QSAR Development for Plasma Protein Binding: Influence of the Ionization State

Abstract

PurposeThis study explored several strategies to improve the performance of literature QSAR models for plasma protein binding (PPB), such as a suitable endpoint transformation, a correct representation of chemicals, more consistency in the dataset, and a reliable definition of the applicability domain.MethodsWe retrieved human fraction unbound (Fu) data for 670 compounds from the literature and carefully checked them for consistency. Descriptors were calculated taking account of the ionization state of molecules at physiological pH (7.4), in order to better estimate the affinity of molecules to blood proteins. We used different algorithms and chemical descriptors to explore the most suitable strategy for modeling the endpoint. SMILES (simplified molecular input line entry system)-based string descriptors were also tested with the CORAL software (CORelation And Logic). We did an outlier analysis to establish the models to use (or not to use) in case of well recognized families.ResultsInternal validation of the selected models returned Q2 values close to 0.60. External validation also gave r2 values always greater than 0.60. The CORAL descriptor based model for √fu was the best, with r2 0.74 in external validation.ConclusionsPerformance in prediction confirmed the robustness of all the derived models and their suitability for real-life purposes, i.e. screening chemicals for their ADMET profiling. Optimization of descriptors can be useful in order to obtain the correct results with a ionized molecule.Electronic supplementary materialThe online version of this article (10.1007/s11095-018-2561-8) contains supplementary material, which is available to authorized users.

Setting the stage for next-generation risk assessment with non-animal approaches: the EU-ToxRisk project experience

“…The toolbox was further enriched by novel in silico approaches, supporting and complementing in vitro cell-based NAMs. An important area here was the extrapolation of NAM concentrations to in vivo doses (IVIVE, PBPK) (Fisher et al 2019 ; Simeon et al 2020 ; Toma et al 2018 ). Additionally, novel QSAR tools and machine learning approaches were put in place to assess chemical similarity among test compounds (Gadaleta et al 2018a , b ; Hemmerich et al 2020 ; Luechtefeld et al 2018 ; Toropova et al 2018 ; Toropov and Toropova 2017 ; Troger et al 2020 ) and to predict toxicological properties from their structure.…”

Comparative Study for Prediction of Low and High Plasma Protein Binding Drugs by Various Machine Learning-Based Classification Algorithms

“…The Naïve Bayes algorithm is a simple, clear, and fast classifier algorithm based on the Bayesian approach 16 . It assumes mutually independent attributes, therefore, called naïve.…”

“…ARFF, CSV, Lib SVM, and C4.5 are the supported file formats in WEKA 13 . WEKA comprises various tools for data filtering, pre-processing, classification models (inbuild and custom), regression analysis, clustering methodologies, association rules, and a few aspects of visualization 16 . It is also well-suited for developing new machine learning schemes.…”

Artificial Intelligence in Drug Discovery: A Comprehensive Review of Data-driven and Machine Learning Approaches

“…The binding rate of drugs to the plasma protein is an important property to predict distribution because the drug molecules have no pharmacological effect when they form the protein-ligand complex, although they have reached the target tissues [127]. With the in vivo or in vitro measured data, the in silico PPB predictive models are still actively studied [128][129][130][131][132]. Wang et al [129] collected a comprehensive PPB dataset from various literatures and the DrugBank database and constructed a prediction model with the optimized feature set and the ensemble of various machine learning models.…”

“…For modeling, engineered molecular descriptors and various machine learning models are used. Toma et al [131] collected in vivo data for PPB prediction modeling. They calculated molecular 2D descriptors and SMILESbased features and trained the RF model.…”