Expanding Predictive Capacities in Toxicology: Insights from Hackathon-Enhanced Data and Model Aggregation.

In the realm of predictive toxicology for small molecules, the applicability domain of QSAR models is often limited by the coverage of the chemical space in the training set. Consequently, classical models fail to provide reliable predictions for wide classes of molecules. However, the emergence of innovative data collection methods such as intensive hackathons have promise to quickly expand the available chemical space for model construction. Combined with algorithmic refinement methods, these tools can address the challenges of toxicity prediction, enhancing both the robustness and applicability of the corresponding models. This study aimed to investigate the roles of gradient boosting and strategic data aggregation in enhancing the predictivity ability of models for the toxicity of small organic molecules. We focused on evaluating the impact of incorporating fragment features and expanding the chemical space, facilitated by a comprehensive dataset procured in an open hackathon. We used gradient boosting techniques, accounting for critical features such as the structural fragments or functional groups often associated with manifestations of toxicity.

Improvement of multi-task learning by data enrichment: application for drug discovery.

Multi-task learning in deep neural networks has become a topic of growing importance in many research fields, including drug discovery. However, applying multi-task learning poses new challenges in improving prediction performance. This study investigated the potential of training data enrichment to enhance multi-task model prediction quality in drug discovery. The study evaluated four scenarios with varying degrees of information capacity of the training data and applied two types of test data to evaluate prediction performance. We used three datasets: ViralChEMBL, which consisted of binary activities of compounds against viral species, was applied for the classification task; pQSAR(159) and pQSAR(4267), which consisted of bio-activities of compounds and assays from the research of the profile-QSAR method, were applied for regression tasks. We built multi-task models based on the feed-forward DNNs using the PyTorch framework. Our findings showed that training data enrichment could be an effective means of enhancing prediction performance in multi-task learning, but the degree of improvement depends on the quality of the training data. The more unique compounds and targets the training data included, the more new compound-target interactions are required for prediction improvement. Also, we found out that even using multi-task learning, one could not predict the interactions of compounds that are highly dissimilar from those used for model training. The study provides some recommendations for effectively employing multi-task learning in drug discovery to improve prediction accuracy and facilitate the discovery of novel drug candidates.

Biphenyl scaffold for the design of NMDA-receptor negative modulators: molecular modeling, synthesis, and biological activity.

NMDA (N-methyl-d-aspartate) receptor antagonists are promising tools for the treatment of a wide variety of central nervous system impairments including major depressive disorder. We present here the activity optimization process of a biphenyl-based NMDA negative allosteric modulator (NAM) guided by free energy calculations, which led to a 100 times activity improvement (IC50 = 50 nM) compared to a hit compound identified in virtual screening. Preliminary calculation results suggest a low affinity for the human ether-a-go-go-related gene ion channel (hERG), a high affinity for which was earlier one of the main obstacles for the development of first-generation NMDA-receptor negative allosteric modulators. The docking study and the molecular dynamics calculations suggest a completely different binding mode (ifenprodil-like) compared to another biaryl-based NMDA NAM EVT-101.

Exploring Chemical Reaction Space with Reaction Difference Fingerprints and Parametric t-SNE.

Humans prefer visual representations for the analysis of large databases. In this work, we suggest a method for the visualization of the chemical reaction space. Our technique uses the t-SNE approach that is parameterized using a deep neural network (parametric t-SNE). We demonstrated that the parametric t-SNE combined with reaction difference fingerprints could provide a tool for the projection of chemical reactions on a low-dimensional manifold for easy exploration of reaction space. We showed that the global reaction landscape projected on a 2D plane corresponds well with the already known reaction types. The application of a pretrained parametric t-SNE model to new reactions allows chemists to study these reactions in a global reaction space. We validated the feasibility of this approach for two commercial drugs, darunavir and montelukast. We believe that our method can help to explore reaction space and will inspire chemists to find new reactions and synthetic ways.

How risky is it to visit a supermarket during the pandemic?

