Table 3. Best predictive gene-gene interaction models identified by multifactor dimensionality reduction analysis.

Model No.	Best Predictive Interaction Modela	Prediction Error (1-TBAb)	10,000 Permutations P-value	c CVC (in %)
I	TLR4Thr399Ile/IL12B+1188	0.318	0.024*	95
II	TLR4Thr399Ile/LTA80	0.358	0.033*	100
III	IL12B+1188/TNF-1031	0.341	0.013*	100
IV	IL12B+1188/TLR9P545P	0.329	0.005*	100
V	IL12B+1188/LTA80	0.366	0.042*	100
VI	TLR4Thr399Ile/IL12B+1188/LTA80	0.305	0.031*	100

^a The best model was selected as the one with the minimum prediction error and maximum CVC.

^b TBA corresponds to the testing balanced accuracy defined as the prediction error (PE) = 1-TBA.

^c CVC corresponds to cross validation consistency.

^* P values for gene-gene interaction models were calculated after 10,000 permutations in MDRpt software.