In Silico Absorption, Distribution, Metabolism, Excretion, and Pharmacokinetics (ADME-PK): Utility and Best Practices. An Industry Perspective from the International Consortium for Innovation through Quality in Pharmaceutical Development

Application of machine learning models for property prediction to targeted protein degraders

Machine learning (ML) systems can model quantitative structure-property relationships (QSPR) using existing experimental data and make property predictions for new molecules. With the advent of modalities such as targeted protein degraders (TPD), the applicability of QSPR models is questioned and ML usage in TPD-centric projects remains limited. Herein, ML models are developed and evaluated for TPDs' property predictions, including passive permeability, metabolic clearance, cytochrome P450 inhibition, plasma protein binding, and lipophilicity. Interestingly, performance on TPDs is comparable to that of other modalities. Predictions for glues and heterobifunctionals often yield lower and higher errors, respectively. For permeability, CYP3A4 inhibition, and human and rat microsomal clearance, misclassification errors into high and low risk categories are lower than 4% for glues and 15% for heterobifunctionals. For all modalities, misclassification errors range from 0.8% to 8.1%. Investigated transfer learning strategies improve predictions for heterobifunctionals. This is the first comprehensive evaluation of ML for the prediction of absorption, distribution, metabolism, and excretion (ADME) and physicochemical properties of TPD molecules, including heterobifunctional and molecular glue sub-modalities. Taken together, our investigations show that ML-based QSPR models are applicable to TPDs and support ML usage for TPDs' design, to potentially accelerate drug discovery.

Identification of novel GSPT1 degraders by virtual screening and bioassay

GSPT1 plays crucial physiological functions, such as terminating protein translation, overexpressed in various tumors. It is a promising anti-tumor target, but is also considered as an "undruggable" protein. Recent studies have found that a class of small molecules can degrade GSPT1 through the "molecular glue" mechanism with strong antitumor activity, which is expected to become a new therapy for hematological malignancies. Currently available GSPT1 degraders are mostly derived from the scaffold of immunomodulatory imide drug (IMiD), thus more active compounds with novel structure remain to be found. In this work, using computer-assisted multi-round virtual screening and bioassay, we identified a non-IMiD acylhydrazone compound, AN5782, which can reduce the protein level of GPST1 and obviously inhibit the proliferation of tumor cells. Some analogs were obtained by a substructure search of AN5782. The structure-activity relationship analysis revealed possible interactions between these compounds and CRBN-GSPT1. Further biological mechanistic studies showed that AN5777 decreased GSPT1 remarkably through the ubiquitin-proteasome system, and its effective cytotoxicity was CRBN- and GSPT1-dependent. Furthermore, AN5777 displayed good antiproliferative activities against U937 and OCI-AML-2 cells, and dose-dependently induced G1 phase arrest and apoptosis. The structure found in this work could be good start for antitumor drug development.

In silico ADME/tox comes of age: twenty years later

In the early 2000s pharmaceutical drug discovery was beginning to use computational approaches for absorption, distribution, metabolism, excretion and toxicity (ADME/Tox, also known as ADMET) prediction. This emphasis on prediction was an effort to reduce the risk of later stage failures from ADME/Tox.Much has been written in the intervening twenty plus years and significant expenditure has occurred in companies developing these in silico capabilities which can be gleaned from publications. It is therefore an appropriate time to briefly reflect on what was proposed then and what the reality is today.20 years ago, we tended to optimise bioactivity and perhaps one ADME/Tox property at a time. Previously pharmaceutical companies needed a whole infrastructure for models - in silico and in vitro experts, IT, champions on a project team, educators and management support. Now we are in the age of generative de novo design where bioactivity and many ADME/Tox properties can be optimised and large language model technologies are available.There are also some challenges such as the focus on very large molecules which may be outside of current ADME/Tox models.We provide an opportunity to look forward with the increasing public data for ADME/Tox as well as expanded types of algorithms available.

