Multi-step retrosynthesis problem can be solved by a search algorithm, such as Monte Carlo tree search (MCTS). The performance of multistep retrosynthesis, as measured by a trade-off in search time and route solvability, therefore depends on the hyperparameters of the search algorithm. In this paper, we demonstrated the effect of three MCTS hyperparameters (number of iterations, tree depth, and tree width) on metrics such as Linear integrated speed-accuracy score (LISAS) and Inverse efficiency score which consider both route solvability and search time. This exploration was conducted by employing three data-driven approaches, namely a systematic grid search, Bayesian optimization over an ensemble of molecules to obtain static MCTS hyperparameters, and a machine learning approach to dynamically predict optimal MCTS hyperparameters given an input target molecule. With the obtained results on the internal dataset, we demonstrated that it is possible to identify a hyperparameter sets which outperform the current AiZynthFinder default setting and appeared optimal across a variety of target input molecules, both on proprietary and public datasets. The settings identified with the in-house dataset reached a solvability of 93% and median search time of 151s for the in-house dataset, and a 74% solvability and 114s for the ChEMBL dataset. These numbers can be compared to the current default settings which solved 85% and 73% during a median time of 110s and 84s, for in-house and ChEMBL, respectively.

Cell-Penetrating Peptides (CPP) are emerging as an alternative to small-molecule drugs to expand the range of biomolecules that can be targeted for therapeutic purposes. Due to the importance of identifying and designing new CPP, a great variety of predictors have been developed to achieve these goals. To establish a ranking for these predictors, a couple of recent studies compared their performances on specific datasets, yet their conclusions cannot determine if the ranking obtained is due to the model, the set of descriptors or the datasets used to test the predictors. We present a systematic study of the influence of the peptide sequence's similarity of the datasets on the predictors' performance. The analysis reveals that the datasets used for training have a stronger influence on the predictors performance than the model or descriptors employed. We show that datasets with low sequence similarity between the positive and negative examples can be easily separated, and the tested classifiers showed good performance on them. On the other hand, a dataset with high sequence similarity between CPP and non-CPP will be a hard dataset, and it should be the one to be used for assessing the performance of new predictors.

Screening peptides with good affinity is an important step in peptide-drug discovery. Recent advancement in computer and data science have made machine learning a useful tool in accurately affinitive-peptide screening. In current study, four different tree-based algorithms, including Classification and regression trees(CART), C5.0 decision tree (C50), Bagged CART (BAG) and Random Forest (RF), were employed to explore the relationship between experimental peptide affinities and virtual docking data, and the performance of each model was also compared in parallel. All four algorithms showed better performances on dataset pre-scaled, -centered and -PCA than other pre-processed dataset. After model re-built and hyperparameter optimization, the optimal C50 model(C50O) showed the best performances in terms of Accuracy, Kappa, Sensitivity, Specificity, F1, MCC and AUC when validated on test data and an unknown PEDV datasets evaluation (Accuracy=80.4%). BAG and RFO (the optimal RF), as two best models during training process, did not performed as expecting during in testing and unknown dataset validations. Furthermore, the high correlation of the predictions of RFO and BAG to C50O implied the high stability and robustness of their prediction. Whereas although the good performance on unknown dataset, the poor performance in test data validation and correlation analysis indicated CARTO could not be used for future data prediction. To accurately evaluate the peptide affinity, the current study firstly gave a tree-model competition on affinitive peptide prediction by using virtual docking data, which would expand the application of machine learning algorithms in studying PepPIs and benefit the development of peptide therapeutics.

Antibiotic-resistant strains are an emerging threat to public health. The usage of antimicrobial peptides (AMPs) is one of the promising approaches to solve this problem. For the development of new AMPs, it is necessary to have reliable prediction methods. Recently, deep learning approaches have been used to predict AMP. In this paper, we want to compare simple and complex methods for these purposes. We used the BERT transformer to create sequence embeddings and the multilayer perceptron (MLP) and light attention (LA) approaches for classification. One of them reached about 80 % accuracy and specificity in benchmark testing, which is on par with the best available methods. For comparison, we proposed a simple method using only the amino acid composition of proteins or peptides. This method has shown good results, at the level of the best methods. We have prepared a special server for predicting the ability of AMPs by amino acid composition: http://bioproteom.protres.ru/antimicrob/.

