Analysis of SNP Data Through Sparse Principal Component Analysis with Altered Similarity Matrix

• Activity Number: 311
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ashley Bonner*+
• Keywords: Sparse PCA ; Loadings ; Tuning Parameters ; SNPs ; Similarity Matrix

Abstract

High-dimensional datasets have become standard in genomic epidemiology research. For example, we often obtain thousands or even millions of single nucleotide polymorphisms (SNPs) (p) from the human genome in hopes to determine which are causal to disease. With only a small number of study participants (n), statistical innovations are essential to our ability to process and learn from this abundance of data. Principal Component Analysis (PCA) is a multivariate method used for dimension reduction and data visualization, but struggles to handle high-dimensional data, especially when n < p. Improvements come via Sparse PCA, a new class of methods that are able to better-translate the underlying data structure through sparse principal components (PCs). However, it traditionally requires continuous-type data to perform appropriately. We utilize Sparse PCA to analyze SNP data and expose complex SNP profiles related to a disease outcome, adjusting the calculations to adapt to non-continuous structure.