Ischemic Stroke (IS) is a severe and complex disorder of high morbidity and mortality rates associated with clinical, environmental, and genetic predisposing factors. Despite previous studies have associated genetic variants to stroke, inconsistent results from different populations pointed to the genetic heterogeneity for IS. Therefore, we may hypothesize that an interaction effect among genetic variants could contribute to IS occurrence rather than genetic variants independently. In this context, we investigated the association and interaction between genetic variants and large-artery atherosclerosis IS (LAAS-IS) and cardioembolic IS (CE-IS). We genotyped 435 patients (195 LAAS-IS; 240 CE-IS) and 535 controls from a population of Joinville, Santa Catarina, Brazil. Association and interaction analysis were performed by chi-square test and Multifactor-dimensionality Reduction test. We found an association between rs2383207*A allele, nearby CDKN2B-AS1, and LAAS-IS [OR 2.35 (95% CI = 1.79-3.08); p = 4.66 × 10-10]. We found an interaction among rs2910829, rs966221 and rs152312, with an accuracy of 0.62 (p = 4.3 × 10-5) demonstrating the interaction effect among variants from different genes can contribute to CE-IS risk. Further prediction analysis confirmed that clinical information, such as hypertension and dyslipidemia, presented high accuracy to predict LAAS-IS (86.47%) and CE-IS (90.47%); however, the inclusion of genetic variant information did not increase the accuracy.