Phenothiazine derivatives can unselectively inhibit the trypanothione-dependent antioxidant system enzyme trypanothione reductase (TR). A virtual screening of 2163 phenothiazine derivatives from the ZINC15 and PubChem databases docked on the active site of T. cruzi TR showed that 285 compounds have higher affinity than the natural ligand trypanothione disulfide. 244 compounds showed higher affinity toward the parasite's enzyme than to its human homolog glutathione reductase. Protein-ligand interaction profiling predicted that the main interactions for the top scored compounds were with residues important for trypanothione disulfide binding: Phe396, Pro398, Leu399, His461, Glu466, and Glu467, particularly His461, which participates in catalysis. Two compounds with the desired profiles, ZINC1033681 (Zn_C687) and ZINC10213096 (Zn_C216), decreased parasite growth by 20 % and 50 %, respectively. They behaved as mixed-type inhibitors of recombinant TR, with Ki values of 59 and 47 μM, respectively. This study provides a further understanding of the potential of phenothiazine derivatives as TR inhibitors.

Conjugated QSPR models for reactions integrate fundamental chemical laws expressed by mathematical equations with machine learning algorithms. Herein we present a methodology for building conjugated QSPR models integrated with the Arrhenius equation. Conjugated QSPR models were used to predict kinetic characteristics of cycloaddition reactions related by the Arrhenius equation: rate constant logk , pre-exponential factor logA , and activation energy Ea . They were benchmarked against single-task (individual and equation-based models) and multi-task models. In individual models, all characteristics were modeled separately, while in multi-task models logk , logA and Ea were treated cooperatively. An equation-based model assessed logk using the Arrhenius equation and logA and Ea values predicted by individual models. It has been demonstrated that the conjugated QSPR models can accurately predict the reaction rate constants at extreme temperatures, at which reaction rate constants hardly can be measured experimentally. Also, in the case of small training sets conjugated models are more robust than related single-task approaches.

QSAR models are widely and successfully used in many research areas. The success of such models highly depends on molecular descriptors typically classified as 1D, 2D, 3D, or 4D. While 3D information is likely important, e.g., for modeling ligand-protein binding, previous comparisons between the performances of 2D and 3D descriptors were inconclusive. Yet in such comparisons the modeled ligands were not necessarily represented by their bioactive conformations. With this in mind, we mined the PDB for sets of protein-ligand complexes sharing the same protein for which uniform activity data were reported. The results, totaling 461 structures spread across six series were compiled into a carefully curated, first of its kind dataset in which each ligand is represented by its bioactive conformation. Next, each set was characterized by 2D, 3D and 2D+3D descriptors and modeled using three machine learning algorithms, namely, k-Nearest Neighbors, Random Forest and Lasso Regression. Models' performances were evaluated on external test sets derived from the parent datasets either randomly or in a rational manner. We found that many more significant models were obtained when combining 2D and 3D descriptors. We attribute these improvements to the ability of 2D and 3D descriptors to code for different, yet complementary molecular properties.

In this study, the specific surface area of various perovskites was modeled using a novel quantitative read-across structure-property relationship (q-RASPR) approach, which clubs both Read-Across (RA) and quantitative structure-property relationship (QSPR) together. After optimization of the hyper-parameters, certain similarity-based error measures for each query compound were obtained. Clubbing some of these error-based measures with the previously selected features along with the Read-Across prediction function, a number of machine learning models were developed using Partial Least Squares (PLS), Ridge Regression (RR), Linear Support Vector Regression (LSVR), Random Forest (RF) regression, Gradient Boost (GBoost), Adaptive Boosting (Adaboost), Multiple Layer Perceptron (MLP) regression and k-Nearest Neighbor (kNN) regression. Based on the repeated cross-validation as well as external prediction quality and interpretability, the PLS model (nTraining  = 38, nTest  = 12, RTrain2 =0.737, QLOO2 = 0.637,  RTest2 = 0.898,  QF1(T,e,s,t)2 = 0.901) was selected as the best predictor which underscored the previously reported results. The finally selected model should efficiently predict specific surface areas of other perovskites for their use in photocatalysis. The new q-RASPR method also appears promising for the prediction of several other property endpoints of interest in materials science.

