Probabilistic Approach for Virtual Screening Based on Multiple Pharmacophores

Abstract

Pharmacophore modeling is usually considered as a special type of virtual screening without probabilistic nature. Correspondence of at least one conformation of a molecule to pharmacophore is considered as evidence of its bioactivity. We show that pharmacophores can be treated as one-class machine learning models, and the probability the reflecting model’s confidence can be assigned to a pharmacophore on the basis of their precision of active compounds identification on a calibration set. Two schemes (Max and Mean) of probability calculation for consensus prediction based on individual pharmacophore models were proposed. Both approaches to some extent correspond to commonly used consensus approaches like the common hit approach or the one based on a logical OR operation uniting hit lists of individual models. Unlike some known approaches, the proposed ones can rank compounds retrieved by multiple models. These approaches were benchmarked on multiple ChEMBL datasets used for ligand-based pharmacophore modeling and externally validated on corresponding DUD-E datasets. The influence of complexity of pharmacophores and their performance on a calibration set on results of virtual screening was analyzed. It was shown that Max and Mean approaches have superior early enrichment to the commonly used approaches. Thus, a well-performing, easy-to-implement, and probabilistic alternative to existing approaches for pharmacophore-based virtual screening was proposed.

QSAR Modeling Based on Conformation Ensembles Using a Multi-Instance Learning Approach

“…Distances between features were binned to allow fuzzy matching of quadruplets with small differences in the position of features. Here, we used the 1 Å bin step as it demonstrated reasonable performance in our previous studies. ,− Three-dimensional pharmacophore signatures were generated for each quadruplet according to the algorithm described in our previous publication . These signatures consider distances between features and their spatial arrangement to recognize the stereoconfiguration of the quadruplets.…”

Hierarchical Graph Representation of Pharmacophore Models

“…Motivated by the recent interest in consensus-based virtual screening methods involving pharmacophore models (Wieder et al, 2017;Polishchuk et al, 2019;Madzhidov et al, 2020), we developed an intuitive hierarchical graph representation of pharmacophore models. A user-friendly interactive visualization of the pharmacophore-based graph provides valuable information for computational chemists toward the understanding of protein-ligand interaction patterns and can aid in the selection of pharmacophore models for VS experiments.…”

“…Recently, Polishchuk et al (2019) improved the workflow by adapting the consensus scoring function to consider the number of conformations of each molecule retrieved by the VS runs. Based on these studies, Madzhidov et al (2020) analyzed the performance of a set of pharmacophore models and developed a probabilistic approach for consensus scoring, leading to a method which is less sensitive to the poor performing models in the pool. Although these consensus-based approaches provided better results than a "classical" pharmacophore approach, they demanded considerable computational resources due to the required multiple VS runs.…”

Applications of Virtual Screening in Bioprospecting: Facts, Shifts, and Perspectives to Explore the Chemo-Structural Diversity of Natural Products

“…Different computational methods have been developed over the years and implemented in virtual screening strategies (Tomar et al, 2018 ), applying knowledge of artificial intelligence (Gupta et al, 2013 ; Yang et al, 2018 ; Schaduangrat et al, 2019 ; Shoombuatong et al, 2019 ; Kong et al, 2020 ), molecular modeling (Semighini et al, 2011 ; Rampogu et al, 2018 ; Da Costa et al, 2019 ; Jin et al, 2020 ; Mascarenhas et al, 2020 ), statistics, and probability (Pire et al, 2015 ; Daina and Zoete, 2016 ; Blanco et al, 2018 ; Madzhidov et al, 2020 ; Cai et al, 2021 ). These methods, when combined with experimental approaches, increase the success to finding novel bioactive compounds (Kumar and Zhang, 2015 ; Coimbra et al, 2020 ; Gorgulla et al, 2020 ; Stokes et al, 2020 ).…”