Refining genetically inferred relationships using treelet covariance smoothing

摘要

Recent technological advances coupled with large sample sets have uncoveredmany factors underlying the genetic basis of traits and the predisposition tocomplex disease, but much is left to discover. A common thread to most geneticinvestigations is familial relationships. Close relatives can be identifiedfrom family records, and more distant relatives can be inferred from largepanels of genetic markers. Unfortunately these empirical estimates can benoisy, especially regarding distant relatives. We propose a new method fordenoising genetically - inferred relationship matrices by exploiting theunderlying structure due to hierarchical groupings of correlated individuals.The approach, which we call Treelet Covariance Smoothing, employs a multiscaledecomposition of covariance matrices to improve estimates of pairwiserelationships. On both simulated and real data, we show that smoothing leads tobetter estimates of the relatedness amongst distantly related individuals. Weillustrate our method with a large genome-wide association study and estimatethe "heritability" of body mass index quite accurately. Traditionallyheritability, defined as the fraction of the total trait variance attributableto additive genetic effects, is estimated from samples of closely relatedindividuals using random effects models. We show that by using smoothedrelationship matrices we can estimate heritability using population-basedsamples. Finally, while our methods have been developed for refining geneticrelationship matrices and improving estimates of heritability, they have muchbroader potential application in statistics. Most notably, forerror-in-variables random effects models and settings that requireregularization of matrices with block or hierarchical structure.