Abstract
Background: Generalized Structured Component Analysis (GSCA) is a component-based alternative to traditional covariance-based structural equation modelling. This method has previously been applied to test for association between candidate genes and clinical phenotypes, contrasting with traditional genetic association analyses that adopt univariate testing of many individual single nucleotide polymorphisms (SNPs) with correction for multiple testing. Methods: We first evaluate the ability of the GSCA method to replicate two previous findings from a genetics association study of developmental language disorders. We then present the results of a simulation study to test the validity of the GSCA method under more restrictive data conditions, using smaller sample sizes and larger numbers of SNPs than have previously been investigated. Finally, we compare GSCA performance against univariate association analysis conducted using PLINK v1.9. Results: Results from simulations show that power to detect effects depends not just on sample size, but also on the ratio of SNPs with effect to number of SNPs tested within a gene. Inclusion of many SNPs in a model dilutes true effects. Conclusions: We propose that GSCA is a useful method for replication studies, when candidate SNPs have been identified, but should not be used for exploratory analysis.
Highlights
Generalized Structured Component Analysis (GSCA) is a component-based alternative to traditional covariance-based structural equation modelling
Part 1: Comparison of GSCA vs conventional association analysis with an existing dataset In part 1 of the study, we used a subset of data from the SLIC cohort to check first whether associations that had been found in the analysis of family data from the full cohort would be detectable in this subset of individuals despite the reduced statistical power and second, whether GSCA is less prone to identify spurious associations than more conventional univariate association analysis conducted using PLINK v1.9 (Purcell et al, 2007)
We included a simulation study for candidate gene association studies using the GSCA method to investigate how power and false-positive rates are affected by effect size, sample size and number of single nucleotide polymorphisms (SNPs)
Summary
Generalized Structured Component Analysis (GSCA) is a component-based alternative to traditional covariance-based structural equation modelling. This method has previously been applied to test for association between candidate genes and clinical phenotypes, contrasting with traditional genetic association analyses that adopt univariate testing of many individual single nucleotide polymorphisms (SNPs) with correction for multiple testing. Methods: We first evaluate the ability of the GSCA method to replicate two previous findings from a genetics association study of developmental language disorders. We present the results of a simulation study to test the validity of the GSCA method under more restrictive data conditions, using smaller sample sizes and larger numbers of SNPs than have previously been investigated. Conclusions: We propose that GSCA is a useful method for replication studies, when candidate SNPs have been identified, but should not be used for exploratory analysis
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