Reducing overconfident errors in molecular property classification using Posterior Network

Deep-learning-based classification models are increasingly used for predicting molecular properties in drug development. However, traditional classification models using the Softmax function often give overconfident mispredictions for out-of-distribution samples, highlighting a critical lack of accurate uncertainty estimation. Such limitations can result in substantial costs and should be avoided during drug development. Inspired by advances in evidential deep learning and Posterior Network, we replaced the Softmax function with a normalizing flow to enhance the uncertainty estimation ability of the model in molecular property classification. The proposed strategy was evaluated across diverse scenarios, including simulated experiments based on a synthetic dataset, ADMET predictions, and ligand-based virtual screening. The results demonstrate that compared with the vanilla model, the proposed strategy effectively alleviates the problem of giving overconfident but incorrect predictions. Our findings support the promising application of evidential deep learning in drug development and offer a valuable framework for further research.

An Automated Robotic Interface for Assays: Facilitating Machine Learning in Drug Discovery by the Automation of Physicochemical Property Assays

Measuring the physicochemical properties of molecules is an iterative but integral process in the drug development process. A strategy to overcome the challenges in maximizing assay throughput relies on the usage of in silico machine learning (ML) prediction models trained on experimental data. Consequently, the performance of these in silico models are dependent on the quality of the utilized experimental data. To improve the data quality, we have designed and implemented an automated robotic system to prepare and run physicochemical property assays (Automated Robotic Interface for Assays, ARIA) with an increase in sample throughput of 6 to10-fold. Through this process, we overcame major challenges and achieved consistent reproducible assay data compared to semiautomated assay preparation.

Ligand-based drug design against Herpes Simplex Virus-1 capsid protein by modification of limonene through in silico approaches

The pharmacological effects of limonene, especially their derivatives, are currently at the forefront of research for drug development and discovery as well and structure-based drug design using huge chemical libraries are already widespread in the early stages of therapeutic and drug development. Here, various limonene derivatives are studied computationally for their potential utilization against the capsid protein of Herpes Simplex Virus-1. Firstly, limonene derivatives were designed by structural modification followed by conducting a molecular docking experiment against the capsid protein of Herpes Simplex Virus-1. In this research, the obtained molecular docking score exhibited better efficiency against the capsid protein of Herpes Simplex Virus-1 and hence we conducted further in silico investigation including molecular dynamic simulation, quantum calculation, and ADMET analysis. Molecular docking experiment has documented that Ligands 02 and 03 had much better binding affinities (- 7.4 kcal/mol and - 7.1 kcal/mol) to capsid protein of Herpes Simplex Virus-1 than Standard Acyclovir (- 6.5 kcal/mol). Upon further investigation, the binding affinities of primary limonene were observed to be slightly poor. But including the various functional groups also increases the affinities and capacity to prevent viral infection of the capsid protein of Herpes Simplex Virus-1. Then, the molecular dynamic simulation confirmed that the mentioned ligands might be stable during the formation of drug-protein complexes. Finally, the analysis of ADMET was essential in establishing them as safe and human-useable prospective chemicals. According to the present findings, limonene derivatives might be a promising candidate against the capsid protein of Herpes Simplex Virus-1 which ultimately inhibits Herpes Simplex Virus-induced encephalitis that causes interventions in brain inflammation. Our findings suggested further experimental screening to determine their practical value and utility.

Computational Applications: Beauvericin from a Mycotoxin into a Humanized Drug

Drug discovery was initially attributed to coincidence or experimental research. Historically, the traditional approaches were complex, lengthy, and expensive, entailing costly random screening of synthesized compounds or natural products coupled with in vivo validation largely depending on the availability of appropriate animal models. Currently, in silico modeling has become a vital tool for drug discovery and repurposing. Molecular docking and dynamic simulations are being used to find the best match between a ligand and a molecule, an approach that could help predict the biomolecular interactions between the drug and the target host. Beauvericin (BEA) is an emerging mycotoxin produced by the entomopathogenic fungus Beauveria bassiana, being originally studied for its potential use as a pesticide. BEA is now considered a molecule of interest for its possible use in diverse biotechnological applications in the pharmaceutical industry and medicine. In this manuscript, we provide an overview of the repurposing of BEA as a potential therapeutic agent for multiple diseases. Furthermore, considerable emphasis is given to the fundamental role of in silico techniques to (i) further investigate the activity spectrum of BEA, a secondary metabolite, and (ii) elucidate its mode of action.

