Aim: To study the pro-arrhythmogenic effects of three gain-of-function mutations –KCNH2 T436M, KCNH2 T895M, KCNE3-V17M- related to atrial fibrillation (AF) by employing 3D human atrial models. Methods: The human atria were represented through 3D geometries composed of 21 regions and 56 subregions to reproduce the heterogeneous histological properties and fibre orientation. The electrophysiological properties were modeled using versions of the Courtemanche-Ramirez-Nattel human atrial model, including the parameters replicating the mutations' effects. A train of 10 stimuli at a 1 Hz frequency was applied to the sinoatrial node (SAN) to stabilize the electrical model. The arrhythmic behavior was induced by stimulating the coronary sinus (CS) region with a 5-pulses train with varying cycle lengths (CL) and by simultaneously maintaining the SAN activity. The temporal vulnerability was investigated for the three mutations. The vulnerable window (VW) was defined as the period when an ectopic stimulus in the CS region elicits arrhythmias. Results: The genetic mutations led to a substrate more susceptible to re-entrant activity. Different arrhythmic patterns were observed. When stimulated with a 160ms CL, the mutation KCNH2 T436M exhibited a VW of 10ms. In this time frame, 90% were macro re-entries, while the remaining 10% were stable rotors, lasting between 3.8s and 5s. In the case of the mutation KCNH2 T895M, pacing the CS region with a CL of 170ms, the VW was 7ms-wide with only macro re-entries perpetuating for the entire simulation time (5s). Finally, when exciting with a 90ms CL, the mutation KNCE3-V17M showed a 24ms-width VW; 79% of induced arrhythmias gave chaotic activations, wave breaks, and unstable rotors, the remaining 21% a stable rotor sustaining till the simulation end. Conclusions: The results support the fact that the three genetic mutations studied on 3D computational models lead to a pro-arrhythmic behavior with the generation of mutation-dependent patterns of arrhythmias.