We found that our PrimerROC/PrimerDimer software consistently outperforms these other tools and can achieve predictive accuracies greater than 92%.
The classifier for testis-selective gene prediction reached predictive performance with overall accuracy at 91.00% and ROC AUC = 0.8433.
The most accurate classifier built using only MAPS variables reaches a predictive accuracy of 92.86% using a total of 17 variables: the individual times taken to complete each Auditory Discrimination test, the total number of correct responses to the Auditory Memory test, and the individual times taken to complete each Sequencing test.
Finally, with a sensitivity of 87% and a specificity of 63%, our tool reaches a reasonable predictive accuracy, while other published tools were similarly specific but often less sensitive and most of the time required complicated data collection falling outside routine care [ 5, 7, 26].
However, the correlation between marbling score EPD and their predicted values in the testing set (referred to hereafter as predictive accuracy) reached its peak (0.863) with a panel size of 700 SNPs, and decreased thereafter.
However, the combined effect of biodiversity components related to the functional structure of communities used in our study consistently reached unprecedented levels of predictive accuracy (up to 84%) whatever the process and for all processes together.
Moreover, we also noticed that, compared with the 96% overall accuracy of data mining method in prediction of NSCLC prognosis and the 88.4% accuracy of gene profiling in predicting RLNM in colorectal cancer (Takahashi et al, 2007; Watanabe et al, 2009), our SVM model reached a lower (72.3%) predictive accuracy in refining RLNM status for RC patients.
Physical models can achieve high predictive accuracy if appropriately built.
However, the correlation between the estimated breeding values and their predicted values in the testing set (referred to as predictive accuracy hereafter) reached its peak (0.8886) with a panel size of 1000 SNP, and then went down slightly.
They become comparable in terms of predictive accuracy only when the number of features reaches more than a few hundred (we do not report these results due to the space limitations).
The reason that the difference in predictive accuracy between the mMIC and the MIV did not reach significance likely resulted from the fact that there was abundant discriminative information in category-dependent patterns.