We performed large-scale numerical simulations using a composite model to investigate the infection spread in a supermarket during a pandemic. The model is composed of the social force, purchasing strategy and infection transmission models. Specifically, we quantified the infection risk for customers while in a supermarket that depended on the number of customers, the purchase strategies and the physical layout of the supermarket. The ratio of new infections compared to sales efficiency (earned profit for customer purchases) was computed as a factor of customer density and social distance. Our results indicate that the social distance between customers is the primary factor influencing infection rate. Supermarket layout and purchasing strategy do not impact social distance and hence the spread of infection. Moreover, we found only a weak dependence of sales efficiency and customer density. We believe that our study will help to establish scientifically-based safety rules that will reduce the social price of supermarket business.

Transformer-based artificial neural networks for the conversion between chemical notations.

We developed a Transformer-based artificial neural approach to translate between SMILES and IUPAC chemical notations: Struct2IUPAC and IUPAC2Struct. The overall performance level of our model is comparable to the rule-based solutions. We proved that the accuracy and speed of computations as well as the robustness of the model allow to use it in production. Our showcase demonstrates that a neural-based solution can facilitate rapid development keeping the required level of accuracy. We believe that our findings will inspire other developers to reduce development costs by replacing complex rule-based solutions with neural-based ones.

Baseline Model for Predicting Protein-Ligand Unbinding Kinetics through Machine Learning.

Derivation of structure-kinetics relationships can help rational design and development of new small-molecule drug candidates with desired residence times. Efforts are now being directed toward the development of efficient computational methods. Currently, there is a lack of solid, high-throughput binding kinetics prediction approaches on bigger datasets. We present a prediction method for binding kinetics based on the machine learning analysis of protein-ligand structural features, which can serve as a baseline for more sophisticated methods utilizing molecular dynamics (MD). We showed that the random forest algorithm is capable of learning the protein binding site secondary structure and backbone/side-chain features to predict the binding kinetics of protein-ligand complexes but still with inferior performance to that of MD-based descriptor analysis. MD simulations had been applied to a limited number of targets and a series of ligands in terms of kinetics analysis, and we believe that the developed approach may guide new studies. The method was trained on a newly curated database of 501 protein-ligand unbinding rate constants, which can also be used for testing and training the binding kinetics prediction models.

Optical Mapping-Validated Machine Learning Improves Atrial Fibrillation Driver Detection by Multi-Electrode Mapping.

BACKGROUND: Atrial fibrillation (AF) can be maintained by localized intramural reentrant drivers. However, AF driver detection by clinical surface-only multielectrode mapping (MEM) has relied on subjective interpretation of activation maps. We hypothesized that application of machine learning to electrogram frequency spectra may accurately automate driver detection by MEM and add some objectivity to the interpretation of MEM findings. METHODS: Temporally and spatially stable single AF drivers were mapped simultaneously in explanted human atria (n=11) by subsurface near-infrared optical mapping (NIOM; 0.3 mm2 resolution) and 64-electrode MEM (higher density or lower density with 3 and 9 mm2 resolution, respectively). Unipolar MEM and NIOM recordings were processed by Fourier transform analysis into 28 407 total Fourier spectra. Thirty-five features for machine learning were extracted from each Fourier spectrum. RESULTS: Targeted driver ablation and NIOM activation maps efficiently defined the center and periphery of AF driver preferential tracks and provided validated annotations for driver versus nondriver electrodes in MEM arrays. Compared with analysis of single electrogram frequency features, averaging the features from each of the 8 neighboring electrodes, significantly improved classification of AF driver electrograms. The classification metrics increased when less strict annotation, including driver periphery electrodes, were added to driver center annotation. Notably, f1-score for the binary classification of higher-density catheter data set was significantly higher than that of lower-density catheter (0.81±0.02 versus 0.66±0.04, P<0.05). The trained algorithm correctly highlighted 86% of driver regions with higher density but only 80% with lower-density MEM arrays (81% for lower-density+higher-density arrays together). CONCLUSIONS: The machine learning model pretrained on Fourier spectrum features allows efficient classification of electrograms recordings as AF driver or nondriver compared with the NIOM gold-standard. Future application of NIOM-validated machine learning approach may improve the accuracy of AF driver detection for targeted ablation treatment in patients.

Recommender Systems in Antiviral Drug Discovery.