Adapting Deep Learning QSPR Models to Specific Drug Discovery Projects

Medicinal chemistry and drug design efforts can be assisted by machine learning (ML) models that relate the molecular structure to compound properties. Such quantitative structure-property relationship models are generally trained on large data sets that include diverse chemical series (global models). In the pharmaceutical industry, these ML global models are available across discovery projects as an "out-of-the-box" solution to assist in drug design, synthesis prioritization, and experiment selection. However, drug discovery projects typically focus on confined parts of the chemical space (e.g., chemical series), where global models might not be applicable. Local ML models are sometimes generated to focus on specific projects or series. Herein, ML-based global models, local models, and hybrid global-local strategies were benchmarked. Analyses were done for more than 300 drug discovery projects at Novartis and ten absorption, distribution, metabolism, and excretion (ADME) assays. In this work, hybrid global-local strategies based on transfer learning approaches were proposed to leverage both historical ADME data (global) and project-specific data (local) to adapt model predictions. Fine-tuning a pretrained global ML model (used for weights' initialization, WI) was the top-performing method. Average improvements of mean absolute errors across all assays were 16% and 27% compared with global and local models, respectively. Interestingly, when the effect of training set size was analyzed, WI fine-tuning was found to be successful even in low-data scenarios (e.g., ∼10 molecules per project). Taken together, this work highlights the potential of domain adaptation in the field of molecular property predictions to refine existing pretrained models on a new compound data distribution.

BioPrint meets the AI age: development of artificial intelligence-based ADMET models for the drug-discovery platform SAFIRE

Background: To prioritize compounds with a higher likelihood of success, artificial intelligence models can be used to predict absorption, distribution, metabolism, excretion and toxicity (ADMET) properties of molecules quickly and efficiently. Methods: Models were trained with BioPrint database proprietary data along with public datasets to predict various ADMET end points for the SAFIRE platform. Results: SAFIRE models performed at or above 75% accuracy and 0.4 Matthew's correlation coefficient with validation sets. Training with both proprietary and public data improved model performance and expanded the chemical space on which the models were trained. The platform features scoring functionality to guide user decision-making. Conclusion: High-quality datasets along with chemical space considerations yielded ADMET models performing favorably with utility in the drug discovery process.

Artificial Intelligence for Quantitative Modeling in Drug Discovery and Development: An Innovation and Quality Consortium Perspective on Use Cases and Best Practices

Recent breakthroughs in artificial intelligence (AI) and machine learning (ML) have ushered in a new era of possibilities across various scientific domains. One area where these advancements hold significant promise is model-informed drug discovery and development (MID3). To foster a wider adoption and acceptance of these advanced algorithms, the Innovation and Quality (IQ) Consortium initiated the AI/ML working group in 2021 with the aim of promoting their acceptance among the broader scientific community as well as by regulatory agencies. By drawing insights from workshops organized by the working group and attended by key stakeholders across the biopharma industry, academia, and regulatory agencies, this white paper provides a perspective from the IQ Consortium. The range of applications covered in this white paper encompass the following thematic topics: (i) AI/ML-enabled Analytics for Pharmacometrics and Quantitative Systems Pharmacology (QSP) Workflows; (ii) Explainable Artificial Intelligence and its Applications in Disease Progression Modeling; (iii) Natural Language Processing (NLP) in Quantitative Pharmacology Modeling; and (iv) AI/ML Utilization in Drug Discovery. Additionally, the paper offers a set of best practices to ensure an effective and responsible use of AI, including considering the context of use, explainability and generalizability of models, and having human-in-the-loop. We believe that embracing the transformative power of AI in quantitative modeling while adopting a set of good practices can unlock new opportunities for innovation, increase efficiency, and ultimately bring benefits to patients.

