Predicting the conformational preferences of flexible compounds is a challenging problem in drug design, where the recognition between ligand and receptor is affected by the ability of the interacting partners to adopt a favorable conformation for the binding. To explore the conformational space of flexible ligands and to obtain the relative free energy of the conformation wells, we have recently reported a multilevel computational strategy that relies on the predominant-state approximation-where the conformational space is partitioned into distinct conformational wells-and combines a low-level method for sampling the conformational minima and high-level ab initio calculations for estimating their relative stability. In this study, we assess the performance of the multilevel strategy for predicting the conformational preferences of a series of structurally related phenylethylamines and streptomycin in aqueous solution. The charged nature of these compounds and the chemical complexity of streptomycin make them a challenging test for the multilevel approach. Furthermore, we explore the suitability of using a molecular mechanics approach as a source of approximate ensembles in the first stage of the multilevel strategy. The results support the reliability of the multilevel approach for obtaining an accurate conformational ensemble of small (bio)organic molecules in aqueous solution.