Identifying quantitative trait loci by model selection allowing for epistasis

Dr. Ani Manichaikul
Department of Biostatistics
Johns Hopkins University



Estimating the number and locations of quantitative trait loci (QTL) and their interactions is a crucial step toward identifying genes responsible for variation in experimental crosses. We treat the problem as one of model selection and present a penalized likelihood approach. We extend the work of Broman and Speed (2002) to allow for pairwise interactions between QTL while still aiming to control false positives, an even greater concern with epistatic interactions in the model space. We formulate a conservative version of the penalized likelihood which provides strict control over the rate of extraneous QTL, as well as a more liberal approach aimed at detecting epistatic QTL with weak main effects. We demonstrate our model selection criteria as exploratory tools in analysis of backcross and intercross data. Through simulation studies, we show that reasonable power to detect true positives is also achieved.