Cross-Species Extrapolation of Biological Data to Guide the Environmental Safety Assessment of Pharmaceuticals-The State of the Art and Future Priorities

The extrapolation of biological data across species is a key aspect of biomedical research and drug development. In this context, comparative biology considerations are applied with the goal of understanding human disease and guiding the development of effective and safe medicines. However, the widespread occurrence of pharmaceuticals in the environment and the need to assess the risk posed to wildlife have prompted a renewed interest in the extrapolation of pharmacological and toxicological data across the entire tree of life. To address this challenge, a biological "read-across" approach, based on the use of mammalian data to inform toxicity predictions in wildlife species, has been proposed as an effective way to streamline the environmental safety assessment of pharmaceuticals. Yet, how effective has this approach been, and are we any closer to being able to accurately predict environmental risk based on known human risk? We discuss the main theoretical and experimental advancements achieved in the last 10 years of research in this field. We propose that a better understanding of the functional conservation of drug targets across species and of the quantitative relationship between target modulation and adverse effects should be considered as future research priorities. This pharmacodynamic focus should be complemented with the application of higher-throughput experimental and computational approaches to accelerate the prediction of internal exposure dynamics. The translation of comparative (eco)toxicology research into real-world applications, however, relies on the (limited) availability of experts with the skill set needed to navigate the complexity of the problem; hence, we also call for synergistic multistakeholder efforts to support and strengthen comparative toxicology research and education at a global level. Environ Toxicol Chem 2024;43:513-525. © 2023 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Design, synthesis, and pharmacological evaluation of indazole carboxamides of N-substituted pyrrole derivatives as soybean lipoxygenase inhibitors

In this paper, we attempted to develop a novel class of compounds against lipoxygenase, a key enzyme in the biosynthesis of leukotrienes implicated in a series of inflammatory diseases. Given the absence of appropriate human 5-lipoxygenase crystallographic data, solved soybean lipoxygenase-1 and -3 structures were used as a template to generate an accurate pharmacophore model which was further used for virtual screening purposes. Eight compounds (1-8) have been derived from the in-house library consisting of N-substituted pyrroles conjugated with 5- or 6-indazole moieties through a carboxamide linker. This study led to the discovery of hit molecule 8 bearing a naphthyl group with the IC50 value of 22 μM according to soybean lipoxygenase in vitro assay. Isosteric replacement of naphthyl ring with quinoline moieties and reduction of carbonyl carboxamide group resulted in compounds 9-12 and 13, respectively. Compound 12 demonstrated the most promising enzyme inhibition. In addition, compounds 8 and 12 were found to reduce the carrageenan-induced paw edema in vivo by 52.6 and 49.8%, respectively. In view of the encouraging outcomes concerning their notable in vitro and in vivo anti-inflammatory activities, compounds 8 and 12 could be further optimized for the discovery of novel 5-lipoxygenase inhibitors in future. A structure-based 3D pharmacophore model was used in the virtual screening of in-house library to discover novel potential 5-lipoxygenase inhibitors.

Structural Dynamics of P-Rex1 Complexed with Natural Leads Establishes the Protein as an Attractive Target for Therapeutics to Suppress Cancer Metastasis

