Artificial intelligence chemistry最新文献

Artificial neural networks (ANN) have been shown for the last several years to be a versatile tool for fitting ab initio potential energy surfaces. We have demonstrated recently how a 60-neuron ANN could successfully fit a four-dimensional ab initio potential energy surface for the rigid rotor HeH⁺ - rigid rotor H₂ system with a root-mean-squared deviation (RMSD) of 35 cm⁻¹. We show in the present study how a (40, 40) neural network with two hidden layers could achieve a better fit with an RMSD of 5 cm⁻¹. Through a follow-up quantum dynamical study of HeH⁺($j_1$)-H₂($j_2$) collisions, it is shown that the two fits lead to slightly different rotational excitation and de-excitation cross sections but are comparable to each other in terms of magnitude and dependence on the relative translational energy of the collision partners. When averaged over relative translational energy, the two sets of results lead to rate coefficients that are nearly indistinguishable at higher temperatures thus demonstrating the reliability of the ANN method for fitting ab initio potential energy surfaces. On the other hand, we also find that the de-excitation rate coefficients obtained using the two different ANN fits differ significantly from each other at low temperatures. The consequences of these findings are discussed in our conclusions.

One of the key steps to make an artificially intelligent (AI) and robotic chemist is the introduction of machine vision for guiding the experiment operation in the AI-redefined laboratory. In order to realize the targets, the prerequisites are to innovate/implement the intelligent vision for the detection of chemistry glassware. Here, we reported a computer vision method based on You only look once (YOLO) with a self-developed Chemical Vessel Identification Dataset (CViG) for the improvement of classification and recognition performance. The training dataset has been collected that includes 4072 images in real-time chemical laboratory. Three models, YOLOv5s, Slim-YOLOv5s and YOLOv7, have been exploited for the recognition of seven types of glassware in the condition of different scenarios (recognition distance, light and dark, stationary and moving). The improved Slim-YOLOv5s exhibited better recognition ability in various scenes, and the recognition accuracy of chemical vessels is improved by 1.51 % compared with YOLOv5s, and the size of the model is reduced from 14.4 MB to 11.0 MB. Slim-YOLOv5s's mAP is similar to YOLOv7's ability with a disadvantage of large volume, suggested that the improved Slim-YOLOv5s clearly has more advantages in terms of embedded requirements. This vision-assisted system capable of classifying chemical containers accurately in the scenarios of real-time chemical experiments will provide a good vision solution in the frontier fields of automated machine chemistry.

The novel coronavirus disease (COVID-19) was caused by a new strain of the virus SARS-CoV-2 in December 2019 emerged as deadly pandemic that affected millions of people worldwide. Factors such as lack of effective drugs, vaccine resistance, gene mutations, and cost of repurposed drugs demand new potential inhibitors. The main protease (Mpro) of SARS-CoV-2 has a key role in viral replication and transcription and is considered as drug target for new lead identification. In this present work, structure-based virtual screening, docking, MM/GBSA, AutoDock, ADME, and MD simulations-based optimization was proposed for the identification of new potential inhibitors against Mpro of SARS-CoV-2. The ligand molecules M1, M3, and M6 were identified as potential leads from lead optimization. Induced fit docking was performed for the identification of the best poses of lead molecules. The best docked poses of potential leads M1 and M3 were subject to 100 ns MD simulations for the evaluation of stability and interaction analysis into Mpro active site. The structures of the top two leads M1 and M3 were optimized based on MD simulation conformational changes and isoster scanning, designed as new leads M7 and M8. The MD simulation trajectories RMSD, RMSF, protein-ligand, ligand-protein interaction plots, and ligand torsion profiles were analyzed for stability interpretation. The docked complexes of M7 and M8 of Mpro exhibited equilibrated and converged plots in 100 ns simulation. The lead molecules M1, M3, M7, and M8 were identified as potential SARS-CoV-2 inhibitors for COVID-19 disease. A comparative docking study was carried out using FDA-approved drugs to support the potential binding affinities of newly identified lead inhibitors.

