paper, we propose a 2-step method for simultaneously testing the genetic association with multiple longitudinal traits. In the first step, a mixed effects model is used to analyze each longitudinal trait. We focus on estimation of the random effect that accounts for the subject-specific genetic contribution to the trait; fixed effects of other confounding covariates are also estimated. This first step enables separation of the genetic effect from other confounding effects for each subject and for each longitudinal trait. Then in the second step, we perform a simultaneous association test on multiple estimated random effects arising from multiple longitudinal traits. The proposed method can efficiently detect pleiotropic effects on multiple longitudinal traits and can flexibly handle traits of different data types such as quantitative, binary, or count data. We apply this method to analyze the 16th Genetic Analysis Workshop (GAW16) Framingham Heart Study (FHS) data. A simulation study is also conducted to validate this 2-step method and evaluate its performance.

No MeSH data available.


Related in: MedlinePlus