Time-Domain Analysis of Molecular Dynamics Trajectories Using Deep Neural Networks: Application to Activity Ranking of Tankyrase Inhibitors

Abstract

Molecular dynamics simulations provide valuable insights into the behavior of molecular systems. Extending the recent trend of using machine learning techniques to predict physicochemical properties from molecular dynamics data, we propose to consider the trajectories as multidimensional time series represented by 2D tensors containing the ligand–protein interaction descriptor values for each time step. Similar in structure to the time series encountered in modern approaches for signal, speech, and natural language processing, these time series can be directly analyzed using long short-term memory (LSTM) recurrent neural networks or convolutional neural networks (CNNs). The predictive regression models for the ligand–protein affinity were built for a subset of the PDBbind v.2017 database and applied to inhibitors of tankyrase, an enzyme of the poly­(ADP-ribose)-polymerase (PARP) family that can be used in the treatment of colorectal cancer. As an additional test set, a subset of the Community Structure–Activity Resource (CSAR) data set was used. For comparison, the random forest and simple neural network models based on the crystal pose or the trajectory-averaged descriptors were used, as well as the commonly employed docking and molecular mechanics Poisson–Boltzmann surface area (MM-PBSA) scores. Convolutional neural networks based on the 2D tensors of ligand–protein interaction descriptors for short (2 ns) trajectories provide the best accuracy and predictive power, reaching the Spearman rank correlation coefficient of 0.73 and Pearson correlation coefficient of 0.70 for the tankyrase test set. Taking into account the recent increase in computational power of modern GPUs and relatively low computational complexity of the proposed approach, it can be used as an advanced virtual screening filter for compound prioritization.

Discovery of Novel Tankyrase Inhibitors through Molecular Docking-Based Virtual Screening and Molecular Dynamics Simulation Studies

“…Also, it is highly desirable to conduct a separate study to evaluate the correlation of FEP energies with available experimental activity values of the known inhibitors. Alternatively, the novel machine learning approach to the analysis of the molecular dynamics trajectories [ 22 ] could be applied to predict the inhibitor activities.…”

Scoring Functions for Protein-Ligand Binding Affinity Prediction Using Structure-based Deep Learning: A Review

“…Yakovenko and Jones (2017) use atomic densities but trained their model on both docked poses and MD trajectory frames to obtain learned representations later used to predict LIE. Berishvili et al (2019) developed 1D descriptors based on GROMACS (Berendsen et al, 1995;Abraham et al, 2015), AutoDock Vina (Trott and Olson 2009), and SMINA (Koes et al, 2013) terms to describe frames from MD trajectories. The descriptor for each frame where stacked together into a matrix of size n descriptor × n frames , representing the whole MD trajectory.…”

Recent Developments in Free Energy Calculations for Drug Discovery

“…Furthermore, due to its reliability MM-PBSA is often used as a baseline comparison or in combination with alternative methods for higher performance. Machine learning methods based on extracting protein-ligand interaction descriptors as features from MD simulation are compared to MM-PBSA on the tankyrase system ( Berishvili et al, 2019 ). Machine learning also accelerates pose prediction methods based on short MD simulation combined with MM-PBSA through the Best Arm Identification method to obtain the correct binding pose with minimal number of runs ( Terayama et al, 2018 ).…”