Phosphatidylinositol 3,4,5-trisphosphate- (PIP3-) dependent Rac exchanger 1 (P-Rex1) functions as Rho guanine nucleotide exchange factor and is activated by synergistic activity of Gβγ and PIP3 of the heterotrimeric G protein. P-Rex1 activates Rac GTPases for regulating cell invasion and migration and promotes metastasis in several human cancers including breast, prostate, and skin cancer. The protein is a promising therapeutic target because of its multifunction roles in human cancers. Herein, the present study attempts to identify selective P-Rex1 natural inhibitors by targeting PIP3-binding pocket using large-size multiple natural molecule libraries. Each library was filtered subsequently in FAF-Drugs4 based on Lipinski's rule of five (RO5), toxicity, and filter pan assay interference compounds (PAINS). The output hits were virtually screened at the PIP3-binding pocket through PyRx AutoDock Vina and cross-checked by GOLD. The best binders at the PIP3-binding pocket were prioritized using a comparative analysis of the docking scores. Top-ranked two compounds with high GOLD fitness score (>80) and lowest AutoDock binding energy (< -12.7 kcal/mol) were complexed and deciphered for molecular dynamics along with control-P-Rex1 complex to validate compound binding conformation and disclosed binding interaction pattern. Both the systems were seen in good equilibrium, and along the simulation time, the compounds are in strong contact with the P-Rex1 PIP3-binding site. Hydrogen bonding analysis towards simulation end identified the formation of 16 and 22 short- and long-distance hydrogen bonds with different percent of occupancy to the PIP3 residues for compound I and compound 2, respectively. Radial distribution function (RDF) analysis of the key hydrogen bonds between the compound and the PIP3 residues demonstrated a strong affinity of the compounds to the mentioned PIP3 pocket. Additionally, MMGB/PBSA energies were performed that confirmed the dominance of Van der Waals energy in complex formation along with favorable contribution from hydrogen bonding. These findings were also cross-validated by a more robust WaterSwap binding energy predictor, and the results are in good agreement with a strong binding affinity of the compounds for the protein. Lastly, the key contribution of residues in interaction with the compounds was understood by binding free energy decomposition and alanine scanning methods. In short, the results of this study suggest that P-Rex1 is a good druggable target to suppress cancer metastasis; therefore, the screened druglike molecules of this study need in vitro and in vivo anti-P-Rex1 validation and may serve as potent leads to fight cancer.

DeepDelta: predicting ADMET improvements of molecular derivatives with deep learning

Established molecular machine learning models process individual molecules as inputs to predict their biological, chemical, or physical properties. However, such algorithms require large datasets and have not been optimized to predict property differences between molecules, limiting their ability to learn from smaller datasets and to directly compare the anticipated properties of two molecules. Many drug and material development tasks would benefit from an algorithm that can directly compare two molecules to guide molecular optimization and prioritization, especially for tasks with limited available data. Here, we develop DeepDelta, a pairwise deep learning approach that processes two molecules simultaneously and learns to predict property differences between two molecules from small datasets. On 10 ADMET benchmark tasks, our DeepDelta approach significantly outperforms two established molecular machine learning algorithms, the directed message passing neural network (D-MPNN) ChemProp and Random Forest using radial fingerprints, for 70% of benchmarks in terms of Pearson's r, 60% of benchmarks in terms of mean absolute error (MAE), and all external test sets for both Pearson's r and MAE. We further analyze our performance and find that DeepDelta is particularly outperforming established approaches at predicting large differences in molecular properties and can perform scaffold hopping. Furthermore, we derive mathematically fundamental computational tests of our models based on mathematical invariants and show that compliance to these tests correlates with overall model performance - providing an innovative, unsupervised, and easily computable measure of expected model performance and applicability. Taken together, DeepDelta provides an accurate approach to predict molecular property differences by directly training on molecular pairs and their property differences to further support fidelity and transparency in molecular optimization for drug development and the chemical sciences.

