Ligand based analysis of the solubility of drugs using in silico techniques

This is the authors' abstract. We don't add key points for this paper.

Solubility is one of the important physicochemical properties required for the pharmacological actions of drugs. The structural analysis of the existing drugs on the basis of their functional groups, atoms, bonds, physicochemical properties, etc. can provide important information on the factors responsible for the increased or decreased solubility of the drugs either in water or lipid. In the present investigation, machine learning based classification and quantitative structure property relationship (QSPR) analyses were performed with 118 drugs exhibiting experimental solubility data (in water and various concentrations of polyethylene glycol (PEG)-water mixture). Physicochemical descriptors and Molecular Access Systems (MACCS) Fingerprints were calculated using PaDEL software, and the machine learning-based classification studies and QSPR studies were conducted using the Waikato Environment for Knowledge Analysis (WEKA) software and BuildQSAR software respectively. Support vector machine (SVM), decision tree (DT), random forest (RF), random tree (RT), and Naïve Bayes (NB) algorithms were used in the machine learning study and multiple linear regression analysis was applied for QSPR study. The results revealed that polar and polarizable functional groups and the related physicochemical descriptors significantly contributed for the solubility prediction. Furthermore, drugs with more than one aromatic ring (MACCSFP125), side chains with nitrogen attached to aromatic atoms (MACCSFP135), triple bonds (MACCSFP19), the number of fluorine atoms (MACCSFP42), sulfur side chains attached to aromatic atoms (MACCSFP59), sulfur atoms connected between the ring and non-ring atoms (MACCSFP64), and the number of sulfur atoms were involved in the classification and prediction of the solubility. In all cases, the descriptors in the models explained that the presence of heteroatoms, polar, and polarizable groups highly influenced the aqueous solubility (with and without PEG) of the drugs. These studies concluded that the substitution of polar/polarizable functional groups in the drugs along with the position of groups and the distance between these polar/polarizable groups could improve the aqueous solubility of the drugs by forming hydrogen bonds with water molecules and the PEG.