A Practical Computational–Experimental Strategy for Drug–Polymer Phase Diagram Prediction and Validation

We present an integrated computational–experimental workflow for the efficient construction of binary drug–polymer phase diagrams. COSMO-RS (Conductor-like Screening Model for Real Solvents) predicts solid–liquid equilibrium (SLE) curves, enabling screening and prioritization of candidate drug–polymer pairs for subsequent experimental validation. Prediction accuracy depends strongly on the quality of the input parameters, particularly the fusion thermodynamics of the drug, including melting temperature (T_m), enthalpy of fusion (ΔH_fus), and, where known, the change in heat capacity on fusion (ΔC_{p, fus}). Accurate COSMO-RS calculations also require a representative molecular geometry, which is improved by identifying the relevant polymorph via powder X-ray diffraction and differential scanning calorimetry (DSC); experimental crystal structures (e.g., from the Cambridge Structural Database) provide a robust starting point for conformer generation. Experimental verification is performed using the DSC-based step-wise dissolution (S-WD) method, whereby a single supersaturated physical mixture is first used to determine the equilibration time at the lowest annealing temperature. The sample is then held at progressively higher temperatures, with the glass-transition temperature (T_g) measured after each step. Drug solubilities are obtained by mapping the measured T_g values onto a modeled composition–T_g relationship. Compared with the conventional melting point depression (MPD) approach, S-WD avoids the need for multiple sample compositions and mitigates ambiguities associated with onset/peak/offset T_m selection, as well as uncertainties in solubility equilibrium attainment. Together, COSMO-RS and S-WD streamline phase diagram construction, reduce experimental workload, and support the rational design of amorphous solid dispersions.