Skip to main content
. 2012 Sep 3;28(18):i582–i588. doi: 10.1093/bioinformatics/bts415

Table 5.

Top three models identified by Cox-MDR with main effect and without main effect

With all 139 SNPs
With 118 SNPs after removing 21 SNPs having strong main effect
Models TRBA TSBA Coeff. P P* Models TRBA TSBA Coeff. P P*
One-way One-way
TYMS rs1004474 0.665 0.665 0.156 0.487 0.00037 TYMS rs1004474 0.665 0.665 0.156 0.487 0.00037
TYMS rs2847153 0.633 0.633 0.206 0.324 0.00194 TYMS rs2847153 0.633 0.633 0.206 0.324 0.00194
CDA rs10799647 0.629 0.630 −0.723 0.076 0.00001 CDA rs10799647 0.629 0.630 −0.723 0.076 0.00001

Two-way Two-way
CDA rs12404655 and 0.719 0.712 −1.125 0.098 0.00001 CDA rs12404655 and 0.719 0.712 −1.125 0.098 0.00001
TYMS rs1004474 TYMS rs1004474
CDA rs532545 and 0.705 0.704 −0.216 0.591 0.00973 CDA rs532545 and 0.705 0.704 −0.216 0.591 0.03013
TYMS rs2847153 TYMS rs2847153
CDA rs10916824 and 0.713 0.704 −1.211 0.117 0.00138 MTHER rs9651118 and 0.721 0.703 −0.714 0.038 0.00035
TYMS rs1004474 TYMS rs1004474

TRBA: Training balanced accuracy; TSBA: Testing balanced accuracy; Coeff.: the estimated effect size of the corresponding SNP effect; P: P-value of main and two-way interactions in the Cox regression model; P*: permutation P-value of main and two-way interaction effects.