Prediction of Progesterone Solubility by Modern Quasi-Chemical and Classical Group Contribution Models

Progesterone is a hydrophobic steroid hormone that plays
a crucial
role in human health. Its limited aqueous solubility presents a significant
challenge for both pharmaceutical applications and academic research,
showing the importance of accurately predicting its solubility behavior
in different solvents. This study assesses the performance of the
modern quasi-chemical equation of state known as COSMO-SAC-Phi (CSP),
in comparison with its underlying COSMO-SAC (CS) activity coefficient
model when modeling solid–liquid equilibrium of progesterone
in 14 different solvents. Pure compound parameters employed in CSP
calculations were obtained from the vapor pressure and liquid volume
data of each pure compound. No binary parameters were adjusted. The
results were compared with experimental solid–liquid equilibrium
data collected from the literature. As a reference, the classical
UNIFAC (Do) group contribution method was also used. The CSP model
generally provided more accurate predictions of phase equilibrium,
captured solubility trends among similar solvents, and reproduced
the correct deviations from ideality for most systems, whereas the
CS model was often less accurate in these aspects. An intermediate
performance was observed for UNIFAC (Do). The mean absolute deviation
in log10 units from experimental solubility data highlights
the advantage of the equation-of-state approach, yielding an average
value of 0.26 for CSP compared to 0.60 for the underlying activity
model and 0.37 for UNIFAC (Do). Nevertheless, CS should be sufficiently
accurate for the preliminary screening of new solvent alternatives.