Ultimately, seismotectonic studies seek to provide ways of assessing the timing, magnitude and spatial extent of future earthquakes. Ample observations document the spatial variability in interseismic coupling, defined as a degree of locking of a fault during the period of stress build-up between seismic events: fully or nearly locked fault patches are often surrounded by aseismically creeping areas. However, it is unclear how these observations could help assess future earthquakes. Here we simulate spontaneous seismic and aseismic fault slip with a fully dynamic numerical model. Our simulations establish the dependence of earthquake rupture patterns and interseismic coupling on spatial variations of fault friction. We consider the long-term evolution of slip on a model fault where two seismogenic, locked segments are separated by an aseismically slipping patch where rupture is impeded. We find that the probability for a large earthquake to break through the rupture-impeding patch is correlated with the interseismic coupling averaged over this patch. In addition, the probability that an earthquake breaks through the rupture-impeding patch and interseismic coupling are both related to fault friction properties through a single non-dimensional parameter. Our study opens the possibility of predicting seismic rupture patterns that a fault system can produce on the basis of observations of its interseismic coupling, and suggests that regions of low interseismic coupling may reveal permanent barriers to large earthquakes.