The human immunodeficiency virus type 1 (HIV-1) is a retrovirus that can cause acquired immunodeficiency syndrome (AIDS), severely weakening the immune system. The United Nations estimates that there are 37.7 million people with HIV worldwide. HIV-1 protease (PR) cleaves polyproteins to create the individual proteins that comprise an HIV virion. Inhibiting PR prevents the creation of new virions, rendering PR an attractive antiviral target. In the present study, a machine-learning regression model was constructed to predict pIC₅₀ bioactivity concentrations using data from 2547 experimentally characterized PR inhibitors. The model achieved Pearson correlation coefficient of 0.88, R-squared of 0.78, and a RMSE of 0.717 in pIC₅₀ units on unseen data using 199 high-variance PubChem substructure fingerprints. The SWEETLEAD database of approximately 4300 traditional medicine compounds and drugs from around the world was screened using the model. Fifty molecules were identified as highly potent, with pIC₅₀ of at least 7.301 (IC₅₀ <= 50 nM). Nine of these molecules, such as lopinavir and ritonavir, are known antiviral drugs. The highly potent molecules were ligand-docked to the 3D structure of HIV protease at the active site. Dihydroergotamine mesylate (daechu alkaloids) had a very strong binding affinity of −13.2, outperforming all known antiviral drugs that were tested. It was also predicted by the model to have an IC₅₀ of 9.16 nM, which is considered very low and desirable. Overall, this study demonstrates the use of machine-learning regression models for virtual screening and highlights several drugs with significant promise for repurposing against HIV-1. Future steps include testing dihydroergotamine mesylate and other candidates in vitro.

Orders-of-coupling representations (representations of multivariate functions with low-dimensional functions that depend on subsets of original coordinates corresponding to different orders of coupling) are useful in many applications, for example, in computational chemistry and other applications, especially where integration is needed. Examples include N-mode approximations and many-body expansions. Such representations can be conveniently built with machine learning methods, and previously, methods building the lower-dimensional terms of such representations with neural networks [e.g. Comput. Phys. Commun. 180 (2009) 2002] and Gaussian process regressions [e.g. Mach. Learn. Sci. Technol. 3 (2022) 01LT02] were proposed. Here, we show that neural network models of orders-of-coupling representations can be easily built by using a recently proposed neural network with optimal neuron activation functions computed with a first-order additive Gaussian process regression [arXiv:2301.05567] and avoiding non-linear parameter optimization. Examples are given of representations of molecular potential energy surfaces.

Monoamine Oxidase-B (MAO-B) is a key neuroprotective target that breaks neurotransmitters such as dopamine and releases highly reactive free radicals as the by-product. Its over-expression in the brain observed due to ageing and neurodegenerative diseases contributes to worsening neuronal degeneration. Being the primary enzyme for dopamine metabolism in the substantia nigra of the brain and due to the lack of efficient drug candidates, MAO-B selective, reversible inhibition is hot topic of research in Parkinson’s disease (PD). This study developed machine learning (ML) models that predict the activity of experimentally tested indole and indazole derivatives against MAO-B using linear genetic function approximation (GFA) and two non-linear support vector machine (SVM) and artificial neural network (ANN) techniques. ANN model with an R² of 0.9704 for the training dataset, $q^2$of 0.9436 for cross-validation and $r^2$of 0.9025 for the test dataset were identified as the best-performing ML model with the seven significant molecular descriptors CATS2D_04_DA, CATS2D_05_DA, CATS3D_06_LL, Mor04u, Mor25m, P_VSA_v_2 and nO. The robust ML model was then employed to design novel MAO-B inhibitors with similar core scaffolds and their biological activity prediction. ANN model was further employed in the virtual screening of 4356 molecules from the ChEMBL database. Applicability domain analysis and pharmacokinetic and toxicity profiles predicted three newly designed molecules (22 N, 23 N and 24 N) and two virtually screened best ChEMBL molecules as potential drug candidates using the ANN ML model. Molecular docking studies of the best-identified compounds were performed to understand the molecular mechanism of interactions having high binding energy and selectivity with the MAO-B enzyme. The current study shortlisted 5 potential lead compounds as potent and selective MAO-B inhibitors, which could further be carried forward for in vitro and in vivo studies to discover small molecules against neurodegenerative disease.

