Estimating Biological Age Using Ensemble-Based Prediction Models in Genomic Data

• Activity Number: 407
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Wendy Shih*+ and Steve Horvath and Roel Ophoff
• Companies: University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles
• Keywords: biomarkers of aging ; ensemble model ; genomics ; aging

Abstract

Biological age (BA) as opposed to chronological age (CA) is meant to measure the true aging process of an individual. The hope is that BA is better than CA when it comes to predicting mortality or age related functional decline. Here we evaluated several statistical methods (multiple regression, penalized regression, principal component analysis, factor analysis, ensemble based models, and epigenetic clock) for estimating BA based on biomarkers of age using simulated data that mirrors applications in DNA methylation and several genomic data applications. Further, we evaluated and adapted techniques from Klemera and Doubal (2006) and proposed a novel ensemble based approach in estimating biological age. Lastly, we evaluated and compared the performance of each method including our novel ensemble based model and found that our newly proposed ensemble approach performed relatively well compared to other methods.