Machine learning-assisted search for novel coagulants: When machine learning can be efficient even if data availability is low.

Design of new drugs is a challenging process: a candidate molecule should satisfy multiple conditions to act properly and make the least side-effect-perfect candidates selectively attach to and influence only targets, leaving off-targets intact. The amount of experimental data about various properties of molecules constantly grows, promoting data-driven approaches. However, the applicability of typical predictive machine learning techniques can be substantially limited by a lack of experimental data about a particular target. For example, there are many known Thrombin inhibitors (acting as anticoagulants), but a very limited number of known Protein C inhibitors (coagulants). In this study, we present our approach to suggest new inhibitor candidates by building an effective representation of chemical space. For this aim, we developed a deep learning model-autoencoder, trained on a large set of molecules in the SMILES format to map the chemical space. Further, we applied different sampling strategies to generate novel coagulant candidates. Symmetrically, we tested our approach on anticoagulant candidates, where we were able to predict their inhibition towards Thrombin. We also compare our approach with MegaMolBART-another deep learning generative model, but exploiting similar principles of navigation in a chemical space.

Toward a quasiphase transition in the single-file chain of water molecules: Simple lattice model.

Recently, Ma et al. [Phys. Rev. Lett. 118, 027402 (2017)] have suggested that water molecules encapsulated in (6,5) single-wall carbon nanotube experience a temperature-induced quasiphase transition around 150 K interpreted as changes in the water dipoles orientation. We discuss further this temperature-driven quasiphase transition performing quantum chemical calculations and molecular dynamics simulations and, most importantly, suggesting a simple lattice model to reproduce the properties of the one-dimensional confined finite arrays of water molecules. The lattice model takes into account not only the short-range and long-range interactions but also the rotations in a narrow tube, and both ingredients provide an explanation for a temperature-driven orientational ordering of the water molecules, which persists within a relatively wide temperature range.

Machine learning assessment of the binding region as a tool for more efficient computational receptor-ligand docking.

We present a combined computational approach to protein-ligand binding, which consists of two steps: (1) a deep neural network is used to locate a binding region on a target protein, and (2) molecular docking of a ligand is performed within the specified region to obtain the best pose using Autodock Vina. Our in-house designed neural network was trained using the PepBDB dataset. Although the training dataset consisted of protein-peptide complexes, we show that the approach is not limited to peptides, but also works remarkably well for a large class of non-peptide ligands. The results are compared with those in which the binding region (first step) was provided by Accluster. In cases where no prior experimental data on the binding region are available, our deep neural network provides a fast and effective alternative to classical software for its localization. Our code is available at https://github.com/mksmd/NNforDocking.

Complex machine learning model needs complex testing: Examining predictability of molecular binding affinity by a graph neural network.

Drug discovery pipelines typically involve high-throughput screening of large amounts of compounds in a search of potential drugs candidates. As a chemical space of small organic molecules is huge, a "navigation" over it urges for fast and lightweight computational methods, thus promoting machine-learning approaches for processing huge pools of candidates. In this contribution, we present a graph-based deep neural network for prediction of protein-drug binding affinity and assess its predictive power under thorough testing conditions. Within the suggested approach, both protein and drug molecules are represented as graphs and passed to separate graph sub-networks, then concatenated and regressed towards a binding affinity. The neural network is trained on two binding affinity datasets-PDBbind and data imported from RCSB Protein Data Bank. In order to explore the generalization capabilities of the model we go beyond traditional random or leave-cluster-out techniques and demonstrate the need for more elaborate model performance assessment - six different strategies for test/train data partitioning (random, time- and property-arranged, protein- and ligand-clustered) with a k-fold cross-validation are engaged. Finally, we discuss the model performance in terms of a set of metrics for different split strategies and fold arrangement. Our code is available at https://github.com/SoftServeInc/affinity-by-GNN.

Ensembling machine learning models to boost molecular affinity prediction.

