Abstract:
|
The prognostic value attributed to the interaction between multiple genetic variants and environmental factors has not been well studied for complex diseases such as cancer. While the next generation sequencing technologies and imputing current GWAS data to larger reference panels (e.g. the 1000 genomes project) provide two efficient ways to discover additional variants, statistical approaches for linking the high-dimensional genetic variants to disease progression and survival are not well-developed. We propose utilizing novel statistical approaches to investigate the joint contributions of multiple imputed or sequenced genetic variants to cancer prognosis, along with how the genetic variant profile effects are modified by other environmental factors. Specifically, we develop a functional proportional hazards model to incorporate data dimension reduction and to model the integrative effects of wide spectrum of genetic variant profiles and environmental factors. We demonstrate the performance of our methods by extensive simulations and application to large genetic epidemiological cancer studies with well-characterized genotype and clinical data.
|