A Computational Approach to Improve Prediction of Off-Target Drug Binding Using Membrane Bilayer Effects

Helgi Ingolfsson | 18-ERD-035

Project Overview

Drugs can alter membrane protein function not only through classical protein-ligand interactions, but also indirectly via changes to the bilayer environment. These bilayer-mediated effects are nonspecific and can lead to severe side effects, especially at large drug dosages. The goal of this project was to elucidate the molecular mechanisms underlying those effects and to develop computational tools to screen for problematic drug candidates.

We developed and validated a high-throughput molecular dynamics assay, utilizing the mini-protein gramicidin A as a probe (like a molecular canary in the coal mine). The assay can quantify the bilayer-modifying potency of different drugs in silico, even before they have been synthesized. We tested a library of more than 5,000 different drug molecules. Then, we used the library to expand our understanding of the molecular mechanisms for how small molecules change bilayer properties, and to train a machine-learning algorithm to computationally and inexpensively predict drug bilayer-modifying potency.

We have advanced our understanding of the underlying mechanisms for what makes molecules alter bilayer properties, and we developed simulation and machine-learning drug-screening methods that can be used to predict bilayer-modifying drugs. These methods provide a new type of safety and off-target checks that can be added to drug development pipelines, leading to safer drugs and reduced drug development costs.

Mission Impact

This research enhances Lawrence Livermore National Laboratory's core competencies in bioscience and bioengineering, as well as high-performance computing, simulation, and data science, by using the Laboratory's expertise in simulation, machine learning, automation, force-field development, drug development, and biological membranes, as well as its significant computational resources, to develop a technology for assessing drug safety. By providing in silico drug safety tools, this project facilitates the process of developing and producing safer drugs and benefits the Laboratory's mission research challenge in chemical and biological countermeasures.

Publications, Presentations, and Patents

Bennett, W. F. D., et al. 2019. "Interfacial and Hydrophobicity Scales for Small Drug-Like Molecules from Atomistic Free Energy Calculations," Biophysical Society Meeting, Baltimore, MA. Biophys. J. Supplement. LLNL-ABS-758788

——— 2020. "Predicting Small Molecule Transfer Free Energies by Combining Molecular Dynamics Simulations and Deep Learning." Journal of Chemical Information and Modeling. doi:10.1021/acs.jcim.0c00318. LLNL-JRNL-774697

Blumer, M., et al. 2020. "Simulations of asymmetric membranes illustrate cooperative leaflet coupling and lipid adaptability." Frontiers Cell Developmental Biology 8:575. doi:10.3389/fcell.2020.00575. LLNL-JRNL-801959

He, S., et al. 2020. "3D Convolutional Neural Network for Predicting Free Energies of Partitioning." Biophysical Society Meeting, San Diego, CA. Biophys. J. Supplement, Poster. LLNL-POST-804161

Ingólfsson, H. I. 2018. "Coarse-grained simulations: introduction and application to membranes. School on computer simulations of biological membranes & free energy calculations of biomolecular systems." University of Los Andes, Bogotá, Colombia. Invited talk. LLNL-PRES-756273

——— 2019. "Predicting Dirty Drugs: Using High-Throughput Simulations and Machine Learning to Detect Troublesome Bilayer-Modifying Molecules." Data Science Institute Workshop, Livermore, CA. Invited talk. LLNL-PRES-782124

——— 2020. "Simulating Plasma Membranes: Effects of Leaflet Asymmetry and Commotional Complexity." Biophysical Society Meeting, Membrane Structure and Function Subgroup, San Diego, CA. Invited talk. LLNL-PRES-805148

Sun, D., et al. 2019a. "Molecular mechanism for gramicidin dimerization and dissociation in bilayers of different thickness." Biophysical Journal 117:1831-1844. doi:10.1016/j.bpj.2019.09.044. LLNL-JRNL-768282

——— 2019b. "Predicting the Promiscuous Effect of Amphipathic Drugs on Gramicidin Channel Stability with Simulations and Experiments." Biophysical Society Meeting, Baltimore, MA. Biophys. J. Supplement. LLNL-ABS-758962

——— 2019c. "Predicting dirty drugs: An in silico approach for predicting bilayer-mediated, off-target drug effects." LLNL Physical and Life Sciences External Review Committee Meeting, Livermore, CA. Invited poster. LLNL-POST-777505

——— 2020a. "Atomistic Characterization of Gramicidin Channel Formation." Journal of Chemical Theory and Computation. doi.org:10.1021/acs.jctc.0c00989. LLNL-JRNL-813847

——— 2020b. "Assessing the Perturbing Effects of Drugs on Lipid Bilayers using Gramicidin Channel-Based in Silico and in Vitro Assays." Journal of Medicinal Chemistry, doi:10.1021/acs.jmedchem.0c00958. LLNL-JRNL-807837

——— 2020c. "Molecular Process of Gramicidin a Dimerization Determined with Milliseconds Atomistic Simulations and Machine Learning." Biophysical Society Meeting, San Diego, CA. Biophys. J. Supplement. LLNL-ABS-791900

Zhang, M., et al. 2018. "Fluorinated alcohols' effects on lipid bilayer properties.” Biophysical Journal 115:679-689. doi:10.1016/j.bpj.2018.07.010. LLNL-JRNL-744560