Table 3 Mean test fold AUC and accuracy from 10-fold CV (standard errors in parentheses) for classification of colorectal cancer samples

No SDC

PoissonPCA + SCPCA

0.803 (0.025)

0.728 (0.022)

0.804 (0.027)

0.762 (0.022)

PLNPCA + SCPCA

0.752 (0.033)

0.723 (0.020)

0.751 (0.032)

0.714 (0.025)

PoissonPCA + FCPCA

0.801 (0.024)

0.739 (0.022)

0.809 (0.026)

0.757 (0.019)

PLNPCA + FCPCA

0.662 (0.036)

0.677 (0.032)

0.687 (0.042)

0.697 (0.024)

PoissonPCA + MSFA, q = 4

0.781 (0.027)

0.725 (0.027)

0.758 (0.028)

0.725 (0.027)

PLNPCA + MSFA, q = 2

0.748 (0.032)

0.674 (0.021)

0.750 (0.028)

0.705 (0.022)

PoissonPCA Alone

0.811 (0.021)

0.742 (0.022)

0.817 (0.021)

0.751 (0.028)

PLNPCA Alone

0.741 (0.018)

0.700 (0.025)

0.765 (0.021)

0.705 (0.019)

SDC

PoissonPCA + SCPCA

0.806 (0.026)

0.759 (0.023)

0.815 (0.026)

0.771 (0.021)

PLNPCA + SCPCA

0.773 (0.033)

0.748 (0.020)

0.790 (0.033)

0.751 (0.031)

PoissonPCA + FCPCA

0.802 (0.025)

0.754 (0.025)

0.822 (0.022)

0.762 (0.016)

PLNPCA + FCPCA

0.635 (0.028)

0.609 (0.021)

0.785 (0.031)

0.756 (0.025)

PoissonPCA + MSFA, q = 4

0.794 (0.027)

0.742 (0.018)

0.782 (0.030)

0.748 (0.019)

PLNPCA + MSFA, q = 3

0.669 (0.027)

0.637 (0.025)

0.701 (0.029)

0.696 (0.034)

PoissonPCA alone

0.800 (0.025)

0.742 (0.025)

0.822 (0.024)

0.745 (0.022)

PLNPCA alone

0.753 (0.019)

0.723 (0.023)

0.768 (0.020)

0.708 (0.017)

Naive

PCA of relative abundance

0.677 (0.015)

0.637 (0.023)

0.670 (0.023)

0.654 (0.024)

PCA of log relative abundance

0.674 (0.031)

0.638 (0.022)

0.686 (0.031)

0.646 (0.033)

PCA of counts

0.659 (0.013)

0.637 (0.023)

0.653 (0.020)

0.640 (0.018)

PCA of log-counts

0.793 (0.027)

0.719 (0.020)

0.798 (0.026)

0.756 (0.02)

AUC

Accuracy

Random forest (all features)

0.853 (0.019)

0.773 (0.012)