Machine Learning Assisted Approach for Finding Novel High Activity Agonists of Human Ectopic Olfactory Receptors

Abstract

Olfactory receptors (ORs) constitute the largest superfamily of G protein-coupled receptors (GPCRs). ORs are involved in sensing odorants as well as in other ectopic roles in non-nasal tissues. Matching of an enormous number of the olfactory stimulation repertoire to its counterpart OR through machine learning (ML) will enable understanding of olfactory system, receptor characterization, and exploitation of their therapeutic potential. In the current study, we have selected two broadly tuned ectopic human OR proteins, OR1A1 and OR2W1, for expanding their known chemical space by using molecular descriptors. We present a scheme for selecting the optimal features required to train an ML-based model, based on which we selected the random forest (RF) as the best performer. High activity agonist prediction involved screening five databases comprising ~23 M compounds, using the trained RF classifier. To evaluate the effectiveness of the machine learning based virtual screening and check receptor binding site compatibility, we used docking of the top target ligands to carefully develop receptor model structures. Finally, experimental validation of selected compounds with significant docking scores through in vitro assays revealed two high activity novel agonists for OR1A1 and one for OR2W1.

Olfactory Receptors as an Emerging Chemical Sensing Scaffold

“…Machine learning (ML) approaches have the potential to accelerate OR deorphanization by leveraging both computational and experimental data. Although ML approaches have been widely implemented in chemosensory research, that is to predict the perception response to an odorant, ML approaches to predict OR-ligand pairs have been limited to expanding the number of ligands for ORs with known ligands, − or making OR-ligand predictions without experimental validation . Other computational approaches have been applied to the identification of alternative ligands for ORs with known ligands, but not to the more difficult challenge of OR deorphanization.…”

Modeling the Orthosteric Binding Site of the G Protein-Coupled Odorant Receptor OR5K1

“…38 Currently, no experimental structures of human ORs are available, and homology modeling techniques have been used to rationalize the binding modes of odorant compounds into ORs and discover new OR ligands. 37,[39][40][41][42][43] AI-based methods are emerging as compelling tools to predict the 3D structure of proteins. [44][45][46] During the CASP (Critical Assessment of Structure Prediction) 14 competition, AlphaFold 2 (AF2) was shown to be able to predict the structure of protein domains at an accuracy matching experimental methods.…”

“…26-31 Despite current efforts in assigning ORs to odorant molecules, or, vice versa, in defining the chemical ligand space of individual ORs, only the molecular recognition ranges of a few ORs have been investigated. 27, 32-37…”

“…[26][27][28][29][30][31] Despite current efforts in assigning ORs to odorant molecules, or vice versa, in defining the chemical ligand space of individual ORs, only the molecular recognition ranges of a few ORs have been investigated. 27,[32][33][34][35][36][37][38] Structure-based virtual screening campaigns have been successfully applied for GPCR ligand discovery and are always more in use with the recent extraordinary advances in GPCR structural biology. 39 Currently, no experimental structures of human ORs are available, and homology modeling techniques have been used to rationalize the binding modes of odorant compounds into ORs and discover new OR ligands.…”

Editorial of Special Issue “Deep Learning and Machine Learning in Bioinformatics”

“…Charoenkwan et al developed a sequence-based predictor, named iBitter-Fuse, to identify bitter peptides by fusing multi-view features [ 16 ]. Jabeen et al adopted a random forest model to identify novel high activity agonists of human ectopic olfactory receptors [ 17 ]. Pouryahya et al proposed a network-based clustering method coupled with optimal mass transport theory to predict cell line-drug sensitivity, and showed that random forest modeling conducted on the resulting cell line-drug clusters outperformed alternative computational methods in predicting in vitro drug responses [ 18 ].…”