Establishment of an <i>in silico</i> phospholipidosis prediction method using descriptors related to molecular interactions causing phospholipid–compound complex formation

Abstract

-Although phospholipidosis (PLD) often affects drug development, there is no convenient in vitro or in vivo test system for PLD detection. In this study, we developed an in silico PLD prediction method based on the PLD-inducing mechanism. We focused on phospholipid (PL)-compound complex formation, which inhibits PL degradation by phospholipase. Thus, we used some molecular interactions, such as electrostatic interactions, hydrophobic interactions, and intermolecular forces, between PL and compounds as descriptors. First, we performed descriptor screening for intermolecular force and then developed a new in silico PLD prediction using descriptors related to molecular interactions. Based on the screening, we identified molecular refraction (MR) as a descriptor of intermolecular force. It is known that ClogP and most-basic pKa can be used for PLD prediction. Thereby, we developed an in silico prediction method using ClogP, most-basic pKa, and MR, which were related to hydrophobic interactions, electrostatic interactions, and intermolecular forces. In addition, a resampling method was used to determine the cut-off values for each descriptor. We obtained good results for 77 compounds as follows: sensitivity = 95.8%, specificity = 75.9%, and concordance = 88.3%. Although there is a concern regarding false-negative compounds for pKa calculations, this predictive ability will be adequate for PLD screening.In conclusion, the mechanism-based in silico PLD prediction method provided good prediction ability, and this method will be useful for evaluating the potential of drugs to cause PLD, particularly in the early stage of drug development, because this method only requires knowledge of the chemical structure.

Artificial Intelligence and Machine Learning in Computational Nanotoxicology: Unlocking and Empowering Nanomedicine

“…[ 78 ] For example, QT prolongation, [ 79 ] hepatotoxicity, [ 80 ] and phospholipidosis. [ 81 ] Such predictive ADMET models are commercially available and can be used in early stages of drug discovery. [ 82 ] However, the models are still only at early stage as the data they are based on are limited.…”

A reaction mechanism-based prediction of mutagenicity: α-halo carbonyl compounds adduct with DNA by S_N2 reaction

“…Therefore, we set EG of 31 ± 2 eV as a "gray zone." The gray zone was defined as co-existence of positive and negative compounds at similar rates (Haranosono et al, 2014(Haranosono et al, , 2016. In a similar way, Fig.…”

“…To predict mutagenicity, we used a "scoring" method (Haranosono et al, 2016). Each compound was given a "score" for each descriptor as follows: positive zone = +1, negative zone = −1, and gray zone = 0.…”