Combining SILCS and Artificial Intelligence for High-Throughput Prediction of the Passive Permeability of Drug Molecules

Abstract

Membrane permeability of drug molecules plays a significant role in the development of new therapeutic agents. Accordingly, methods to predict the passive permeability of drug candidates during a medicinal chemistry campaign offer the potential to accelerate the drug design process. In this work, we combine the physics-based site identification by ligand competitive saturation (SILCS) method and data-driven artificial intelligence (AI) to create a high-throughput predictive model for the passive permeability of druglike molecules. In this study, we present a comparative analysis of four regression models to predict membrane permeabilities of small druglike molecules; of the tested models, Random Forest was the most predictive yielding an R 2 of 0.81 for the independent data set. The input feature vector used to train the developed prediction model includes absolute free energy profiles of ligands through a POPC-cholesterol bilayer based on ligand grid free energy (LGFE) profiles obtained from the SILCS approach. The use of the membrane free energy profiles from SILCS offers information on the physical forces contributing to ligand permeability, while the use of AI yields a more predictive model trained on experimental PAMPA permeability data for a collection of 229 molecules. This combination allows for rapid estimations of ligand permeability at a level of accuracy beyond currently available predictive models while offering insights into the contributions of the functional groups in the ligands to the permeability barrier, thereby offering quantitative information to facilitate rational ligand design.

Application of Generative Artificial Intelligence in Predicting Membrane Partitioning of Drugs: Combining Denoising Diffusion Probabilistic Models and MD Simulations Reduces the Computational Cost to One-Third

“…The choice of US sampling windows to be used for training the models was important, as the best results were obtained from nearly equally spaced windows, including the bounds of the windows at 0 and 30, but sampling more windows representing regions around the headgroup/water and headgroup/core interfaces. This observation is in agreement with previous work where restraint centers around these regions were determined to be the most important features for predicting the passive permeability of drug molecules …”

“…While limitations have already been described that make it challenging to calculate absolute membrane permeability, various methods have been developed to make such predictions. These include modeling based on experimental in vitro data, implicit membrane combined with machine learning, MD simulation, and machine learning . A major limitation of most of these methods has been the lack of atomistic data, making it difficult to predict the absolute permeability and the underlying free energy surface.…”

“…These include modeling based on experimental in vitro data, 56 implicit membrane combined with machine learning, 57 MD simulation, and machine learning. 58 A major limitation of most of these methods has been the lack of atomistic data, making it difficult to predict the absolute permeability and the underlying free energy surface. While certain methods have been developed to address this issue, 58 they still require extensive MD simulations, which are both timeintensive and computationally expensive to obtain the data needed to train these models.…”

Enhancing SILCS-MC via GPU Acceleration and Ligand Conformational Optimization with Genetic and Parallel Tempering Algorithms

“…This approach allows for assessment of GFE scores of individual classified atoms within the ligand that may be summed yielding the LGFE score that represents the ligand binding affinity. ,,, Notably, it has been shown that ligand ranking using the SILCS methodology is comparable with the FEP approaches while offering an improvement in computational speed of over 2 orders of magnitude. In addition, SILCS-MC has been used in conjunction with FragMaps for lipid bilayers to calculate the free energy profiles of ligands across the bilayers, information subsequently used in conjunction with a deep neural net to develop an artificial intelligence tool for prediction of ligand passive permeability. − …”

“…SILCS-MC has been successfully applied in a number of studies and is computationally efficient, − ,− requiring minutes on a single CPU core for full docking of a ligand . However, the algorithm requires multiple SILCS-MC runs to obtain adequate convergence defined as 0.5 kcal/mol in the context of the LGFE.…”