Search

In this work, we propose a machine learning approach to generate novel molecules starting from a seed compound, its three-dimensional (3D) shape, and its pharmacophoric features. The pipeline draws inspiration from generative models used in image analysis and represents a first example of the de novo design of lead-like molecules guided by shape-based features. A variational autoencoder is used to perturb the 3D representation of a compound, followed by a system of convolutional and recurrent neural networks that generate a sequence of SMILES tokens. The generative design of novel scaffolds and functional groups can cover unexplored regions of chemical space that still possess lead-like properties.

Drug discovery suffers from high attrition because compounds initially deemed as promising can later show ineffectiveness or toxicity resulting from a poor understanding of their activity profile. In this work, we describe a deep self-normalizing neural network model for the prediction of molecular pathway association and evaluate its performance, showing an AUC ranging from 0.69 to 0.91 on a set of compounds extracted from ChEMBL and from 0.81 to 0.83 on an external data set provided by Novartis. We finally discuss the applicability of the proposed model in the domain of lead discovery. A usable application is available via PlayMolecule.org .

Molecular recognition is a crucial issue when aiming to interpret the mechanism of known active substances as well as to develop novel active candidates. Unfortunately, simulating the binding process is still a challenging task because it requires classical MD experiments in a long microsecond time scale that are affordable only with a high-level computational capacity. In order to overcome this limiting factor, we have recently implemented an alternative MD approach, named supervised molecular dynamics (SuMD), and successfully applied it to G protein-coupled receptors (GPCRs). SuMD enables the investigation of ligand-receptor binding events independently from the starting position, chemical structure of the ligand, and also from its receptor binding affinity. In this article, we present an extension of the SuMD application domain including different types of proteins in comparison with GPCRs. In particular, we have deeply analyzed the ligand-protein recognition pathways of six different case studies that we grouped into two different classes: globular and membrane proteins. Moreover, we introduce the SuMD-Analyzer tool that we have specifically implemented to help the user in the analysis of the SuMD trajectories. Finally, we emphasize the limit of the SuMD applicability domain as well as its strengths in analyzing the complexity of ligand-protein recognition pathways.

Recent advances in structural biology revealed that water molecules play a crucial structural role in the protein architecture and ligand binding of G protein-coupled receptors. In this work, we present an alternative approach to monitor the time-dependent organization of water molecules during the final stage of the ligand-receptor recognition process by means of membrane molecular dynamics simulations. We inspect the variation of fluid dynamics properties of water molecules upon ligand binding with the aim to correlate the results with the binding affinities. The outcomes of this analysis are transferred into a bidimensional graph called water fluid dynamics maps, that allow a fast graphical identification of protein "hot-spots" characterized by peculiar shape and electrostatic properties that can play a critical role in ligand binding. We hopefully believe that the proposed approach might represent a valuable tool for structure-based drug discovery that can be extended to cases where crystal structures are not yet available, or have not been solved at high resolution.

G protein-coupled receptors (GPCRs) represent the largest family of cell-surface receptors and about one-third of the actual targets of clinically used drugs. Following the progress made in the field of GPCRs structural determination, docking-based screening for novel potent and selective ligands is becoming an increasingly adopted strategy in the drug discovery process. However, this methodology is not yet able to anticipate the "bioactive" binding mode and discern it among other conformations. In the present work, we present a novel approach consisting in the integration of molecular docking and membrane MD simulations with the aim to merge the rapid sampling of ligand poses into in the binding site, typical of docking algorithms, with the thermodynamic accuracy of MD simulations in describing, at the molecular level, the stability a GPCR-ligand complex embedded into explicit lipid-water environment. To validate our approach, we have chosen as a key study the human A(2A) adenosine receptor (hA(2A) AR) and selected four receptor-antagonist complexes and one receptor-agonist complex that have been recently crystallized. In light of the obtained results, we believe that our novel strategy can be extended to other GPCRs and might represent a valuable tool to anticipate the "bioactive" conformation of high-affinity ligands.