Each exported dataset was subject to an interval mapping analysis, which uses GeneNetwork's embedded MapManager software (Manly et al. 2001) to perform Haley–Knott regression. Empirical P-values were derived using 1000 permutations using the incorporated permutation feature of WebQTL. The peak of each statistically significant (P-value < 0.05) or suggestive (P-value < 0.63) (Lander & Kruglyak 1995) QTL was determined based on empirical P-values (Doerge & Churchill 1996). A one-LOD drop-off was used to determine the QTL confidence interval about each peak. Positional candidates residing within an one-LOD drop from the peak of each statistically significant and suggestive QTL were identified. Trait data were correlated against the following GeneNetwork gene expression tissue databases: whole brain [INIA Brain mRNA 430 (June 2006) RMA (Peirce et al. 2006)], neocortex [HQF BXD NeoCortex ILM6v1.1 (February 2008) RankInv (Gaglani et al. 2009)], striatum [HBP Rosen Striatum M430V2 (April 2005) RMA (Rosen et al. 2009)], cerebellum [SJUT Cerebellum mRNA M430 (March 2005) RMA (Chesler et al. 2005)] and hippocampus [Hippocampus Consortium M430V2 (June 2006) RMA (Kempermann et al. 2006)]. Lists of genes were generated based on their correlation to each phenotype using a correlation P-value <0.001.
