Virtual Screening Models Research Articles

A Combination of Pharmacophore-Based Virtual Screening, Structure-Based Lead Optimization, and DFT Study for the Identification of S. epidermidis TcaR Inhibitors.

The transcriptional regulator (TcaR) enzyme plays an important role in biofilm formation. Prevention of TcaR-DNA complex formation leads to inhibit the biofilm formation is likely to reveal therapeutic ways for the treatment of bacterial infections. To identify the novel ligands for TcaR and to provide a new idea for drug design, two efficient drug design methods, such as pharmacophore modeling and structure-based drug design, were used for virtual screening of database and lead optimization, respectively. Gemifloxacin (FDA-approved drug) was considered to generate the pharmacophore model for virtual screening of the ZINC database, and five hits, namely ZINC77906236, ZINC09550296, ZINC77906466, ZINC09751390, and ZINC01269201, were identified as novel inhibitors of TcaR with better binding energies. Using structure-based drug design, a set of 7a–7p inhibitors of S. epidermidis were considered, and Mol34 was identified with good binding energy and high fitness score with improved pharmacological properties. The active site residues ARG110, ASN20, HIS42, ASN45, ALA38, VAL63, VAL68, ALA24, VAL43, ILE57, and ARG71 are playing a promising role in inhibition process. In addition, we performed DFT simulations of final hits to understand the electronic properties and their significant role in driving the inhibitor to adopt apposite bioactive conformations in the active site. Conclusively, the newly identified and designed hits from both the methods are promising inhibitors of TcaR, which can hinder biofilm formation.

Estrogen receptor potentially stable conformations from molecular dynamics as a structure-based pharmacophore model for mapping, screening, and identifying ligands—a new paradigm shift in pharmacophore screening

Despite rigorous research on breast cancer has increased in recent decades, only few drugs are in practice to combat against the disease. Due to excessive usage, these drugs attain resistance is an avertable phenomenon resulting from inadequate treatment. A novel, and real-time approaches are expected to overcome to find the solution for the drug resistance. The molecular dynamics based multi-conformational sampling technique via computer-aided drug-designing approach, may be a promising route to identify the lead candidates from real-time generated frames. The estrogenic receptor, being one of the most widely targeted receptors for various breast cancer drugs namely, tamoxifen, raloxifene and GW5 (tamoxifen-resistance inhibitor) was used for simulating the molecular dynamics to obtain various real time frames. The energetically stable frames were funnelled based on Gibbs free binding energy, interaction energy and active site interaction to generate pharmacophores model for virtual screening of compounds. Generated pharmacophores are validated by receiver operating characteristic area under curve greater than 0.8. Further, screening of compounds with validated structure-based pharmacophore model of different estrogen bound drug complex conformations and binding orientations are complement for tamoxifen and tamoxifen-resistance inhibitor frames. Moreover, the best mapped compounds were docked and probed for ADMET, TopKat® and Lipinski’s rule of five is more favourable for compound Andrographidine F sourced from medicinal herbal plant Andrographis paniculata. Hence, this compound had to be further analysed in in-vitro and in-vivo to prove the same. Communicated by Ramaswamy H. Sarma

Insect RDL Receptor Models for Virtual Screening: Impact of the Template Conformational State in Pentameric Ligand-Gated Ion Channels.

The RDL receptor is one of the most relevant protein targets for insecticide molecules. It belongs to the pentameric ligand-gated ion channel (pLGIC) family. Given that the experimental structures of pLGICs are difficult to obtain, homology modeling has been extensively used for these proteins, particularly for the RDL receptor. However, no detailed assessments of the usefulness of homology models for virtual screening (VS) have been carried out for pLGICs. The aim of this study was to evaluate which are the determinant factors for a good VS performance using RDL homology models, specially analyzing the impact of the template conformational state. Fifteen RDL homology models were obtained based on different pLGIC templates representing the closed, open, and desensitized states. A retrospective VS process was performed on each model, and their performance in the prioritization of active ligands was assessed. In addition, the three best-performing models among each of the conformations were subjected to molecular dynamics simulations (MDS) in complex with a representative active ligand. The models showed variations in their VS performance parameters that were related to the structural properties of the binding site. VS performance tended to improve in more constricted binding cavities. The best performance was obtained with a model based on a template in the closed conformation. MDS confirmed that the closed model was the one that best represented the interactions with an active ligand. These results imply that different templates should be evaluated and the structural variations between their channel conformational states should be specially examined, providing guidelines for the application of homology modeling for VS in other proteins of the pLGIC family.