Recommender systems (RSs), which underwent rapid development and had an enormous impact on e-commerce, have the potential to become useful tools for drug discovery. In this paper, we applied RS methods for the prediction of the antiviral activity class (active/inactive) for compounds extracted from ChEMBL. Two main RS approaches were applied: collaborative filtering (Surprise implementation) and content-based filtering (sparse-group inductive matrix completion (SGIMC) method). The effectiveness of RS approaches was investigated for prediction of antiviral activity classes ("interactions") for compounds and viruses, for which some of their interactions with other viruses or compounds are known, and for prediction of interaction profiles for new compounds. Both approaches achieved relatively good prediction quality for binary classification of individual interactions and compound profiles, as quantified by cross-validation and external validation receiver operating characteristic (ROC) score >0.9. Thus, even simple recommender systems may serve as an effective tool in antiviral drug discovery.

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.

Computational characterization of the glutamate receptor antagonist perampanel and its close analogs: density functional exploration of conformational space and molecular docking study.

Perampanel approved by FDA in 2012 is a first-in-class antiepileptic drug which inhibits α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor currents. It is markedly more active than many of its close analogs, and the reasons for this activity difference are not quite clear. Recent crystallographic studies allowed the authors to identify the location of its binding site. Unfortunately, the resolution is low, and the detailed description of perampanel binding mode is still in part speculative. Here we provide a detailed DFT-level conformational analysis of perampanel in a vacuum and in the solvents, mimicking the protein environment, followed by quantum theory of atoms in molecules (QTAIM), non-covalent interactions (NCI), and natural bond orbital (NBO) analyses. The findings indicate the electrostatic nature of the intramolecular interactions which contribute to energy differences of the conformations in a vacuum whereas the increase of dielectric constant leads to the energy equalization of conformations. Based on these results, the docking study was performed to investigate possible binding modes of perampanel and its close analogs in AMPA receptors. The influence of the pyridine nitrogen and cyano group position was explained based on the results of conformational analysis and molecular docking. These findings may contribute to the design of novel antiepileptic drugs and the development of novel approaches to treat neurodegenerative diseases and major depressive disorder.

Comparative Study of Multitask Toxicity Modeling on a Broad Chemical Space.

Acute toxicity is one of the most challenging properties to predict purely with computational methods due to its direct relationship to biological interactions. Moreover, toxicity can be represented by different end points: it can be measured for different species using different types of administration, etc., and it is questionable if the knowledge transfer between end points is possible. We performed a comparative study of prediction multitask toxicity for a broad chemical space using different descriptors and modeling algorithms and applied multitask learning for a large toxicity data set extracted from the Registry of Toxic Effects of Chemical Substances (RTECS). We demonstrated that multitask modeling provides significant improvement over single-output models and other machine learning methods. Our research reveals that multitask learning can be very useful to improve the quality of acute toxicity modeling and raises a discussion about the usage of multitask approaches for regulation purposes. Our MultiTox models are freely available in OCHEM platform ( ochem.eu/multitox ) under CC-BY-NC license.

Chemical space exploration guided by deep neural networks.

A parametric t-SNE approach based on deep feed-forward neural networks was applied to the chemical space visualization problem. It is able to retain more information than certain dimensionality reduction techniques used for this purpose (principal component analysis (PCA), multidimensional scaling (MDS)). The applicability of this method to some chemical space navigation tasks (activity cliffs and activity landscapes identification) is discussed. We created a simple web tool to illustrate our work (http://space.syntelly.com).

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

In this work, we present a new method for predicting complex physical-chemical properties of organic molecules. The approach utilizes 3D convolutional neural network (ActivNet4) that uses solvent spatial distributions around solutes as input. These spatial distributions are obtained by a molecular theory called three-dimensional reference interaction site model. We have shown that the method allows one to achieve a good accuracy of prediction of bioconcentration factor which is difficult to predict by direct application of methods of molecular theory or simulations. Our research demonstrates that combination of molecular theories with modern machine learning approaches can be effectively used for predicting properties that are otherwise inaccessible to purely theory-based models.

Interfacial structure and structural forces in mixtures of ionic liquid with a polar solvent.

