Cargando…
Using parental phenotypes in case-parent studies
In studies of case-parent triads, information is often collected about history of the condition in the parents, but typically parental phenotypes are ignored. Including that information in analyses may increase power to detect genetic association for autosomal variants. Our proposed approach uses pa...
Autores principales: | , , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
Frontiers Media S.A.
2015
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4477179/ https://www.ncbi.nlm.nih.gov/pubmed/26157456 http://dx.doi.org/10.3389/fgene.2015.00221 |
Sumario: | In studies of case-parent triads, information is often collected about history of the condition in the parents, but typically parental phenotypes are ignored. Including that information in analyses may increase power to detect genetic association for autosomal variants. Our proposed approach uses parental phenotypes to assess association independently of the usual case-parent-based association test, enabling cross-generational internal replication for findings based on offspring and their parents. Our model for parental phenotypes also resists bias due to population stratification. We combine the information from the two generations into a single coherent model that can exploit approximate equality of parental and offspring relative risks to improve power and can also test that equality. We call the resulting procedure the Parent-phenotype Informed Likelihood Ratio Test (PPI-LRT). When some parental genotypes are missing, one can use the expectation-maximization algorithm to fit the combined model. We also develop a second composite test (PPI-CT) based on a linear combination of the parent-phenotype-based test statistic and that from the traditional log-linear, transmission-based test. We evaluate the proposed methods through non-centrality parameter calculations and simulation studies and compare them to the previously proposed approaches, parenTDT and combTDT. We show that incorporation of parental phenotype data often improves statistical power. As illustration, we apply our method to a study of young-onset breast cancer and find that it improved precision for SNPs in FGFR2 and that estimated relative risks based on triads are closely replicated using the parental data. |
---|