Combined Machine Learning and GRID-Independent Molecular Descriptor (GRIND) Models to Probe the Activity Profiles of 5-Lipoxygenase Activating Protein Inhibitors.

Hafiza Aliza Khan,Ishrat Jabeen

doi:10.3389/fphar.2022.825741

Abstract

Leukotrienes (LTs) are pro-inflammatory lipid mediators derived from arachidonic acid (AA), and their high production has been reported in multiple allergic, autoimmune, and cardiovascular disorders. The biological synthesis of leukotrienes is instigated by transfer of AA to 5-lipoxygenase (5-LO) via the 5-lipoxygenase-activating protein (FLAP). Suppression of FLAP can inhibit LT production at the earliest level, providing relief to patients requiring anti-leukotriene therapy. Over the last 3 decades, several FLAP modulators have been synthesized and pharmacologically tested, but none of them could be able to reach the market. Therefore, it is highly desirable to unveil the structural requirement of FLAP modulators. Here, in this study, supervised machine learning techniques and molecular modeling strategies are adapted to vaticinate the important 2D and 3D anti-inflammatory properties of structurally diverse FLAP inhibitors, respectively. For this purpose, multiple machine learning classification models have been developed to reveal the most relevant 2D features. Furthermore, to probe the 3D molecular basis of interaction of diverse anti-inflammatory compounds with FLAP, molecular docking studies were executed. By using the most probable binding poses from docking studies, the GRIND model was developed, which indicated the positive contribution of four hydrophobic, two hydrogen bond acceptor, and two shape-based features at certain distances from each other towards the inhibitory potency of FLAP modulators. Collectively, this study sheds light on important two-dimensional and three-dimensional structural requirements of FLAP modulators that can potentially guide the development of more potent chemotypes for the treatment of inflammatory disorders.

Highlights

The 5-LO pathway is responsible for the biological synthesis of leukotrienes (LTs) using arachidonic acid (AA) predominately by inflammatory cells like polymorphonuclear leukocytes, activated macrophages, and mast cells upon arrival of immunologic and non-immunologic stimuli (Hedi and Norbert 2004)
Six machine learning (ML) models were developed by different algorithms (SVM, logistic regression (LR), multilayer perceptron (MLP), decision tree (DT), random forest (RF), and gradient boost decision tree (GBDT)) using two-dimensional structural features of FLAP inhibitors
We developed multiple ML models including XGboost (GBDT), random forest (RF), decision tree (DT), support vector machine (SVM), logistic regression (LR), and multilayer perceptron (MLP), and in comparison, XGBoost and RF were able to classify our training set and predict the test set with significant classification and prediction accuracies

Summary

Introduction

The 5-LO pathway is responsible for the biological synthesis of leukotrienes (LTs) using arachidonic acid (AA) predominately by inflammatory cells like polymorphonuclear leukocytes, activated macrophages, and mast cells upon arrival of immunologic and non-immunologic stimuli (Hedi and Norbert 2004). LTA4 acts as a common precursor for biosynthesis of chemoattractant leukotriene B4 (LTB4) by a zinc-bound LTA4 hydrolase (LTA4H) and bronchoconstrictive cysteinyl leukotrienes (CysLTs or LTC4) with the help of LTC4 synthase (LTC4S) (Jakschik and Kuo 1983; Haeggström 2000). Both LTB4 and CysLTs are physiologically active final products of the 5-LO pathway and are exported out of the cell through specific transport proteins while extracellular peptidases metabolize LTC4 to LTD4, which is converted into LTE4 depending on type of inflammatory signal and cell demand (Jedlitschky and Keppler 2002). LTs bind with respective G-proteincoupled receptors, e.g., LTB4 binds with BLT1 and BLT2, whereas CysLTs activate CysLT1 and CYsLT2 receptors to incite further proinflammatory signaling cascades (Ghosh et al, 2016)

Methods

Results

Discussion

Conclusion