In order to analyze the Chimiothèque Nationale (CN) - The French National Compound Library - in the context of screening and biologically relevant compounds, the library was compared with ZINC in-stock collection and ChEMBL. This includes the study of chemical space coverage, physicochemical properties and Bemis-Murcko (BM) scaffold populations. More than 5 K CN-unique scaffolds (relative to ZINC and ChEMBL collections) were identified. Generative Topographic Maps (GTMs) accommodating those libraries were generated and used to compare the compound populations. Hierarchical GTM («zooming») was applied to generate an ensemble of maps at various resolution levels, from global overview to precise mapping of individual structures. The respective maps were added to the ChemSpace Atlas website. The analysis of synthetic accessibility in the context of combinatorial chemistry showed that only 29,7 % of CN compounds can be fully synthesized using commercially available building blocks.

Cooperative molecular contacts play an important role in protein structure and ligand binding. Here, we constructed a PostgreSQL database that stores structural information in the form of atomic environments and allows flexible mining of molecular contacts. Taking the Ser-His-Asp/Glu catalytic triad as a first test case, we demonstrate that the presence of a carboxylate oxygen atom in the vicinity of a His is associated with shorter Ser-OH..N-His bond in the PDB30 subset. We prospectively mine catalytic triads in unannotated proteins, suggesting catalytic functions for unannotated proteins. As a second test case, we demonstrate that this database system can include ligand atoms, represented by Sybyl atom types, by evaluating the proportion of counter-ions for ligand carboxylate oxygens.

Identification of novel chemotypes with biological activity similar to a known active molecule is an important challenge in drug discovery called 'scaffold hopping'. Small-, medium-, and large-step scaffold hopping efforts may lead to increasing degrees of chemical structure novelty with respect to the parent compound. In the present paper, we focus on the problem of large-step scaffold hopping. We assembled a high quality and well characterized dataset of scaffold hopping examples comprising pairs of active molecules and including a variety of protein targets. This dataset was used to build a benchmark corresponding to the setting of real-life applications: one active molecule is known, and the second active is searched among a set of decoys chosen in a way to avoid statistical bias. This allowed us to evaluate the performance of computational methods for solving large-step scaffold hopping problems. In particular, we assessed how difficult these problems are, particularly for classical 2D and 3D ligand-based methods. We also showed that a machine-learning chemogenomic algorithm outperforms classical methods and we provided some useful hints for future improvements.

The main protease (Mpro ) is an essential enzyme for the life cycle of SARS-CoV-2 and a validated target for treatment of COVID-19 infection. Structure-based pharmacophore modeling combined with QSAR calculations were employed to identify new chemical scaffolds of Mpro inhibitors from natural products repository. Hundreds of pharmacophore models were manually built from their corresponding X-ray crystallographic structures. A pharmacophore model that was validated by receiver operating characteristic (ROC) curve analysis and selected using the statistically optimum QSAR equation was implemented as a 3D-search tool to mine AnalytiCon Discovery database of natural products. Captured hits that showed the highest predicted inhibitory activities were bioassayed. Three active Mpro inhibitors (pseurotin A, lactupicrin, and alpinetin) were successfully identified with IC50 values in low micromolar range.

