*2.2. Energy Framework Analysis*

Energy frameworks were constructed in *CrystalExplorer* using the CE-B3LYP model (basis set: 6-31G(d,p)), taking into account only intermolecular interactions within the radius of 3.8 Å from the centroid on the central molecule [30]. All frameworks were scaled equally so as to facilitate comparisons.

### *2.3. Polymorph Assessment through Hydrogen Bond Propensity Models*

Initially, the definition of a hydrogen bond was set as established earlier; however, this resulted in an excessive number of contacts within each form, some of which were chemically incorrect. Therefore, the bond angle was modified to be larger than 133◦ (the lower limit of the hydrogen bonds detected in both forms). All hydrogen bond acceptors and donors were selected, with the latter list including carbon. Systems with errors or disorder were excluded from the study, together with organometallic compounds and entries with an R-factor > 0.075.

A set of data containing entries whose chemistry was relevant to ganciclovir was generated so as to build the statistical model for prediction. Potential bias or error was decreased by ensuring that each functional group (5 substructures; Please refer to Supplementary Materials) was represented by an adequate number of hits. After analysis, donor or acceptor candidates with a low number of relevant hits were omitted to avoid regression failures. A logistic regression model was fitted on the data, producing its corresponding area under the receiver operating characteristic (ROC) curve value as a measure of the extent of correct predictions. This procedure was performed separately for both polymorphs.

### **3. Results and Discussion**
