Validity of the family-based association test for copy number variant data in the case of non-linear intensity-genotype relationship.
View Article:
PubMed Central - PubMed
Affiliation: Wellcome Trust Sanger Institute, Hinxton, United Kingdom.
: Ionita-Laza and colleagues have proposed a family-based association test (FBAT) based on raw intensity data from copy number variant (CNV) assays rather than genotype calls [... This work is motivated by the difficulty of obtaining reliable discrete CNV calls owing to the limited resolution of CNV assays, especially for complex and multi-allelic CNVs... If this is not the case then the expected CNV data of the affected offspring (Xi) will differ from the midparent intensity, under the hypothesis of no association and for some pairs of parental genotypes... In this non-linear case, the contribution of each trio to the score statistic U does not have zero expectation conditional on the genotype of the parents... The result shown in 3 also shows that any population structure in the parental population that disrupts HWE also leads to a non-zero expectation for Ui in the non-linear case... Hence, the robustness to population structure does not extend from the linear to non-linear case... The trio design as developed in the FBAT CNV test can alleviate these concerns by comparing raw intensity data within families, rather than comparisons across cases and controls... In the non-linear intensity case, if the batch effect can be modelled using a family-specific covariate that acts on the raw intensity data in an additive manner, this covariate will have the same effect on the parental and offspring intensities... Therefore, the effect of this covariate will vanish in the computation of the score statistic... Thus, even in the case of a non-linear effect, the FBAT CNV statistic is expected to be robust to a range of technical artefacts... Approximately 250 of these 372 CNVs are deletions for which the intensity is typically non-linear with copy number... A small subset of CNVs showed significant association (Fig... In this case, the FBAT CNV test is appropriate and the initially proposed estimate of the score variance estimator V is consistent... Moreover, in a non-linear genotype-intensity context, the FBAT CNV test is robust to family-specific technical covariates. |
Related In:
Results -
Collection
License getmorefigures.php?uid=PMC3569870&req=5
Mentions: In array-generated CNV data, it is typical to observe a non-linear relationship between CNV genotype and raw intensity data. In the example of a common deletion presented in Figure 1, for which the intensity-CNV state relationship is clearly not linear, if both parents are homozygous with copy number states 0 and 2 then the affected offspring will be heterozygous and the CNV intensity will be systematically higher than the midparent intensity (Table 1). Hence, the score statistic Ui will be biased toward positive values. Conversely, if parents are both heterozygous with copy number states 1, Ui will be biased toward negative values (Table 1). In fact, if the mean positions for the three clusters (copy numbers 0, 1, and 2) in Figure 1 are denoted by a, b, and c, and the respective frequencies of the genotypes are f0, f1 and f2, a straightforward analysis of all possible parental and offspring genotypes (Table 1) shows that the marginal expectation of Ui is:which is in general different from 0. |
View Article: PubMed Central - PubMed
Affiliation: Wellcome Trust Sanger Institute, Hinxton, United Kingdom.
: Ionita-Laza and colleagues have proposed a family-based association test (FBAT) based on raw intensity data from copy number variant (CNV) assays rather than genotype calls [... This work is motivated by the difficulty of obtaining reliable discrete CNV calls owing to the limited resolution of CNV assays, especially for complex and multi-allelic CNVs... If this is not the case then the expected CNV data of the affected offspring (Xi) will differ from the midparent intensity, under the hypothesis of no association and for some pairs of parental genotypes... In this non-linear case, the contribution of each trio to the score statistic U does not have zero expectation conditional on the genotype of the parents... The result shown in 3 also shows that any population structure in the parental population that disrupts HWE also leads to a non-zero expectation for Ui in the non-linear case... Hence, the robustness to population structure does not extend from the linear to non-linear case... The trio design as developed in the FBAT CNV test can alleviate these concerns by comparing raw intensity data within families, rather than comparisons across cases and controls... In the non-linear intensity case, if the batch effect can be modelled using a family-specific covariate that acts on the raw intensity data in an additive manner, this covariate will have the same effect on the parental and offspring intensities... Therefore, the effect of this covariate will vanish in the computation of the score statistic... Thus, even in the case of a non-linear effect, the FBAT CNV statistic is expected to be robust to a range of technical artefacts... Approximately 250 of these 372 CNVs are deletions for which the intensity is typically non-linear with copy number... A small subset of CNVs showed significant association (Fig... In this case, the FBAT CNV test is appropriate and the initially proposed estimate of the score variance estimator V is consistent... Moreover, in a non-linear genotype-intensity context, the FBAT CNV test is robust to family-specific technical covariates.