Stroke is a common complex trait and does not follow Mendelian pattern of inheritance. Gene-gene or gene-environment interactions may be responsible for the complex trait. How the interactions contribute to stroke is still under research. This study aimed to explore the association between gene-gene interactions and stroke in Chinese in a large case-control study. Nearly 4,000 participants were recruited from seven clinical centers. Eight variants in five candidate genes were examined for stroke risk. Gene-gene interactions were explored by using Generalized Multifactor Dimensionality Reduction (GMDR). A significant gene-gene interaction was found by GMDR. The best model including MTHFR C677T, ALOX5AP T2354A and NOTCH3 C381T scored 10 for Cross-Validation Consistency and 9 for Sign Test (P = 0.0107). The individuals with combination of MTHFR 677TT, ALOX5AP 2354AA and NOTCH3 381TT/TC had a significantly higher risk of thrombotic stroke (OR 3.165, 95% CI 1.461-6.858, P = 0.003). Our results show that combination of these alleles conferred higher risk for stroke than single risk allele. The gene-gene interaction may serve as a novel area for stroke research. The three-locus combination may change the susceptibility of particular subjects to the disease.