We developed and tested a novel, powerful composite experimental design and effective analytical tool for genetic dissection of complex traits regulated by polygenes with epistatic interactions. The design included two datasets: 1) an extended RIX cross consisting of 94 isogenic lines (13 CXB RI strains, 78 RIX diallel F1s generated from CXBs, both progenitor strains, BALB/cByJ, C57BL/6ByJ, and the nonreciprocal CXB F1); and 2) a complementary but independent CXB F2 population. Cerebellum weight, which is known to be under polygenic control, was used to test the analytic approach. The first dataset (extended CXB RIX) was used to solve dimensionality and linkage. Specifically, multiple regression analysis combined with a beam search procedure was used to identify a set of solutions derived from the optimal segregation model characterized by the minimal number of loci able to account for among-line variation relative to within–line environmental noise. Subsequently, solutions for the optimal segregation model whose model component SDPs were significantly linked to marker loci in the CXB RI set were selected for further characterization. These solutions were used to compare expected and observed histograms in the second dataset (CXB F2) using Monte Carlo method and G test. Solutions that successfully passed this test were used to predict model multilocus genotypes in the CXB F2 dataset and to perform multiple regression analysis of the joint datasets to estimate integral LRS. Finally we developed the composite genome-wide permutation test by using an “RIX shuffle” procedure and ordinary permutations of the F2 phenotype data. In conclusion, we extracted a single multilocus solution that exceeded the 95th percentile permutation threshold. This final solution contained four epistatically interacting loci, which included eight main and 24 pairwise epistatic genetic effects. Each locus was characterized by complete concordance with the corresponding marker locus in the CXB RI set. The approach based on two complementary but independent datasets together with the developed analytical tool demonstrates synergistic results and high effectiveness even for a small set of RI strains.
