Today over 350 million patients worldwide are affected with orphan diseases. To tackle the associated challenges, in silico models and virtual clinical trials are increasingly explored. In this study we combined mechanistic modeling with data-driven modeling in an investigative in silico clinical trial to assess the (beneficial) effect of bone morphogenetic protein (BMP) treatment on fracture healing in patients with congenital pseudarthrosis of the tibia (CPT). Although the exact etiology of CPT is still highly debated, it is hypothesized that a mutation in the Neurofibromatosis type 1 (NF1) gene results in an altered phenotype of the skeletal cells and impaired bone healing. In this study, we generated a set of 200 virtual patients from a previously established multiscale model of bone regeneration by altering the parameter values of eight key factors which describe the aberrant cellular behaviour of cells affected by NF1 mutation. Each virtual patient was simulated to receive no treatment and BMP treatment. We show that the degree of severity of CPT is significantly reduced with BMP treatment, although the effect is highly patient-specific. Moreover, machine learning techniques identified four distinct patient groups: adverse responders, non-responders, responders and asymptomatic. This study demonstrates how mechanistic and data-driven modeling are useful tools to simulate and mine data from in silico clinical trials, stratify patient populations, and improve current treatment strategies.