If the proportion of T2D variance that physical activity explains independently of genes was known to be only 3%, say, then the integrated predictor's maximum AUC would be calculated based on a proportion of variance explained of 29% (sum of 26% and 3%), which yields a maximum AUC of 90%.
These principal components were then summed, weighted according to the relative proportion of variance explained, to compute the continuous metabolic syndrome score.
They then multiplied each component by factor loading and weighed the sum by the relative contribution in the proportion of variance explained by each factor.
Then, to establish the shared variance among the four predictors, we subtracted the sum of all squared semi-partial correlations 2 from the proportion of variance explained by the four predictors, R²(.18) – total sr(.06) =.06.06
Finally, to establish the shared variance among the three predictors, we subtracted the sum of all squared semi-partial correlations 4 from the proportion of variance explained by the three predictors, R²(.28) – total sr(.11) =.11.11
Note that Epost(%PVE) are estimates for single markers and simply summing them up does not yield an estimate for the entire proportion of variance explained by markers, as covariances due to LD between markers are missed in this sum.
Table 1 shows the composition and the percent of variance explained (the eigenvalue for that component divided by the sum of all eigenvalues) for the first two components.
The fraction of variance explained by tissues and species was calculated by computing the fractions of sum of squares explained by the factor tissue and the factor species in a linear model.
For example, after merging windows at 75 and 76 Mb on chromosome 17 in Shetland pony mares and performing another GWAS, the percentage of variance explained by this 2 Mb genomic region was 0.426, which roughly equals the sum of genetic variance explained by the two separate 1 Mb windows (Table 4).
In addition, the overall percentage of variance explained by life-course area was calculated from the cross-classified models by summing area variance for all 3 study years and dividing by the total variance.
