Table 5 |
||||
Determining the contribution of gene weighting and moderated t-scores in PADOG when analyzing 229 KEGG metabolic and non-metabolic pathways | ||||
noM | noW | PADOG | noMnoW | |
p geomean | 0.0480 | 0.1330 | 0.0486 | 0.1225 |
p med | 0.092 | 0.1695 | 0.091 | 0.1595 |
% p.value<0.05 | 33.3 | 16.7 | 33.3 | 16.7 |
% q.value<0.05 | 8.3 | 0 | 4.2 | 0 |
rank mean | 20.52 | 22.33 | 18.95 | 22.48 |
rank med | 14.38 | 15.71 | 13.05 | 16.81 |
p Wilcox. | 0.0260 | 0.371 | 0.002 | reference |
p LME | 0.0463 | 0.314 | 0.0030 | reference |
coef. LME | -1.96 | -0.15 | -3.53 | reference |
^{}The table shows statistics computed from nominal and adjusted p-values, and ranks of the 24 target pathways only, including geometric mean, median and percentages of pathways significant at 0.05 level based on nominal and adjusted p-values (q-values). The results of comparing the ranks of each method against noMnoW method, using a paired Wilcoxon test and a linear mixed-effects model, are included. The best value for each criterion is shown in bold. PADOG is compared against simpler approaches that i) use gene weights but regular rather than moderated t-scores (noM), ii) use moderated t-scores but no gene weights (noW) and iii) use neither moderated t-scores nor gene weights (noMnoW).
Tarca et al.
Tarca et al. BMC Bioinformatics 2012 13:136 doi:10.1186/1471-2105-13-136