3D matters! 3D-RISM and 3D convolutional neural network for accurate bioaccumulation prediction

graphLambda: Fusion Graph Neural Networks for Binding Affinity Prediction

Predicting the binding affinity of protein-ligand complexes is crucial for computer-aided drug discovery (CADD) and the identification of potential drug candidates. The deep learning-based scoring functions have emerged as promising predictors of binding constants. Building on recent advancements in graph neural networks, we present graphLambda for protein-ligand binding affinity prediction, which utilizes graph convolutional, attention, and isomorphism blocks to enhance the predictive capabilities. The graphLambda model exhibits superior performance across CASF16 and CSAR HiQ NRC benchmarks and demonstrates robustness with respect to different types of train-validation set partitions. The development of graphLambda underscores the potential of graph neural networks in advancing binding affinity prediction models, contributing to more effective CADD methodologies.

gr Predictor: A Deep Learning Model for Predicting the Hydration Structures around Proteins

Among the factors affecting biological processes such as protein folding and ligand binding, hydration, which is represented by a three-dimensional water site distribution function around the protein, is crucial. The typical methods for computing the distribution functions, including molecular dynamics simulations and the three-dimensional reference interaction site model (3D-RISM) theory, require a long computation time ranging from hours to tens of hours. Here, we propose a deep learning (DL) model that rapidly estimates the distribution functions around proteins obtained using the 3D-RISM theory from the protein 3D structure. The distribution functions predicted using our DL model are in good agreement with those obtained using the 3D-RISM theory. Particularly, the coefficient of determination between the distribution function obtained by the DL model and that obtained using the 3D-RISM theory is approximately 0.98. Furthermore, using a graphics processing unit, the prediction by the DL model is completed in less than 1 min, more than 2 orders of magnitude faster than the calculation time of the 3D-RISM theory. The position of water molecules around the protein was estimated based on the distribution function obtained by our DL model, and the position of waters estimated by our DL model was in good agreement with that of water molecules estimated using the 3D-RISM theory and of crystallographic waters. Therefore, our DL model provides a practical and efficient way to calculate the three-dimensional water site distribution functions and to estimate the position of water molecules around the protein. The program called "gr Predictor" is available under the GNU General Public License from https://github.com/YoshidomeGroup-Hydration/gr-predictor.

Computational Indicator Approach for Assessment of Nanotoxicity of Two-Dimensional Nanomaterials

The increasing growth in the development of various novel nanomaterials and their biomedical applications has drawn increasing attention to their biological safety and potential health impact. The most commonly used methods for nanomaterial toxicity assessment are based on laboratory experiments. In recent years, with the aid of computer modeling and data science, several in silico methods for the cytotoxicity prediction of nanomaterials have been developed. An affordable, cost-effective numerical modeling approach thus can reduce the need for in vitro and in vivo testing and predict the properties of designed or developed nanomaterials. We propose here a new in silico method for rapid cytotoxicity assessment of two-dimensional nanomaterials of arbitrary chemical composition by using free energy analysis and molecular dynamics simulations, which can be expressed by a computational indicator of nanotoxicity (CIN_2D). We applied this approach to five well-known two-dimensional nanomaterials promising for biomedical applications: graphene, graphene oxide, layered double hydroxide, aloohene, and hexagonal boron nitride nanosheets. The results corroborate the available laboratory biosafety data for these nanomaterials, supporting the applicability of the developed method for predictive nanotoxicity assessment of two-dimensional nanomaterials.

Biomolecular Simulations with the Three-Dimensional Reference Interaction Site Model with the Kovalenko-Hirata Closure Molecular Solvation Theory

The statistical mechanics-based 3-dimensional reference interaction site model with the Kovalenko-Hirata closure (3D-RISM-KH) molecular solvation theory has proven to be an essential part of a multiscale modeling framework, covering a vast region of molecular simulation techniques. The successful application ranges from the small molecule solvation energy to the bulk phase behavior of polymers, macromolecules, etc. The 3D-RISM-KH successfully predicts and explains the molecular mechanisms of self-assembly and aggregation of proteins and peptides related to neurodegeneration, protein-ligand binding, and structure-function related solvation properties. Upon coupling the 3D-RISM-KH theory with a novel multiple time-step molecular dynamic (MD) of the solute biomolecule stabilized by the optimized isokinetic Nosé-Hoover chain thermostat driven by effective solvation forces obtained from 3D-RISM-KH and extrapolated forward by generalized solvation force extrapolation (GSFE), gigantic outer time-steps up to picoseconds to accurately calculate equilibrium properties were obtained in this new quasidynamics protocol. The multiscale OIN/GSFE/3D-RISM-KH algorithm was implemented in the Amber package and well documented for fully flexible model of alanine dipeptide, miniprotein 1L2Y, and protein G in aqueous solution, with a solvent sampling rate ~150 times faster than a standard MD simulation in explicit water. Further acceleration in computation can be achieved by modifying the extent of solvation layers considered in the calculation, as well as by modifying existing closure relations. This enhanced simulation technique has proven applications in protein-ligand binding energy calculations, ligand/solvent binding site prediction, molecular solvation energy calculations, etc. Applications of the RISM-KH theory in molecular simulation are discussed in this work.