Screening and bioavailability evaluation of anti-oxidative selenium-containing peptides from soybeans based on specific structures.

Our previous study has evaluated the antioxidant capacity and identified the sequences of soybean selenium-containing peptides. Herein, pharmacophore screening, gastrointestinal simulation and in vivo pharmacokinetics were performed to predict the potentials of selenium-containing peptides in terms of antioxidant activity, safety and bioavailability. A pharmacophore model with 6 structure features was constructed for virtual screening to determine the potential activities of 85 selenium sequences from soybean peptides. Strong reversing effects (p < 0.05) of the targeted sequences were observed in tumor necrosis factor-α (TNF-α)-induced inflammatory cytokines and adhesion factors burst in EA·hy926/Caco-2 co-culture cell models. Ser-Phe-Gln-SeMet (SFQSeM), a promising peptide selected from both virtual screening and cell models, was proved to be stable in the gastrointestinal tract and could be transported across the Caco-2 monolayer via the paracellular pathway. Additionally, SFQSeM showed a long residence time (89.42 ± 1.34 min) and half-life (81.60 ± 11.88 min) after consumption, and it induced lower liver alanine/aspartate transaminase (ALT/AST) and serum nitric oxide (NO) levels compared to Na2SeO3 and SeMet (p < 0.05). The potency of SFQSeM against oxidative stress as well as its oral bioavailability and low risk highlight its potential utility as an effective Se nutritional supplement.

Active Site Sequence Representations of Human Kinases Outperform Full Sequence Representations for Affinity Prediction and Inhibitor Generation: 3D Effects in a 1D Model.

Recent advances in deep learning have enabled the development of large-scale multimodal models for virtual screening and de novo molecular design. The human kinome with its abundant sequence and inhibitor data presents an attractive opportunity to develop proteochemometric models that exploit the size and internal diversity of this family of targets. Here, we challenge a standard practice in sequence-based affinity prediction models: instead of leveraging the full primary structure of proteins, each target is represented by a sequence of 29 discontiguous residues defining the ATP binding site. In kinase-ligand binding affinity prediction, our results show that the reduced active site sequence representation is not only computationally more efficient but consistently yields significantly higher performance than the full primary structure. This trend persists across different models, data sets, and performance metrics and holds true when predicting pIC50 for both unseen ligands and kinases. Our interpretability analysis reveals a potential explanation for the superiority of the active site models: whereas only mild statistical effects about the extraction of three-dimensional (3D) interaction sites take place in the full sequence models, the active site models are equipped with an implicit but strong inductive bias about the 3D structure stemming from the discontiguity of the active sites. Moreover, in direct comparisons, our models perform similarly or better than previous state-of-the-art approaches in affinity prediction. We then investigate a de novo molecular design task and find that the active site provides benefits in the computational efficiency, but otherwise, both kinase representations yield similar optimized affinities (for both SMILES- and SELFIES-based molecular generators). Our work challenges the assumption that the full primary structure is indispensable for modeling human kinases.

QSAR and deep learning model for virtual screening of potential inhibitors against Inosine 5’ Monophosphate dehydrogenase (IMPDH) of Cryptosporidium parvum

Cryptosporidium parvum (Cp) causes a gastro-intestinal disease called Cryptosporidiosis. C. parvum Inosine 5’ monophosphate dehydrogenase (CpIMPDH) is responsible for the production of guanine nucleotides. In the present study, 37 known urea-based congeneric compounds were used to build a 2D and 3D QSAR model against CpIMPDH. The built models were validated based on OECD principles. A deep learning model was adopted from a framework called Deep Purpose. The model was trained with 288 known active compounds and validated using a test set. From the training set of the 3D QSAR, a pharmacophore model was built and the best pharmacophore hypotheses were scored and sorted using a phase-hypo score. A phytochemical database was screened using both the pharmacophore model and a deep learning model. The screened compounds were considered for glide XP docking, followed by quantum polarized ligand docking. Finally, the best compound among them was considered for molecular dynamics simulation study.