Adverse drug reactions (ADRs) and drug-induced toxicity are major challenges in drug discovery, threatening patient safety and dramatically increasing healthcare expenditures. Since ADRs and toxicity are not as visible as infectious diseases, the potential consequences are considerable. Early detection of ADRs and drug-induced toxicity is an essential indicator of a drug's viability and safety profile. The introduction of artificial intelligence (AI) and machine learning (ML) approaches has resulted in a paradigm shift in the field of early ADR and toxicity detection. The application of these modern computational methods allows for the rapid, thorough, and precise prediction of probable ADRs and toxicity even before the drug’s practical synthesis as well as preclinical and clinical trials, resulting in more efficient and safer medications with a lesser chance of drug’s withdrawal. This present review offers an in-depth examination of the role of AI and ML in the early detection of ADRs and toxicity, incorporating a wide range of methodologies ranging from data mining to deep learning followed by a list of important databases, modeling algorithms, and software that could be used in modeling and predicting a series of ADRs and toxicity. This review also provides a complete reference to what has been performed and what might be accomplished in the field of AI and ML-based early identification of ADRs and drug-induced toxicity. By shedding light on the capabilities of these technologies, it highlights their enormous potential for revolutionizing drug discovery and improving patient safety.

As increasingly more data science-driven approaches have been applied for compound properties predictions in the domain of drug discovery, such kinds of methods have displayed considerable accuracy compared to conventional ones. In this work, we proposed an interpretable graph learning representation model, SolubNet, for drug aqueous solubility prediction. The comprehensive evaluation demonstrated that SolubNet can successfully capture the quantitative structure-property relationship and can be interpreted with layer-wise relevance propagation (LRP) algorithm regarding how prediction values are generated from original input structures. The key advantage of SolubNet lies in the fact that it includes 3 layers of Topology Adaptive Graph Convolutional Networks which can efficiently perceive chemical local environments. SolubNet showed high performance in several tasks for drugs’ aqueous solubility prediction. LRP revealed that SolubNet can identify high and low polar regions of a given molecule, assigning them reasonable weights to predict the final solubility, in a way highly compatible with chemists’ intuition. We are confident that such a flexible yet interpretable and accurate tool will largely enhance the efficiency of drug discovery, and will even contribute to the methodology development of computational pharmaceutics.

In the present research, a few well-known artificial intelligence tools were explored for efficient hit selection which could be further utilized for the discovery of CaMK-II inhibitors for the Treatment of arrhythmia. To achieve the desired goals pharmacophore modelling, database retrieval, molecular docking studies, and toxicity prediction were performed. Pharmacophore modelling was performed with the Pharmit open-source database which gave the features viz. Hydrogen Bond Donor, Hydrogen Bond Acceptor, and Hydrophobic. This pharmacophore is generated with the aid of the protein of CaMK-II (PDB ID: 2WEL) and co-crystallized ligand K88. Further, this generated pharmacophore was screened through the various Pharmit databases which include CHEMBL30, ChemDiv, ChemSpace, MCULE, MolPort, NCI Open Chemical Repository, Lab Network, and ZINC. Further, the top two hits from each database that has maximum similarity with the pharmacophore have been selected for the molecular docking and ADMET studies. Among, all the hits CHEMBL 1952032 showed good binding interactions with CaMK-II. Also, it was found to be non-toxic upon evaluation through the OSIRIS property explorer. In the future, it can be explored against the CaMK-II for the development of novel CaMK-II inhibitors which can be used for the mitigation of arrhythmia.