Skip to main content

Table 5 pairwise comparison of predictions made by FAM and SVM for 369 GO functions classified into 4 groups based on how informative they are. A GO function is said to be more informative if fewer numbers of genes are associated with it. N = number of GO functions in each group. H, MH, ML, L denote the 4 groups with high, medium high, medium low, and low levels of informativeness. For each GO function, the performance of FAM and SVM are compared using averaged performance metrics over the 4 cross-validation folds. "Win," "Tie" and "Lose" refer to the number of GO functions for which FAM's predictions are better than, tied with, or worse than that of SVM, respectively. Notice that the last row is identical to that in Table 2.

From: A factor analysis model for functional genomics

   Sensitivity10% Specificity10% AUC
Group N Win Draw Lose Win Draw Lose Win Draw Lose
H 95 46 22 27 46 20 29 55 0 40
MH 144 64 12 68 76 12 56 74 0 70
ML 71 43 2 26 39 2 30 39 0 32
L 59 31 0 28 33 0 26 33 0 26
ACV 369 184 36 149 194 34 141 201 0 168