Trends in Deep Learning for Property-driven Drug Design.

It is more pressing than ever to reduce the time and costs for the development of lead compounds in the pharmaceutical industry. The co-occurrence of advances in high-throughput screening and the rise of deep learning (DL) have enabled the development of large-scale multimodal predictive models for virtual drug screening. Recently, deep generative models have emerged as a powerful tool to explore the chemical space and raise hopes to expedite the drug discovery process. Following this progress in chemocentric approaches for generative chemistry, the next challenge is to build multimodal conditional generative models that leverage disparate knowledge sources when mapping biochemical properties to target structures. Here, we call the community to bridge drug discovery more closely with systems biology when designing deep generative models. Complementing the plethora of reviews on the role of DL in chemoinformatics, we specifically focus on the interface of predictive and generative modelling for drug discovery. Through a systematic publication keyword search on PubMed and a selection of preprint servers (arXiv, biorXiv, chemRxiv, and medRxiv), we quantify trends in the field and find that molecular graphs and VAEs have become the most widely adopted molecular representations and architectures in generative models, respectively. We discuss progress on DL for toxicity, drug-target affinity, and drug sensitivity prediction and specifically focus on conditional molecular generative models that encompass multimodal prediction models. Moreover, we outline future prospects in the field and identify challenges such as the integration of deep learning systems into experimental workflows in a closed-loop manner or the adoption of federated machine learning techniques to overcome data sharing barriers. Other challenges include, but are not limited to interpretability in generative models, more sophisticated metrics for the evaluation of molecular generative models, and, following up on that, community-accepted benchmarks for both multimodal drug property prediction and property-driven molecular design.

Structure-based virtual screening of CYP1A1 inhibitors: towards rapid tier-one assessment of potential developmental toxicants

Cytochrome P450 1A1 (CYP1A1) metabolizes estrogens, melatonin, and other key endogenous signaling molecules critical for embryonic/fetal development. The enzyme has increasing expression during pregnancy, and its inhibition or knockout increases embryonic/fetal lethality and/or developmental problems. Here, we present a virtual screening model for CYP1A1 inhibitors based on the orthosteric and predicted allosteric sites of the enzyme. Using 1001 reference compounds with CYP1A1 activity data, we optimized the decision thresholds of our model and classified the training compounds with 68.3% balanced accuracy (91.0% sensitivity and 45.7% specificity). We applied our final model to 11 known CYP1A1 orthosteric binders and related compounds, and found that our ranking of the known orthosteric binders generally agrees with the relative activity of CYP1A1 in metabolizing these compounds. We also applied the model to 22 new test compounds with unknown/unclear CYP1A1 inhibitory activity, and predicted 16 of them are CYP1A1 inhibitors. The CYP1A1 potency and modes of inhibition of these 22 compounds were experimentally determined. We confirmed that most predicted inhibitors, including drugs contraindicated during pregnancy (amiodarone, bicalutamide, cyproterone acetate, ketoconazole, and tamoxifen) and environmental agents suspected to be endocrine disruptors (bisphenol A, diethyl and dibutyl phthalates, and zearalenone), are indeed potent inhibitors of CYP1A1. Our results suggest that virtual screening may be used as a rapid tier-one method to screen for potential CYP1A1 inhibitors, and flag them out for further experimental evaluations.