Many applications of ionic liquids involve their mixtures with neutral molecular solvents. The chemical physics of these high-concentration electrolytes, in particular at interfaces, still holds many challenges. In this contribution we begin to unravel the relationship between measurements of structural ('solvation') forces in mixtures of ionic liquid with polar solvent and the corresponding structure determined by molecular dynamics simulations of the same mixtures. In order to make the quantitative link between experiments with mica surfaces and simulations with fixed-charge surfaces, we present an experimental procedure for determining the effective surface charge on mica in ionic liquid. We find that a structural cross-over recently inferred from force measurements appears to be supported by the simulations: at the cross-over, the charge-oscillatory structure switches to charge-monotonic, and solvent layering becomes dominant. Finally, we map out a phase diagram in composition-surface charge space delineating regions of charge-oscillatory interfacial structure and regions of charge-monotonic decay. We note that these features of structure and oscillatory forces are distinct from (acting simultaneously with) the recently reported longer range monotonic forces arising from anomalously long bulk screening lengths in high-concentration electrolytes.

The nanostructure of a lithium glyme solvate ionic liquid at electrified interfaces.

Solvate ionic liquids are a subclass of ionic liquids that have the potential to be used in a range of electrochemical devices. We present molecular dynamics simulations of the interfacial structure of thin films of one such lithium based solvate ionic liquid, [Li(G4)][TFSI], an equimolar solution of tetraglyme and lithium bistriflimide. This solvate ionic liquid is shown to form a novel interfacial structure at a plane electrode, which differs in a number of ways from the nanostructure observed for a conventional ionic liquid at similar interfaces. This paper explores the structural composition of the interfacial layers of this solvate ionic liquid, including their variation with surface charge, and the relation between chemical structure and interfacial arrangement.

Molecular dynamics simulation of the structure and interfacial free energy barriers of mixtures of ionic liquids and divalent salts near a graphene wall.

A molecular dynamics study of mixtures of 1-butyl-3-methylimidazolium tetrafluoroborate ([BMIm][BF4]) with magnesium tetrafluoroborate (Mg[BF4]2) confined between two parallel graphene walls is reported. The structure of the system is analyzed by means of ionic density profiles, lateral structure of the first layer close to the graphene surface and angular orientations of imidazolium cations. Free energy profiles for divalent magnesium cations are calculated using two different methods in order to evaluate the height of the potential barriers near the walls, and the results are compared with those of mixtures of the same ionic liquid and a lithium salt (Li[BF4]). Preferential adsorption of magnesium cations is analyzed using a simple model and compared to that of lithium cations, and vibrational densities of states are calculated for the cations close to the walls analyzing the influence of the graphene surface charge. Our results indicate that magnesium cations next to the graphene wall have a roughly similar environment to that in the bulk. Moreover, they face higher potential barriers and are less adsorbed on the charged graphene walls than lithium cations. In other words, magnesium cations have a more stable solvation shell than lithium ones.

Salting-out effects by pressure-corrected 3D-RISM.

We demonstrate that using a pressure corrected three-dimensional reference interaction site model one can accurately predict salting-out (Setschenow's) constants for a wide range of organic compounds in aqueous solutions of NaCl. The approach, based on classical molecular force fields, offers an alternative to more heavily parametrized methods.

Predicting Solvation Free Energies Using Parameter-Free Solvent Models.

We present a new approach for predicting solvation free energies in nonaqueous solvents. Utilizing the corresponding states principle, we estimate solvent Lennard-Jones parameters directly from their critical points. Combined with atomic solutes and the pressure corrected three-dimensional reference interaction site model (3D-RISM/PC+), the model gives accurate predictions for a wide range of nonpolar solvents, including olive oil. The results, obtained without electrostatic interactions and with a very coarse-grained solvent, provide an interesting alternative to widely used and heavily parametrized models.

Hydration Free Energies of Molecular Ions from Theory and Simulation.

We present a theoretical/computational framework for accurate calculation of hydration free energies of ionized molecular species. The method is based on a molecular theory, 3D-RISM, combined with a recently developed pressure correction (PC+). The 3D-RISM/PC+ model can provide ∼3 kcal/mol hydration free energy accuracy for a large variety of ionic compounds, provided that the Galvani potential of water is taken into account. The results are compared with direct atomistic simulations. Several methodological aspects of hydration free energy calculations for charged species are discussed.