Mathematically Based General Framework for Integrating Multiple Heterogeneous Existing Data Sets Into a Novel Data Collection

• Activity Number: 418
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: IMS
• Author(s): Jozsef Bukszar*+ and Amit Khachane and Karolina Aberg and Youfang Liu and Joseph McClay and Patrick Sullivan and Edwin van den Oord
• Companies: Virginia Commonwealth University and Virginia Commonwealth University and Virginia Commonwealth University and The University of North Carolina at Chapel Hill and Virginia Commonwealth University and The University of North Carolina at Chapel Hill and Virginia Commonwealth University
• Keywords: data integration ; genetics ; heterogeneous data ; multiple-hypothesis testing ; applied mathematics

Abstract

We present a general framework that integrates information from multiple heterogeneous existing data sets (EDS) into a novel data collection (NDC) in order to find genetic units related to a certain disease. The framework is general in the sense that the EDS-s can be of any type whose genetic units (SNP-s, genes, chromosomal regions) can be ranked, where ties are allowed. The NDC may be of any type that has a statistic value assigned to each of its genetic unit, e.g. next-generation sequencing data or GWAS. The methods of the framework rely on exact mathematical formulas that ultimately provide the posterior probability that a genetic unit in the NDC has an effect. While the formulas rely on numerous unknown parameters, almost all of these unknown parameters can be aggregated into a single term, which, therefore, can be estimated precisely. The framework also includes some other tools, such as methods testing EDS for being informative to the NDC, and assessing the increase in power of adding another EDS. The methods have been validated through simulations, cross-validations on multiple schizophrenia GWAS, and an actual replication study.