Ammonium hexafluorosilicates as potential anti-caries agents: the problem of cation selection

In the last decade, ammonium hexafluorosilicate (AHFS) and ammonium hexafluorosilicates with biologically active cations (AHBAC), which have certain advantages over traditional fluoride medicinal substances, have been actively studied as anti-caries agents. In particular, an important feature of the action of AHFS is its ability to cause prolonged occlusion of the dentinal tubules with a precipitate of calcium fluoride; when using AHBAС there is a possibility of strengthening the anti-caries activity of the substance due to the pharmacological potential of the cation.&#x0D; The purpose of the review is to analyze the effects of the cation on the physicochemical properties and biological activity of ammonium hexafluorosilicates as potential anti-caries agents.&#x0D; Research methods – bibliosemantic, content analysis.&#x0D; It was drew the attention to the peculiarity of the AHBAC structure: salt structures are formed on the basis of systems of strong interionic H-bonds, mainly of the NH···F type, which have a significant effect on the properties of hexafluorosilicates in the crystalline state and their behavior in solutions. It was demonstrated the non-trivial nature of the change of solubility in water of AHBAC with heterocyclic and aromatic cations, which consists in the decrease of solubility with increasing number of hydrophilic fragments in the structure of cations. Adequate 2D QSPR models for interpretation and virtual screening of AHBAC water solubility have been constructed. Accounting for the effect of H-bonds on the solubility of AHBAC was detailed. It was assumed that the process of hydrolysis of AHBAC in aqueous solutions can be stimulated by elongation of the Si–F anion bonds due to the effects of H-bonds. It is shown that the thermal stability of AHBAC with pyridinium cations symbatically correlates with the number of strong and medium H-bonds in salt structures.&#x0D; The action of the pharmacological effects of the cation on the biological activity of AHBAC is manifested in the form of an increase in the caries-prophylactic efficacy of AHBAC in comparison with a similar effect of AHFS. Attempts to establish a relationship between the anti-caries activity of AHBAC and a certain pharmacological action of the cation have led to mixed results. This obviously reflects the complex mechanism of the influence of the biological activity of the cation on the caries-prophylactic efficacy of hexafluorosilicates, which is not limited to any one, albeit dominant, type of activity.

Discovery of novel selective PI3Kγ inhibitors through combining machine learning-based virtual screening with multiple protein structures and bio-evaluation

IntroductionPhosphoinositide 3-kinase gamma (PI3Kγ) has been regarded as a promising drug target for the treatment of various diseases, and the diverse physiological roles of class I PI3K isoforms (α, β, δ, and γ) highlight the importance of isoform selectivity in the development of PI3Kγ inhibitors. However, the high structural conservation among the PI3K family makes it a big challenge to develop selective PI3Kγ inhibitors. ObjectivesA novel machine learning-based virtual screening with multiple PI3Kγ protein structures was developed to discover novel PI3Kγ inhibitors. MethodsA large chemical database was screened using the virtual screening model, the top-ranked compounds were then subjected to a series of bio-evaluations, which led to the discovery of JN-KI3. The selective inhibition mechanism of JN-KI3 against PI3Kγ was uncovered by a theoretical study. Results49 hits were identified through virtual screening, and the cell-free enzymatic studies found that JN-KI3 selectively inhibited PI3Kγ at a concentration as low as 3,873 nM but had no inhibitory effect on Class IA PI3Ks, leading to the selective cytotoxicity on hematologic cancer cells. Meanwhile, JN-KI3 potently blocked the PI3K signaling, finally led to distinct apoptosis of hematologic cell lines at a low concentration. Lastly, the key residues of PI3Kγ and the structural characteristics of JN-KI3, which both would influence γ isoform-selective inhibition, were highlighted by systematic theoretical studies. ConclusionThe developed virtual screening model strongly manifests the robustness to find novel PI3Kγ inhibitors. JN-KI3 displays a specific cytotoxicity on hematologic tumor cells, and significantly promotes apoptosis associated with the inhibition of the PI3K signaling, which depicts PI3Kγ as a potential target for the hematologic tumor therapy. The theoretical results reveal that those key residues interacting with JN-KI3 are less common compared to most of the reported PI3Kγ inhibitors, indicating that JN-KI3 has novel structural characteristics as a selective PIK3γ inhibitor.

