TDT for Human QTL Mapping and Genome - Wide Association Study

Abstract

An efficient and economical sampling approach ERS is proposed to extend TDT to QTL mapping, and the effect on the power of various TDT is studied. The simulation study is carried out to compare the power of TDT with ERS and with conventional truncation approach. We also introduce a generalized TDT approach by penalized logistic model and a new variable selection criterion EBIC to apply TDT in genome-wide association study. The validation and the advantages of this approach are demonstrated by comparison of PSR and FDR with multiple-comparison method. Our numerical studies also illustrate that compared to EBIC, the traditional variable selection criterion BIC tends to select too many spurious variables in high dimensional space. Handling the data in a certain way, we have a logistic model with grouped covariates. An efficient algorithm for sparse solution is proposed based on a series of optimality conditions of the optimization programming.