A Machine Learning Approach for the Discovery of Ligand-Specific Functional Mechanisms of GPCRs

Abstract

G protein-coupled receptors (GPCRs) play a key role in many cellular signaling mechanisms, and must select among multiple coupling possibilities in a ligand-specific manner in order to carry out a myriad of functions in diverse cellular contexts. Much has been learned about the molecular mechanisms of ligand-GPCR complexes from Molecular Dynamics (MD) simulations. However, to explore ligand-specific differences in the response of a GPCR to diverse ligands, as is required to understand ligand bias and functional selectivity, necessitates creating very large amounts of data from the needed large-scale simulations. This becomes a Big Data problem for the high dimensionality analysis of the accumulated trajectories. Here we describe a new machine learning (ML) approach to the problem that is based on transforming the analysis of GPCR function-related, ligand-specific differences encoded in the MD simulation trajectories into a representation recognizable by state-of-the-art deep learning object recognition technology. We illustrate this method by applying it to recognize the pharmacological classification of ligands bound to the 5-HT2A and D2 subtypes of class-A GPCRs from the serotonin and dopamine families. The ML-based approach is shown to perform the classification task with high accuracy, and we identify the molecular determinants of the classifications in the context of GPCR structure and function. This study builds a framework for the efficient computational analysis of MD Big Data collected for the purpose of understanding ligand-specific GPCR activity.

Ligand-Dependent Conformational Transitions in Molecular Dynamics Trajectories of GPCRs Revealed by a New Machine Learning Rare Event Detection Protocol

“…Central among these tools and approaches are MD simulations that offer the ability to follow the dynamic changes in GPCR molecular structure in response to the environmental condition (especially the membrane), interactions with other proteins, and the effects of ligand binding [32]. The need for, and the success of, using MD simulations for to study such processes involving the allosteric mechanisms of molecular machines in the membrane are well known [32][33][34][35], and so are the challenges presented by the very large amounts of resulting data that encode the desired information [7,36,37]. This difficulty is compounded by the fact that mechanistically important state-to-state transitions of complex molecular machines such as GPCRs must constitute rare events in the MD trajectories, reflecting the need to preserve the specificity of their regulation.…”

“…The conformational transitions underlying the functions of such systems involve collective motions that occur rarely in the dynamics of trajectories due to the high barrier associated with the simultaneous involvement of various structural elements [5,6]. The identification of such conformational transitions in MD trajectories of GPCRs has proven essential in revealing the dynamic elements of a receptor's response to ligands that differ in their pharmacological properties [7] and, even more intriguingly, the role of specific conformational dynamics of the receptor that prepare for differential coupling (i.e., functional selectivity [8]). We have demonstrated the collective nature of these conformational transitions in studies of ligand-dependent functional selectivity of the 5-HT 2A serotonin receptor (5-HT 2A R) and the dopamine D2 receptor (e.g., see [7,[9][10][11]), and have shown that understanding such ligand-determined GPCR functions depends on a rigorous identification and analysis of the diverse function-related conformational transitions induced by various ligands.…”

IMPROvER: the Integral Membrane Protein Stability Selector

“…Instead, we extracted a linear scoring matrix, allowing us to observe trends in membrane protein stability and to devise a computationally-efficient tool that helped us stabilise three membrane proteins with different folds and modes of action. To apply a neural network approach that would work broadly with membrane proteins as has been done for GPCRs 56 – 58 would require access to large systematic stabilisation datasets, with a record of both stabilising and destabilising variants of membrane proteins from diverse folds. We hope to increase the reporting, sharing and depositing of such data by releasing IMPROvER for use in the academic community http://improver.ddns.net/IMPROvER/ .…”

DROIDS 3.0—Detecting Genetic and Drug Class Variant Impact on Conserved Protein Binding Dynamics

“…Therefore, the high-performance accelerated computation used to generate simulated protein motions for comparison can be effectively partnered with high-performance methods for optimally extracting and learning the underlying dynamic feature differences that define the different functional states of proteins. Although machine learning has recently been applied to individual MD studies for a variety of specific tasks (17)(18)(19), there is no current software platform for the general application of machine learning to general comparative problems in protein dynamics.…”