Comprehensive 3D-RISM analysis of the hydration of small molecule binding sites in ligand-free protein structures

Hydration is a critical factor in the ligand binding process. Herein, to examine the hydration states of ligand binding sites, the three‐dimensional distribution function for the water oxygen site, g_O(r), is computed for 3,706 ligand‐free protein structures based on the corresponding small molecule–protein complexes using the 3D‐RISM theory. For crystallographic waters (CWs) close to the ligand, g_O(r) reveals that several CWs are stabilized by interaction networks formed between the ligand, CW, and protein. Based on the g_O(r) for the crystallographic binding pose of the ligand, hydrogen bond interactions are dominant in the highly hydrated regions while weak interactions such as CH‐O are dominant in the moderately hydrated regions. The polar heteroatoms of the ligand occupy the highly hydrated and moderately hydrated regions in the crystallographic (correct) and wrongly docked (incorrect) poses, respectively. Thus, the g_O(r) of polar heteroatoms may be used to distinguish the correct binding poses.

Multisolvent Models for Solvation Free Energy Predictions Using 3D-RISM Hydration Thermodynamic Descriptors

The potential to predict solvation free energies (SFEs) in any solvent using a machine learning (ML) model based on thermodynamic output, extracted exclusively from 3D-RISM simulations in water is investigated. The models on multiple solvents take into account both the solute and solvent description and offer the possibility to predict SFEs of any solute in any solvent with root mean squared errors less than 1 kcal/mol. Validations that involve exclusion of fractions or clusters of the solutes or solvents exemplify the model's capability to predict SFEs of novel solutes and solvents with diverse chemical profiles. In addition to being predictive, our models can identify the solute and solvent features that influence SFE predictions. Furthermore, using 3D-RISM hydration thermodynamic output to predict SFEs in any organic solvent reduces the need to run 3D-RISM simulations in all these solvents. Altogether, our multisolvent models for SFE predictions that take advantage of the solvation effects are expected to have an impact in the property prediction space.

COVER: conformational oversampling as data augmentation for molecules

Training neural networks with small and imbalanced datasets often leads to overfitting and disregard of the minority class. For predictive toxicology, however, models with a good balance between sensitivity and specificity are needed. In this paper we introduce conformational oversampling as a means to balance and oversample datasets for prediction of toxicity. Conformational oversampling enhances a dataset by generation of multiple conformations of a molecule. These conformations can be used to balance, as well as oversample a dataset, thereby increasing the dataset size without the need of artificial samples. We show that conformational oversampling facilitates training of neural networks and provides state-of-the-art results on the Tox21 dataset.

graphDelta: MPNN Scoring Function for the Affinity Prediction of Protein-Ligand Complexes

In this work, we present graph-convolutional neural networks for the prediction of binding constants of protein-ligand complexes. We derived the model using multi task learning, where the target variables are the dissociation constant (K d), inhibition constant (K i), and half maximal inhibitory concentration (IC50). Being rigorously trained on the PDBbind dataset, the model achieves the Pearson correlation coefficient of 0.87 and the RMSE value of 1.05 in pK units, outperforming recently developed 3D convolutional neural network model K deep.

Effects of Mutating Trp42 Residue on γD-Crystallin Stability

Oligomerization and aggregation of γD-crystallins (HγDC) in the eye lens is one of the main causes of cataract development. To date, several congenital mutations related to this protein are known to promote the formation of aggregates. Previous studies have demonstrated that mutations in W42 residue of HγDC lead to the generation of partially unfolded intermediates that are more prone to aggregate. To understand the role of W42 in the stability of HγDC, we performed alchemical free-energy calculations and all-atom molecular dynamics simulations of different W42 mutant models. Our results suggest that substitution of W42 by small size and/or polar residues promotes HγDC denaturation due to the entry of water molecules into the hydrophobic core of the N-terminal domain. Similar behavior was observed in the C-terminal domain of HγDC when mutating the W130 residue located in a homologous position. Moreover, the exposure of the hydrophobic core residues could lead to the formation of aggregation-prone partially unfolded species. Overall, this study takes a step toward understanding the role of HγDC in cataract development.

Thermodynamic Stability of Human γD-Crystallin Mutants Using Alchemical Free-Energy Calculations

γD-Crystallin (HγDC) is a key structural protein in the human lens, whose aggregation has been associated with the development of cataracts. Single-point mutations and post-translational modifications destabilize HγDC interactions, forming partially folded intermediates, where hydrophobic residues are exposed and thus triggering its aggregation. In this work, we used alchemical free-energy calculations to predict changes in thermodynamic stability (ΔΔG) of 10 alanine-scanning variants and 12 HγDC mutations associated with the development of congenital cataract. Our results show that W42R is the most destabilizing mutation in HγDC. This has been corroborated through experimental determination of ΔΔG employing differential scanning calorimetry. Calculations of hydration free energies from the HγDC WT and the W42R mutant suggested that the mutant has a higher aggregation propensity. Our combined theoretical and experimental results contribute to understand HγDC destabilization and aggregation mechanisms in age-onset cataracts.