Cheminformatics and biomolecular dynamics studies towards the discovery of anti-staphylococcal nuclease domain-containing 1 (SND1) inhibitors to treat metastatic breast cancer

Metastatic breast cancer is a prime health concern and leading health burden across the globe. Previous efforts have shown that protein-protein interaction between Metadherin and Staphylococcal nuclease domaincontaining 1 (SND1) promotes initiation of breast cancer, progression, therapy resistance and metastasis. Therefore, small drug molecules that can interrupt the Metadherin and SND1 interaction may be ideal to suppress tumor growth, metastasis and increases chemotherapy sensitivity of triple negative breast cancer. Here, in this study, structure based virtual screening was conducted against the reported active site of SND1 enzyme, which revealed three promising lead molecules from Asinex library. These compounds were; BAS_00381028, BAS_00327287, and BAS_01293454 with binding energy score -10.25 kcal/mol, -9.65 kcal/mol and -9.32 kcal/mol, respectively. Compared to control (5-chloro-2-methoxy-N-([1,2,4]triazolo[1,5-a]pyridin-8-yl)benzene-1-sulfonamide) the lead molecules showed robust hydrophilic and hydrophobic interactions with the enzyme and revealed stable docked conformation in molecular dynamics simulation. During the simulation time, the compounds reported stable dynamics with no obvious fluctuation in binding mode and interactions noticed. The mean root mean square deviation (RMSD) of BAS_00381028, BAS_00327287, and BAS_01293454 complexes were 1.87 Å, 1.75 Å, 1.34 Å, respectively. Furthermore, the MM/GBSA analysis was conduction on the simulation trajectories of complexes that unveiled binding energy score of -19.25 kcal/mol, -27.03 kcal/mol, -34.6 kcal/mol and -29.61 kcal/mol for control, BAS_00381028, BAS_00327287, and BAS_01293454, respectively. In MM/PBSA, the binding energy value of for control, BAS_00381028, BAS_00327287, and BAS_01293454 was -20.45 kcal/mol, -27.89 kcal/mol, -36.41 kcal/mol and -32.01 kcal/mol, respectively. Additionally, the compounds were classified as druglike and have favorable pharmacokinetic properties. The compounds were predicted as promising leads and might be used in experimental investigation to study their anti-SND1 activity.

Quercetin Antagonizes the Sedative Effects of Linalool, Possibly through the GABAergic Interaction Pathway

Sedatives promote calmness or sleepiness during surgery or severely stressful events. In addition, depression is a mental health issue that negatively affects emotional well-being. A group of drugs called anti-depressants is used to treat major depressive illnesses. The aim of the present work was to evaluate the effects of quercetin (QUR) and linalool (LIN) on thiopental sodium (TS)-induced sleeping mice and to investigate the combined effects of these compounds using a conventional co-treatment strategy and in silico studies. For this, the TS-induced sleeping mice were monitored to compare the occurrence, latency, and duration of the sleep-in response to QUR (10, 25, 50 mg/kg), LIN (10, 25, 50 mg/kg), and diazepam (DZP, 3 mg/kg, i.p.). Moreover, an in silico investigation was undertaken to assess this study's putative modulatory sedation mechanism. For this, we observed the ability of test and standard medications to interact with various gamma-aminobutyric acid A receptor (GABA_A) subunits. Results revealed that QUR and LIN cause dose-dependent antidepressant-like and sedative-like effects in animals, respectively. In addition, QUR-50 mg/kg and LIN-50 mg/kg and/or DZP-3 mg/kg combined were associated with an increased latency period and reduced sleeping times in animals. Results of the in silico studies demonstrated that QUR has better binding interaction with GABA_A α3, β1, and γ2 subunits when compared with DZP, whereas LIN showed moderate affinity with the GABA_A receptor. Taken together, the sleep duration of LIN and DZP is opposed by QUR in TS-induced sleeping mice, suggesting that QUR may be responsible for providing sedation-antagonizing effects through the GABAergic interaction pathway.

