GPRC Molecular Dynamics Forecasting using Recurrent Neural Networks

Abstract

G protein-coupled receptors (GPCRs) are a large super-family of cell membrane proteins that play an important physiological role as transmitters of extra-cellular signals.Signal transmission through the cell membrane depends upon conformational changes in the transmembrane region of the receptor, which makes the investigation of the dynamics in these regions particularly relevant.Molecular dynamics simulations can provide information about the conformational states of the receptor at the atom level and Machine Learning methods can be useful tools for knowledge discovery from this information. In this paper, Recurrent and Convolutional Neural Networks are used to predict the molecular dynamics of GPCRs in different conformations, using the \textbeta 2AR GPRC as an illustration of the process and focusing on specific receptor regions.Active and inactive states of the GPRC are analysed in six scenarios involving APO, Full Agonist (BI-167107) and Partial Inverse Agonist (carazolol) of the receptor. Three Machine Learning models with increasing complexity in terms of neural network architecture are evaluated and their results compared.Results show that the transmembrane helices are the regions whose dynamics are best predicted by these models.