Solubility is a frequently recurring issue within pharmaceutical industry, and new methods to proactively resolve this are of fundamental importance. Here, a novel methodology is reported for intrinsic solubility improvement, using in silico prediction of crystal structures, by perturbing key interactions in the crystalline solid state. The methodology was evaluated with a set of benzodiazepine molecules, using the two‐dimensional molecular structure as the only a priori input. The overall trend in intrinsic solubility was correctly predicted for the entire set of benzodiazepines molecules. The results also indicate that, in drug compound series where the melting point is relatively high (i.e., “brick dust” compounds), the reported methodology should be very suitable for identifying strategically important molecular substitutions to improve solubility. As such, this approach could be a useful predictive tool for rational compound design in the early stages of drug development.