Knowledge of the melting properties of materials, especially at extreme pressure conditions, represents a long-standing scientific challenge. For instance, there is currently considerable uncertainty over the melting temperatures of the high-pressure mantle mineral, bridgmanite (MgSiO3-perovskite), with current estimates of the melting T at the base of the mantle ranging from 4800 K to 8000 K. The difficulty with experimentally measuring high pressure melting temperatures has motivated the use of ab initio methods, however, melting is a complex multi-scale phenomenon and the timescale for melting can be prohibitively long. Here we show that a combination of empirical and ab-initio molecular dynamics calculations can be used to successfully predict the melting point of multicomponent systems, such as MgSiO3 perovskite. We predict the correct low-pressure melting T, and at high-pressure we show that the melting temperature is only 5000 K at 120 GPa, a value lower than nearly all previous estimates. In addition, we believe that this strategy is of general applicability and therefore suitable for any system under physical conditions where simpler models fail.