A cheminformatics-biophysics correlate to identify promising lead molecules against matrix metalloproteinase-2 (MMP-2) enzyme: A promising anti-cancer target

Matrix metalloproteinase-2 (MMP-2) is an endopeptidase enzyme that is devoted to extracellular matrix proteins degradation. The enzyme is warranted as promising drugs target for different light threating diseases such as arthritis, cancer and fibrosis. Herein, in this study, three drug molecules: CMNPD8322, CMNPD8320, and CMNPD8318 were filtered as high affinity binding compounds with binding energy score of -9.75 kcal/mol, -9.11 kcal/mol, -9.05 kcal/mol, respectively. The control binding energy score was -9.01 kcal/mol. The compounds docked deeply inside the pocket interacting with S1 pocket residues. The docked complexes dynamics in real time at cellular environment was then done to decipher the stable binding conformation and intermolecular interactions network. The compounds complexes achieved very stable dynamics with root mean square deviation (RMSD) with mean value of around 2-3 Å compared to control complex that showed higher fluctuations of 5 Å. The simulation trajectories frames based binding free energy demonstrated all the compounds-MMP-2 complexes reported highly stable energy, particularly the van der Waals energy dominate the overall net energy. Similarly, the complexes revalidation of WaterSwap based energies also disclosed the complexes highly stable in term docked conformation. Also, the compounds illustrated the compounds favorable pharmacokinetics and were non-toxic and non-mutagenic. Thus, the compounds might be used thorough experimental assays to confirm compounds selective biological potency against MMP-2 enzyme.

Prospective Validation of Machine Learning Algorithms for Absorption, Distribution, Metabolism, and Excretion Prediction: An Industrial Perspective

Absorption, distribution, metabolism, and excretion (ADME), which collectively define the concentration profile of a drug at the site of action, are of critical importance to the success of a drug candidate. Recent advances in machine learning algorithms and the availability of larger proprietary as well as public ADME data sets have generated renewed interest within the academic and pharmaceutical science communities in predicting pharmacokinetic and physicochemical endpoints in early drug discovery. In this study, we collected 120 internal prospective data sets over 20 months across six ADME in vitro endpoints: human and rat liver microsomal stability, MDR1-MDCK efflux ratio, solubility, and human and rat plasma protein binding. A variety of machine learning algorithms in combination with different molecular representations were evaluated. Our results suggest that gradient boosting decision tree and deep learning models consistently outperformed random forest over time. We also observed better performance when models were retrained on a fixed schedule, and the more frequent retraining generally resulted in increased accuracy, while hyperparameters tuning only improved the prospective predictions marginally.

Evaluation of Free Online ADMET Tools for Academic or Small Biotech Environments

For a new molecular entity (NME) to become a drug, it is not only essential to have the right biological activity also be safe and efficient, but it is also required to have a favorable pharmacokinetic profile including toxicity (ADMET). Consequently, there is a need to predict, during the early stages of development, the ADMET properties to increase the success rate of compounds reaching the lead optimization process. Since Lipinski's rule of five, the prediction of pharmacokinetic parameters has evolved towards the current in silico tools based on empirical approaches or molecular modeling. The commercial specialized software for performing such predictions, which is usually costly, is, in many cases, not among the possibilities for research laboratories in academia or at small biotech companies. Nevertheless, in recent years, many free online tools have become available, allowing, more or less accurately, for the prediction of the most relevant pharmacokinetic parameters. This paper studies 18 free web servers capable of predicting ADMET properties and analyzed their advantages and disadvantages, their model-based calculations, and their degree of accuracy by considering the experimental data reported for a set of 24 FDA-approved tyrosine kinase inhibitors (TKIs) as a model of a research project.

Novel Sunifiram-carbamate hybrids as potential dual acetylcholinesterase inhibitor and NMDAR co-agonist: simulation-guided analogue design and pharmacological screening