Graph generative models have recently emerged as an interesting approach to construct molecular structures atom-by-atom or fragment-by-fragment. In this study, we adopt the fragment-based strategy and decompose each input molecule into a set of small chemical fragments. In drug discovery, a few drug molecules are designed by replacing certain chemical substituents with their bioisosteres or alternative chemical moieties. This inspires us to group decomposed fragments into different fragment clusters according to their local structural environment around bond-breaking positions. In this way, an input structure can be transformed into an equivalent three-layer graph, in which individual atoms, decomposed fragments, or obtained fragment clusters act as graph nodes at each corresponding layer. We further implement a prototype model, named multi-resolution graph variational autoencoder (MRGVAE), to learn embeddings of constituted nodes at each layer in a fine-to-coarse order. Our decoder adopts a similar but conversely hierarchical structure. It first predicts the next possible fragment cluster, then samples an exact fragment structure out of the determined fragment cluster, and sequentially attaches it to the preceding chemical moiety. Our proposed approach demonstrates comparatively good performance in molecular evaluation metrics compared with several other graph-based molecular generative models. The introduction of the additional fragment cluster graph layer will hopefully increase the odds of assembling new chemical moieties absent in the original training set and enhance their structural diversity. We hope that our prototyping work will inspire more creative research to explore the possibility of incorporating different kinds of chemical domain knowledge into a similar multi-resolution neural network architecture.

In recent years, thanks to advances in computer hardware and dataset availability, data-driven approaches (like machine learning) have become one of the essential parts of the drug design framework to accelerate drug discovery procedures. Constructing a new scoring function, a function that can predict the binding score for a generated protein-ligand pose during docking procedure or a crystal complex, based on machine and deep learning has become an active research area in computer-aided drug design. GB-Score is a state-of-the-art machine learning-based scoring function that utilizes distance-weighted interatomic contact features, PDBbind-v2019 general set, and Gradient Boosting Trees algorithm to the binding affinity prediction. The distance-weighted interatomic contact featurization method used the distance between different ligand and protein atom types for numerical representation of the protein-ligand complex. GB-Score attains Pearson's correlation 0.862 and RMSE 1.190 on the CASF-2016 benchmark test in the scoring power metric. GB-Score's codes are freely available on the web at https://github.com/miladrayka/GB_Score.

The toxicity of compounds is closely related to the effectiveness and safety of drug development, and accurately predicting the toxicity of compounds is one of the most challenging tasks in medicinal chemistry and pharmacology. In this paper, we construct three types of models for single and multi-tasking based on 2D and 3D descriptors, fingerprints and molecular graphs, and then validate the models with benchmark tests on the Tox21 data challenge. We found that due to the information sharing mechanism of multi-task learning, it could address the imbalance problem of the Tox21 data sets to some extent, and the prediction performance of the multi-task was significantly improved compared with the single task in general. Given the complement of the different molecular representations and modeling algorithms, we attempted to integrate them into a robust Co-Model. Our Co-Model performs well in various evaluation metrics on the test set and also achieves significant performance improvement compared to other models in the literature, which clearly demonstrates its superior predictive power and robustness.

This study examines how two popular drug-likeness concepts used in early development, Lipinski Rule of Five (Ro5) and Veber's Rules, possibly affected drug profiles of FDA approved drugs since 1997. Our findings suggest that when all criteria are applied, relevant compounds may be excluded, addressing the harmfulness of blindly employing these rules. Of all oral drugs in the period used for this analysis, around 66 % conform to the RO5 and 85 % to Veber's Rules. Molecular Weight and calculated LogP showed low consistent values over time, apart from being the two least followed rules, challenging their relevance. On the other hand, hydrogen bond related rules and the number of rotatable bonds are amongst the most followed criteria and show exceptional consistency over time. Furthermore, our analysis indicates that topological polar surface area and total count of hydrogen bonds cannot be used as interchangeable parameters, contrary to the original proposal. This research enhances the comprehension of drug profiles that were FDA approved in the post-Lipinski period. Medicinal chemists could utilize these heuristics as a limited guide to direct their exploration of the oral bioavailability chemical space, but they must also steer the wheel to break these rules and explore different regions when necessary.

