The presentation will discuss how virtual diffraction experiments and data analytics can be used simultaneously to derive numerically efficient models that can relate diffraction peak broadening to dislocation content in a deformed microstructure. Specifically, a new algorithm will be presented to generate virtual diffraction peaks from synthetic deformed microstructures generated from discrete dislocation dynamics simulations. This algorithm is then used to generate a database of synthetic whole diffraction line profiles. As a first application, virtual XRD line profiles from 221 DDD generated Al microstructures are processed into individual peaks and normalized by intensity and breadth. A principal component analysis is performed to capture peak shape variations. The reduced dataset is employed to train and test an ensemble of Gaussian process regression models. The ensemble method employed predicts dislocation densities with a coefficient of determination of 0.987 in validation. Second, this approach is applied to the case of plastically deformed Ta samples. For this application, the effect of instrumental broadening is necessarily taken into account. The 6 experimentally-gathered whole line profile diffraction profiles are then analyzed using the surrogate modeling framework. It is shown that the proposed approach yields realistic estimates of the dislocation content as a function of plastic deformation.