a1 Biostatistics Graduate Group, University of California, Davis, CA 95616, USA
a2 Gladstone Institutes and Department of Epidemiology and Biostatistics, University of California, San Francisco, CA 94158, USA
Increasingly researchers are conducting quantitative trait locus (QTL) mapping in metabolomics and proteomics studies. These data often are distributed as a point-mass mixture, consisting of a spike at zero in combination with continuous non-negative measurements. Composite interval mapping (CIM) is a common method used to map QTL that has been developed only for normally distributed or binary data. Here we propose a two-part CIM method for identifying QTLs when the phenotype is distributed as a point-mass mixture. We compare our new method with existing normal and binary CIM methods through an analysis of metabolomics data from Arabidopsis thaliana. We then conduct a simulation study to further understand the power and error rate of our two-part CIM method relative to normal and binary CIM methods. Our results show that the two-part CIM has greater power and a lower false positive rate than the other methods when a continuous phenotype is measured with many zero observations.
(Received August 17 2009)
(Revised November 30 2009)
(Online publication March 03 2010)