Most associated linear combinations of phenotypes at genome-wide significant SNPs.

*indicates that the SNP did not have a univariate genome-wide significant P value. Each row indicates the linear combination of phenotypes (given by the corresponding regression coefficients) which is most associated with the given SNP under the MultiPhen regression, after removing the most associated phenotype. The regression coefficients have been scaled so that the CHOL coefficient is always equal to one. The last row contains the expected coefficients according to the Friedewald Formula (Equation 1).