Artificial neural network‐based quantitative structure–activity relationships model and molecular docking for virtual screening of novel potent acetylcholinesterase inhibitors

AbstractThe deficit in cholinergic neurotransmission is the main incentives for the development of new therapeutic drugs for the treatment of Alzheimer's disease (AD). In this study, we employed cheminformatics tools to explore new molecules that can serve as effective therapeutic agents against AD. Artificial neural networks (ANNs) were applied in a quantitative structure anti‐acetylcholinesterase (AChE)‐activity relationship study on a series of AChE inhibitors (AChEIs). The best computational neural network had an [5‐10‐12‐1] architecture, with a low value of mean squared error (MSE = 0.06) and a high value of R2 (0.96). All validations showed that the ANN model can be used quite satisfactorily for the screening of a new series of molecules having anti‐AChE activity. The virtual screening based on the molecular similarity method and applicability domain of ANN‐quantitative structure–activity relationships allowed the discovery of novel anti‐AChE candidates with improved activity. Docking simulation carried out on these novel AChEIs has identified eight best hits with a higher binding affinity toward their target (4EY7). These eight stable complexes were in good agreement with the biological activity and they have an inhibition profile at a similar rate as the reference drug donepezil. The results showed that these compounds were strongly bound up with the AChE enzyme active site with the optimal conformations.

Deep learning model for virtual screening of novel 3C-like protease enzyme inhibitors against SARS coronavirus diseases

In the context of the recently emerging COVID-19 pandemic, we developed a deep learning model that can be used to predict the inhibitory activity of 3CLpro in severe acute respiratory syndrome coronavirus (SARS-CoV) for unknown compounds during the virtual screening process. This paper proposes a novel deep learning-based method to implement virtual screening with convolutional neural network (CNN) architecture. The descriptors represent chemical molecules, and these descriptors are input into the CNN framework to train a model and predict active compounds. When compared to other machine learning methods, including random forest, naive Bayes, decision tree, and support vector machine, the proposed CNN model's evaluation of the test set showed an accuracy of 0.86, a sensitivity of 0.45, a specificity of 0.96, a precision of 0.73, a recall of 0.45, an F-measure of 0.55, and a ROC of 0.71. The CNN model screened 17 out of 918 phytochemical compounds; 60 out of 423 from the natural product NCI divset IV; 17,831 out of 112,267 from the ZINC natural product database; and 315 out of 1556 FDA-approved drugs as anti-SARS-CoV agents. Further, to prioritize drug-like compounds, Lipinski's rule of five was applied to screen anti-SARS-CoV compounds (excluding FDA-approved drugs), resulting in 10, 59, and 14,025 hit molecules. Out of 10 phytochemical compounds, 9 anti-SARS-CoV agents belonged to the flavonoid group. In conclusion, the proposed CNN model can prove useful for developing novel target-specific anti-SARS-CoV compounds.

Could graph neural networks learn better molecular representation for drug discovery? A comparison study of descriptor-based and graph-based models

Graph neural networks (GNN) has been considered as an attractive modelling method for molecular property prediction, and numerous studies have shown that GNN could yield more promising results than traditional descriptor-based methods. In this study, based on 11 public datasets covering various property endpoints, the predictive capacity and computational efficiency of the prediction models developed by eight machine learning (ML) algorithms, including four descriptor-based models (SVM, XGBoost, RF and DNN) and four graph-based models (GCN, GAT, MPNN and Attentive FP), were extensively tested and compared. The results demonstrate that on average the descriptor-based models outperform the graph-based models in terms of prediction accuracy and computational efficiency. SVM generally achieves the best predictions for the regression tasks. Both RF and XGBoost can achieve reliable predictions for the classification tasks, and some of the graph-based models, such as Attentive FP and GCN, can yield outstanding performance for a fraction of larger or multi-task datasets. In terms of computational cost, XGBoost and RF are the two most efficient algorithms and only need a few seconds to train a model even for a large dataset. The model interpretations by the SHAP method can effectively explore the established domain knowledge for the descriptor-based models. Finally, we explored use of these models for virtual screening (VS) towards HIV and demonstrated that different ML algorithms offer diverse VS profiles. All in all, we believe that the off-the-shelf descriptor-based models still can be directly employed to accurately predict various chemical endpoints with excellent computability and interpretability.

