Differential Bias Introduced by the Copy Number Variation Inheritance Model

• Activity Number: 179
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Sulgi Kim*+ and Ellen M. Wijsman and Debby W. Tsuang
• Companies: University of Washington and University of Washington and University of Washington
• Keywords: CNV ; Genetics ; Family data

Abstract

Copy Number Variation (CNV) calling for SNP chip data is challenging. Wang et al. (2008) proposed a Hidden Markov Model for family data that jointly models all members of a trio (parents and their offspring) using Mendelian Inheritance Laws, with distribution of an implementation in the package, "PennCNV Joint-calling". Conceptually, this formal model is expected to improve CNV-detection in related samples because it uses more information, which the authors showed with simulated datasets. We show that CNVs inferred by this inheritance model have unequal marginal sensitivity between a parent and an offspring thus introducing a differential bias. Therefore, a typical family study where the phenotype is correlated with the family membership (e.g. the case-parent design) violates the assumption that case and control should have errors equally. This results in an inflated CNV detection rate was inflated in the offspring in our real data. We show the bias of the model theoretically, with simulation, and then with two real datasets.