Introduction: Multiple genetic variants have been identified in GWAS as being associated with ischemic cardiopathy (IC). To better analyze the gene-gene interaction, new computational and statistical methods emerged beyond logistic regression. Objectives: Study the best gene-gene interaction model and predictor of coronary disease (CAD), using new data mining methods such as Multifactor Dimensionality Redution (MDR). Methods: We included 2888 participants (mean age 53±7.9 years, 77.8% male), namely 1566 coronary patients documented by angiography with one or more epicardial stenoses greater than 75% and 1322 controls adjusted for age and gender. Taqman SNP genotyping (Applied Biosytems) was used and then a gene-to-gene analysis was performed between 33 variants associated with CAD. MDR was applied to obtain the best genetic predictor for IC, by using the 10 most significant variants. Results: In the one-gene model, the MDR projected the TCF21 gene as the most significant genetic risk factor for CAD. The model with two genes demonstrated synergistic interaction between the TCF21 and APOE. The genetic bivariate model of TCF21 and APOE was the best predictive model with an OR of 1.51 (95% CI: 1.29–1.76; p<0.0001) and with adequate cross-validation (10/10), with no evidence of overfitting model. The acuity of the best G-G predictor model of CAD was 0.55. A reasonable sensitivity (60%) and specificity (50%) were obtained from this model.