Screening of Natural Product and Natural Product like Molecules against SARS–CoV–2 Main Protease Using Molecular Modeling Methods

&#x0D; Objective: To determine possible MPro enzyme inhibitors by using structure-based virtual screening methods, in the ZINC Biogenic Data Set containing natural products and natural product-like molecules.&#x0D; Materials and Methods: QVina, an AutoDockVina derivative, was used in virtual screening operations, GROMACS in molecular dynamics studies and SwissAdme server in ADME (Absorption, Distribution, Metabolism, and Excretion) calculations. KNIME (Konstanz Information Miner) and ChemAxon software were used for filtering data and creating three-dimensional structures of the molecules.&#x0D; Results: Seven out of totally screened 51535 natural products or natural products like molecules were identified as possible candidate to be used as SARS–CoV–2 Main Protease (MPro) enzyme inhibitors based on the results obtained from structure based virtual screening and ADME models.&#x0D; Conclusion: Among the seven potent molecules, two of them (ZINC000604382012 and ZINC000514288074) were selected as candidate molecules for further studies according to the results obtained from g_mmpbsa simulations and synthetic accessibility models. In addition, a workflow has been established to identify novel or potent Mpro enzyme inhibitors.&#x0D;

MGCN: Graph Convolutional Network for Drug Molecular Screening

The intersection of drug research and development and neural network research is getting closer, new methods and new technologies are emerging one after another, using neural network to improve the efficiency of drug research and development, and rapidly innovate the treatment of diseases. In drug development, one target protein corresponds to multiple drug molecules. In order to save the time and cost of new drug development, virtual screening of key steps in new drug development is studied. A virtual screening model for drug molecules is proposed, and the graph convolutional neural network is used to classify drug molecules to achieve the purpose of preliminary screening. The graph convolutional network model is constructed based on DGL and combined with cross-entropy and fully connected networks. The model is carried out by experimental means analysis. The experimental results show that the graph neural network can complete the classification task of drug molecules. Through the verification of molecular docking software, the screened molecules are real and effective, reducing the time spent on virtual screening, and it has become an inspiration for further exploration of graph neural networks in drug development.

A probable means to an end: exploring P131 pharmacophoric scaffold to identify potential inhibitors of Cryptosporidium parvum inosine monophosphate dehydrogenase

Compound P131 has been established to inhibit Cryptosporidium parvum's inosine monophosphate dehydrogenase (CpIMPDH). Its inhibitory activity supersedes that of paromomycin, which is extensively used in treating cryptosporidiosis. Through the per-residue energy decomposition approach, crucial moieties of P131 were identified and subsequently adopted to create a pharmacophore model for virtual screening in the ZINC database. This search generated eight ADMET-compliant hits that were examined thoroughly to fit into the active site of CpIMPDH via molecular docking. Three compounds ZINC46542062, ZINC58646829, and ZINC89780094, with favorable docking scores of - 8.3 kcal/mol, - 8.2 kcal/mol, and - 7.5 kcal/mol, were selected. The potential inhibitory mechanism of these compounds was probed using molecular dynamics simulation and Molecular Mechanics Generalized Poisson Boltzmann Surface Area (MM/PBSA) analyses. Results revealed that one of the hits (ZINC46542062) exhibited a lower binding free energy of - 39.52 kcal/mol than P131, which had - 34.6 kcal/mol. Conformational perturbation induced by the binding of the identified hits to CpIMPDH was similar to P131, suggesting a similarity in inhibitory mechanisms. Also, in silico investigation of the properties of the hit compounds implied superior physicochemical properties with regards to their synthetic accessibility, lipophilicity, and number of hydrogen bond donors and acceptors in comparison with P131. ZINC46542062 was identified as a promising hit compound with the highest binding affinity to the target protein and favorable physicochemical and pharmacokinetic properties relative to P131. The identified compounds can serve as a basis for conducting further experimental investigations toward the development of anticryptosporidials, which can overcome the challenges of existing therapeutic options. Graphical abstract P131 and the identified compounds docked in the NAD+ binding site of Cryptosporidium parvum IMPDH.