This study unites six popular machine learning approaches to enhance the prediction of a molecular binding affinity between receptors (large protein molecules) and ligands (small organic molecules). Here we examine a scheme where affinity of ligands is predicted against a single receptor - human thrombin, thus, the models consider ligand features only. However, the suggested approach can be repurposed for other receptors. The methods include Support Vector Machine, Random Forest, CatBoost, feed-forward neural network, graph neural network, and Bidirectional Encoder Representations from Transformers. The first five methods use input features based on physico-chemical properties of molecules, while the last one is based on textual molecular representations. All approaches do not rely on atomic spatial coordinates, avoiding a potential bias from known structures, and are capable of generalizing for compounds with unknown conformations. Within each of the methods, we have trained two models that solve classification and regression tasks. Then, all models are grouped into a pipeline of two subsequent ensembles. The first ensemble aggregates six classification models which vote whether a ligand binds to a receptor or not. If a ligand is classified as active (i.e., binds), the second ensemble predicts its binding affinity in terms of the inhibition constant Ki.

Toward efficient generation, correction, and properties control of unique drug-like structures.

Efficient design and screening of the novel molecules is a major challenge in drug and material design. This paper focuses on a multi-stage pipeline, in which several deep neural network models are combined to map discrete molecular representations into continuous vector space to later generate from it new molecular structures with desired properties. Here, the Attention-based Sequence-to-Sequence model is added to "spellcheck" and correct generated structures, while the oversampling in the continuous space allows generating candidate structures with desired distribution for properties and molecular descriptors, even for a small reference datasets. We further use computer simulation to validate the desired properties in the numerical experiment. With the focus on the drug design, such a pipeline allows generating novel structures with a control of Synthetic Accessibility Score and a series of metrics that assess the drug-likeliness. Our code is available at https://github.com/SoftServeInc/novel-molecule-generation.

Explicit water molecular dynamics study of the mobility of halide ions in presence of ionene oligocations.

We present explicit water molecular dynamics simulations of solutions of aliphatic 3,3- and 6,6-ionene oligocations neutralized with (i) fluoride, chloride, bromide, or iodide counterions, respectively, or (ii) with a 1:1 mixture of chloride and bromide anions in presence of a low molecular weight salt at 298 K. The SPC/E model was used to describe water molecules. Results of the simulation are presented in form of the pair distribution functions between various atoms on the ionene oligoion and counterions in solution. In addition, we were interested in the dynamics of counterions around model ionenes. We showed that counterions residing in the vicinity of the oligoion exchange rapidly with those in the bulk solution, with the frequency depending on the nature of the counterion and on the charge density of the oligoion. We calculated the average residence times of the various counterion species to the oligoions and proposed the model which divides the counterions into "free" and "bound" and calculated the fraction of "free" counterions. In the second part of the study, we investigated interaction of the sodium chloride and sodium bromide, being simultaneously present in the solution, with differently charged ionenes in water. The selectivity effect was clearly observed: bromide ions tend to replace chloride ions in the immediate vicinity of the ionene oligoions. Simulation results are discussed in light of our recent measurements of thermodynamic and transport properties of aqueous ionene solutions.

Isothermal titration calorimetry and molecular dynamics study of ion-selectivity in mixtures of hydrophobic polyelectrolytes with sodium halides in water.

Aliphatic x,y-ionenes are polyelectrolytes in which x and y denote the numbers of methylene groups separating quaternary ammonium ions. They represent useful model substances for studying hydrophobic and charge effects in aqueous solutions. We used isothermal titration calorimetry to measure the enthalpies of mixing, ΔH(mix), of 3,3- and 6,6-ionene fluorides and bromides with low molecular weight salts (NaF, NaCl, NaBr, and NaI) at 298 K in water. The signs and magnitudes of the measured enthalpies depend on the hydrophobicity of the ionene and on the nature of the added salt. For example, addition of sodium fluoride to solutions of 3,3- and 6,6-ionene fluorides produced endothermic effects, while addition of sodium bromide to 3,3-ionene bromide resulted in a strong exothermic effect. Interestingly, mixing of 6,6-ionene bromide and NaBr solutions in water gave a small exothermic heat effect. Polyelectrolyte theories, based on continuum-solvent models, predict enthalpies of mixing to be positive (endothermic) for all the solutions examined in this work. The ion-specific effect is more strongly expressed in ionene solutions with higher charge density (3,3-ionene). The most important result of this work is the finding that the enthalpy of mixing of 3,3- (and of 6,6-ionene) fluorides with sodium halides can be expressed as a linear function of the enthalpy of hydration of the halide counterions. The experimental results were complemented with an explicit water molecular dynamics simulation of solutions of oligoions modelling 3,3- and 6,6-ionenes. The computer simulation results for various nitrogen-counterion pair distribution functions were in most cases consistent with the enthalpy measurements.