In this work we propose and study four novel search procedures for identifying susceptibility genes for complex human diseases among thousands of candidate genes genome-wide and, in particular gene-gene interaction. The proposed methods are based on a float search algorithm utilizing criterion functions based on entropy based measures of association. The search procedures utilize genotype data from a large set of independent markers under case-control design. We show, through extensive simulations and comparisons, that the proposed search procedure may have higher discovery rates than other methods for identification of gene-gene interaction currently used. |