Aggregate particles strongly influence the behaviour and performance of asphalt mixtures. This study applies a refined version of a recently developed microstructure generator (MG) to computationally produce random aggregate particles with controlled shape parameters (i.e. angularity index [AI] and form index [FI]). These sets of aggregates were used to generate virtual microstructures of asphalt mixtures that were transferred to finite elements and subjected to mechanical loading. A main advantage of using the MG to produce several random microstructures of asphalt mixtures is that it permits to capture the heterogeneity inherent to the material and, therefore, to include the uncertainty related to its mechanical response. The results showed that virtual specimens with aggregates having low values of AI (at a constant value of FI) and high values of FI (at a constant value of AI) presented lower damaged area and smaller overall modulus of the mixture.