BACKGROUND: Despite tremendous efforts made by scientific community during the outbreak of COVID-19 pandemic, this disease still remains as a public health concern. Although different types of vaccines were globally used to reduce the mortality, emergence of new variants of SARS-CoV-2 is a challenging issue in COVID-19 pharmacotherapy. In this context, target therapy of SARS-CoV-2 by small ligands is a promising strategy. METHODS: In this investigation, we applied ligand-based virtual screening for finding novel molecules based on nirmatrelvir structure. Various criteria including drug-likeness, ADME, and toxicity properties were applied for filtering the compounds. The selected candidate molecules were subjected to molecular docking and dynamics simulation for predicting the binding mode and binding free energy, respectively. Then the molecules were experimentally evaluated in terms of antiviral activity against SARS-CoV-2 and toxicity assessment. RESULTS: The results demonstrated that the identified compounds showed inhibitory activity towards SARS-CoV-2 Mpro . CONCLUSION: In summary, the introduced compounds may provide novel scaffold for further structural modification and optimization with improved anti SARS-CoV-2 Mpro activity.

SARS-CoV-2 gained crucial mutations at the receptor binding domain (RBD) that often changed the course of the pandemic leading to new waves with increased case fatality. Variants are observed with enhanced transmission and immune invasion abilities. Thus, predicting future variants with enhanced transmission ability is a problem of utmost research interest. Here, we have developed a multi-tier exhaustive SARS-CoV-2 mutation screening platform combining MM/GBSA, extensive molecular dynamics simulations, and steered molecular dynamics to identify RBD mutants with enhanced ACE2 binding capability. We have identified four RBM mutations (F490K, S494K, G504F, and the P499L) with significantly higher ACE2 binding abilities than wild-type RBD. Compared to wild-type RBD, they all form stable complexes with more hydrogen bonds and salt-bridge interactions with ACE2. Our simulation data suggest that these mutations allosterically alter the packing of the RBM interface of the RBD-ACE2 complex. As a result, the rupture force required to break the RBD-ACE2 contacts is significantly higher for these mutants.

AutoDock is a widely used software for flexible ligand docking problems since it is open source and easy to be implemented. In this paper, a novel hybrid algorithm is proposed and applied in the docking environment of AutoDock version 4.2.6 in order to enhance the accuracy and the efficiency for dockings with flexible ligands. This search algorithm, called entropy-based Lamarckian quantum-behaved particle swarm optimization (ELQPSO), is a combination of the QPSO with an entropy-based update strategy and the Solis and Wet local search (SWLS) method. By using the PDBbind core set v.2016, the ELQPSO is compared with the Lamarckian genetic algorithm (LGA), Lamarckian particle swarm optimization (LPSO) and Lamarckian QPSO (LQPSO). The experimental results reveal that the corresponding docking program of ELQPSO, named as EQDOCK in this paper, has a competitive performance in dealing with the protein-ligand docking problems. Moreover, for the test cases with different number of torsions, the EQDOCK outperforms the other three docking programs in finding docking conformations with small root mean squared deviation (RMSD) values in most cases. In particular, it has an advantage of solving highly flexible ligand docking problems over the others.

Computational drug repurposing aims to discover new treatment regimens by analyzing approved drugs on the market. This study proposes previously approved compounds that can change the expression profile of disease-causing proteins by developing a network theory-based drug repurposing approach. The novelty of the proposed approach is an exploration of module similarity between a disease-causing network and a compound-specific interaction network; thus, such an association leads to more realistic modeling of molecular cell responses at a system biology level. The overlap of the disease network and each compound-specific network is calculated based on a shortest-path similarity of networks by accounting for all protein pairs between networks. A higher similarity score indicates a significant potential of a compound. The approach was validated for breast and lung cancers. When all compounds are sorted by their normalized-similarity scores, 36 and 16 drugs are proposed as new candidates for breast and lung cancer treatment, respectively. A literature survey on candidate compounds revealed that some of our predictions have been clinically investigated in phase II/III trials for the treatment of two cancer types. As a summary, the proposed approach has provided promising initial results by modeling biochemical cell responses in a network-level data representation.