基于贝叶斯多表型模型提高 SNP 遗传力的估计。
Improved estimation of SNP heritability using Bayesian multiple-phenotype models.
机构信息
Department of Medical and Molecular Genetics, King's College London, London, England.
Department of Statistics, London School of Economics and Political Science, London, England.
出版信息
Eur J Hum Genet. 2018 May;26(5):723-734. doi: 10.1038/s41431-018-0100-z. Epub 2018 Feb 13.
Linear mixed models (LMM) are widely used to estimate narrow sense heritability explained by tagged single-nucleotide polymorphisms (SNPs). However, those estimates are valid only if large sample sizes are used. We propose a Bayesian covariance component model (BCCM) that takes into account the genetic correlation among phenotypes and genetic correlation among individuals. The use of the BCCM allows us to circumvent issues related to small sample sizes, including overfitting and boundary estimates. Using expression of genes in breast cancer pathway, obtained from the Multiple Tissue Human Expression Resource (MuTHER) project, we demonstrate a significant improvement in the accuracy of SNP-based heritability estimates over univariate and likelihood-based methods. According to the BCCM, except CHURC1 (h = 0.27, credible interval = (0.2, 0.36)), all tested genes have trivial heritability estimates, thus explaining why recent progress in their eQTL identification has been limited.
线性混合模型（LMM）广泛用于估计由标记单核苷酸多态性（SNP）解释的狭义遗传率。然而，只有使用大样本量时，这些估计才有效。我们提出了一种贝叶斯协方差分量模型（BCCM），该模型考虑了表型之间的遗传相关性和个体之间的遗传相关性。使用 BCCM 可以避免与小样本量相关的问题，包括过度拟合和边界估计。使用来自多组织人类表达资源（MuTHER）项目的乳腺癌途径中的基因表达数据，我们证明了基于 SNP 的遗传率估计在准确性上优于单变量和似然方法。根据 BCCM，除了 CHURC1（h=0.27，置信区间=（0.2，0.36））之外，所有测试的基因都具有微不足道的遗传率估计，这解释了为什么最近在它们的 eQTL 鉴定方面进展有限。
Zotero 插件