Data-driven algorithms for inverse design of polymers.

The ever-increasing demand for novel polymers with superior properties requires a deeper understanding and exploration of the chemical space. Recently, data-driven approaches to explore the chemical space for polymer design have emerged. Among them, inverse design strategies for designing polymers with specific properties have evolved to be a significant materials informatics platform by learning hidden knowledge from materials data as well as smartly navigating the chemical space in an optimized way. In this review, we first summarize the progress in the representation of polymers, a prerequisite step for the inverse design of polymers. Then, we systematically introduce three data-driven strategies implemented for the inverse design of polymers, i.e., high-throughput virtual screening, global optimization, and generative models. Finally, we discuss the challenges and opportunities of the data-driven strategies as well as optimization algorithms employed in the inverse design of polymers.

Automated Framework for Developing Predictive Machine Learning Models for Data-Driven Drug Discovery

The increasing availability of extensive collections of chemical compounds associated with experimental data provides an opportunity to build predictive quantitative structure-activity relationship (QSAR) models using machine learning (ML) algorithms. These models can promote data-driven decisions and have the potential to speed up the drug discovery process and reduce their failure rates. However, many essential aspects of data preparation and modeling are not available in any standalone program. Here, we developed an automated framework for the curation of chemogenomics data and to develop QSAR models for virtual screening using the open-source KoNstanz Information MinEr (KNIME) program. The workflow includes four modules: (i) dataset preparation and curation; (ii) chemical space analysis and structure-activity relationships (SAR) rules; (iii) modeling; and (iv) virtual screening (VS). As case studies, we applied these workflows to four datasets associated with different endpoints. The implemented protocol can efficiently curate chemical and biological data in public databases and generates robust QSAR models. We provide scientists a simple and guided cheminformatics workbench following the best practices widely accepted by the community, in which scientists can adapt to solve their research problems. The workflows are freely available for download at GitHub and LabMol web portals.

Drug Screening of Potential Multiple Target Inhibitors for Estrogen Receptor-α-positive Breast Cancer.

Estrogen receptor α (ERα) antagonist is the most common treatment for ERα-positive breast cancer. However, compensatory signaling contributes to resistance to ERα antagonists. Thus, to explore the potential agents for targeting compensatory signaling, we screened multiple target inhibitors for breast cancer treatment. We attempted to build a structure-based virtual screening model that can find potential compounds and assay the anticancer ability of these drugs by overall cell survival assay. The downstream compensatory phosphorylated signaling was measured by immunoblotting. Hamamelitannin and glucocheirolin were hits for ERα, phosphoinositide 3-kinase (PI3K), and KRAS proto-oncogene, GTPase (KRAS), which were active against estrogen and epidermal growth factor-triggered proliferation. Additionally, we select aminopterin as a hit for ERα, PI3K, KRAS, and SRC proto-oncogene, non-receptor tyrosine kinase (SRC) with inhibitory activities toward AKT serine/threonine kinase 1 (AKT) and mitogen-activated protein kinase kinase (MEK) signaling. Our structure-based virtual screening model selected hamamelitannin, glucocheirolin, aminopterin, and pemetrexed as compounds that may act as potential inhibitors for improving endocrine therapies for breast cancer.

Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models.

Generative models are becoming a tool of choice for exploring the molecular space. These models learn on a large training dataset and produce novel molecular structures with similar properties. Generated structures can be utilized for virtual screening or training semi-supervized predictive models in the downstream tasks. While there are plenty of generative models, it is unclear how to compare and rank them. In this work, we introduce a benchmarking platform called Molecular Sets (MOSES) to standardize training and comparison of molecular generative models. MOSES provides training and testing datasets, and a set of metrics to evaluate the quality and diversity of generated structures. We have implemented and compared several molecular generation models and suggest to use our results as reference points for further advancements in generative chemistry research. The platform and source code are available at https://github.com/molecularsets/moses.