Ballasted tracks play an essential role in its economy through transporting freight and bulk commodities between major cities and ports, and carrying passengers, particularly in urban areas. Ballast usually consists of hard and strong angular particles, which are derived from high strength unweathered rocks. Ballast aggregates undergo gradual and continuing degradation under cyclic train loadings. In this study, the load-deformation responses of ballasted rail tracks subjected to cyclic loading are studied experimentally using a large-scale Track Process Simulation Apparatus (TPSA), and numerically through a coupled discrete-continuum approach, namely, coupled DEM-FEM. Laboratory tests are carried out to examine the deformation and degradation responses of ballast subjected to cyclic train loading under a frequency of f=15 Hz and a lateral confinement of σxx=10 kPa. Test results reveal that significant settlements are observed during the initial load cycles, followed by gradually increased deformation, arriving at a steady value towards the end of testing. A rigorous coupling model based on discrete element method (DEM) and finite element method (FEM) is introduced to predict the load-deformation behaviour of the ballast assembly considering the interaction of discrete ballast grains and continuum subgrade. In this coupled model, the discrete ballast grains are modelled by DEM and the subgrade domain is modelled as a continuum by FEM. Interface elements are introduced to transmit the interacting forces and displacements between adjoining material domains (i.e. discrete and continuum) whereby the DEM transfers contact forces to the FEM, and then the FEM updates the displacements back to the DEM. The coupled model is validated by comparing the predicted ballast settlement responses with those obtained experimentally. Contact force distributions, stress contours and a corresponding number of broken bonds are analysed. This combined DEM-FEM model is also used to analyse the load-deformation of a fully instrumented track in Singleton, Australia, and the numerical predictions are compared with the field data.