OPLS model quality was evaluated based on the parameters R Y (goodness of fitting, i.e., fraction of Y variation modelled by the predictive component) and Q Y (goodness of prediction) as determined through the default leave-1/7th-out cross validation test.
In the present study, accuracy of prediction as determined by adjusted R-square values was between 30.8% and 4.1%.
As shown in Author response image 1, the performances as judged by ROC analyses were found similar whether TP53 mutation was integrated into the prediction (as determined either by all exon sequencing [blue curve] or only exon 5-8 sequencing [green curve]) or not (red curve).
Regardless of the estrogen or the regression used, the strength of the prediction model, as determined by the coefficient of determination (r2) and other factors, was always better for Kd than Koc values.
The goodness of fit of the model was evaluated in terms of both R (which is analogous to the Pearson product moment correlation coefficient) and the goodness of prediction Q as determined by a crossvalidation procedure [Vandervoet, 1994; Wold, 1978].
bNumber of transmembrane domains in the predicted protein, as determined by InterPro prediction software (http://www.ebi.ac.uk/interpro/).ac.uk/interpro/
The discriminatory ability of these complement proteins (AUC = 0.816) appears superior or at least similar to the discriminatory ability of genetic markers of complement genes for the prediction of AMD as determined in this and other similarly designed studies [24].
The final model will contain simulation results for the number of poses (looped over using index j) for which the standard deviation error in prediction (SDEP) is lowest as determined in leave-one-out (LOO) cross validation.
The analysis identified a signature of 23 deregulated miRNAs that best discriminate the two groups with a prediction accuracy of 96% as determined by 10-fold cross-validation (Supplementary Figure S1).
As discussed in the Methods section, when comparing microRNA target predictions to actual microRNA targets (as determined by microarray experiments) two filtering steps were performed to increase the suitability of the target predictions for the data – filtering by both microarray and by cell type.
