Description:

Aastha Khatiwada
PhD Candidate in Biostatistics
Medical University of South Carolina

Presentation Title:

GPA-Tree: Statistical Approach for Functional-Annotation-Tree- Guided Prioritization of  GWAS Results

Abstract:

Genome-wide association studies (GWAS) have successfully identified over two hundred thousand genotype-trait associations. However, deeper understanding of the genetic basis of complex traits remains challenging due to two major reasons. First, complex traits are often associated with many genetic variants, each with small or moderate effect sizes, which are hard to detect given the sample sizes in current GWAS. Second, our understanding of the functional mechanisms through which genetic variants are associated with complex traits remains limited. To address these challenges, we developed GPA-Tree, a statistical approach to integrate GWAS summary statistics and functional annotation information within a unified framework. GPA-Tree combines a machine learning algorithm (decision tree) and an Expectation-Maximization algorithm within a hierarchical modeling architecture and delivers biological insights by simultaneously identifying trait-risk-associated genetic variants and the key combinations of functional annotations that are enriched for the trait-risk-associated genetic variants. Simulation studies show that GPA-Tree outperforms existing statistical approaches in detecting risk-associated variants while also identifying the true combinations of functional annotations with high accuracy. Application of GPA-Tree to a systemic lupus erythematosus (SLE) GWAS and cell- and tissue-specific functional annotation data highlights the dysregulation of blood immune cells, including but not limited to primary B, memory helper T, regulatory T, neutrophils and CD8+ memory T cells in SLE. These results demonstrate that GPA-Tree can be a powerful tool that improves association mapping while facilitating understanding of the underlying genetic architecture of complex traits and potential mechanisms linking risk-associated variants with complex traits.