Abstract Motivation: Canonical correlation analysis (CCA) measures the association between two sets of multidimensional variables. We reasoned that CCA could provide an efficient and powerful approach for both univariate and multivariate gene-based tests of association without the need for permutation testing. Results: Compared with a commonly used permutation-based approach, CCA (i) is faster; (ii) has appropriate type-I error rate for normally distributed quantitative traits; (iii) provides comparable power for small to medium-sized genes (<100 kb); (iv) provides greater power when the causal variants are uncommon; (v) provides considerably less power for larger genes (≥100 kb) when the causal variants have a broad minor allele frequency (MAF) spectrum. Application to a GWAS of leukocyte levels identified SAFB and a histone gene cluster as novel putative loci harboring multiple independent variants regulating lymphocyte and neutrophil counts. Availability: http://genepi.qimr.edu.au/staff/manuelF/gene/main.html Contact: [email protected] Supplementary information: Supplementary material is available at Bioinformatics online.

Abstract Summary: Although genetic association studies often test multiple, related phenotypes, few formal multivariate tests of association are available. We describe a test of association that can be efficiently applied to large population-based designs. Availability: A C++ implementation can be obtained from the authors. Contact: [email protected] Supplementary information: Supplementary figures are available at Bioinformatics online.