An efficient method for synthesising NMDAR co-agonist Sunifiram (DM235), in addition to Sunifram-carbamate and anthranilamide hybrids, has been developed in high yields via protecting group-free stepwise unsymmetric diacylation of piperazine using N-acylbenzotiazole. Compounds 3f, 3d, and 3i exhibited promising nootropic activity by enhancing acetylecholine (ACh) release in A549 cell line. Moreover, the carbamate hybrid 3f was found to exhibit higher in vitro potency than donepezil with IC50 = 18 ± 0.2 nM, 29.9 ± 0.15 nM for 3f and donepezil, respectively. 3f was also found to effectively inhibit AChE activity in rat brain (AChE = 1.266 ng/mL) compared to tacrine (AChE = 1.137 ng/ml). An assessment of the ADMET properties revealed that compounds 3f, 3d, and 3i are drug-like and can penetrate blood-brain barrier. Findings presented here showcase highly potential cholinergic agents, with expected partial agonist activity towards glycine binding pocket of NMDAR which could lead to development and optimisation of novel nootropic drugs.

Combined Bioinformatics and Combinatorial Chemistry Tools to Locate Drug-Able Anti-TB Phytochemicals: A Cost-Effective Platform for Natural Product-Based Drug Discovery

Based on extensive experimental studies, a huge number of phytochemicals showed potential activity against tuberculosis (TB) at a lower minimum inhibitory concentration (MIC) and fewer toxicity profiles. However, these promising drugs have not been able to convert from 'lead' to 'mainstream' due to inadequate drug-ability profiles. Thus, early drug-prospective analyses are required at the primary stage to accelerate natural-product-based drug discovery with limited resources and time. In the present study, we have selected seventy-three potential anti-TB phytochemicals (MIC value ≤10 μg/mL) and assessed the drug-ability profiles using bioinformatics and combinatorial chemistry tools, systematically. Primarily, the molecular docking study was done against two putative drug targets, catalase-peroxidase enzyme (katG) and RNA polymerase subunit-β (rpoB) of Mycobacterium tuberculosis (Mtb) using AutoDock 4.2 software. Further, assessed the drug-ability score from Molsoft, toxicity profiles from ProTox, pharmacokinetics from SwisADME, hierarchical cluster analysis (HCA) by ChemMine tools and frontier molecular orbitals (FMOs) with Avogadro and structural activity relationships (SAR) analysis with ChemDraw 18.0 software. Above analyses indicated that, lower MIC exhibited anti-TB phytochemicals, abietane, 12-demethylmulticaulin exhibited poor docking and drug-ability scores, while tiliacorinine, 2-nortiliacorinine showed higher binding energy and drug-ability profiles. Overall, tiliacorinine, 2-nortiliacorinine, 7α-acetoxy-6β-hydroxyroyleanone (AHR), (2S)-naringenin and isovachhalcone were found as the most active and drug-able anti-TB candidates from 73 candidates. Phytochemicals are always a vital source of mainstream drugs, but the MIC value of a phytochemical is not sufficient for it to be promoted. An ideal drug-ability profile is therefore essential for achieving clinical success, where advanced bioinformatics tools help to assess and analyse that profile. Additionally, several natural pharmacophores found in existing anti-TB drugs in SAR analyses also provide crucial information for developing potential anti-TB drug. As a conclusion, combined bioinformatics and combinatorial chemistry are the most effective strategies to locate potent-cum-drug-able candidates in the current drug-development module.

Nano-bio interactions: A major principle in the dynamic biological processes of nano-assemblies

Controllable nano-assembly with stimuli-responsive groups is emerging as a powerful strategy to generate theranostic nanosystems that meet unique requirements in modern medicine. However, this prospective field is still in a proof-of-concept stage due to the gaps in our understanding of complex-(nano-assemblies)-complex-(biosystems) interactions. Indeed, stimuli-responsive assembly-disassembly is, in and of itself, a process of nano-bio interactions, the key steps for biological fate and functional activity of nano-assemblies. To provide a comprehensive understanding of these interactions in this review, we first propose a 4W1H principle (Where, When, What, Which and How) to delineate the relevant dynamic biological processes, behaviour and fate of nano-assemblies. We further summarize several key parameters that govern effective nano-bio interactions. The effects of these kinetic parameters on ADMET processes (absorption, distribution, metabolism, excretion and transformation) are then discussed. Furthermore, we provide an overview of the challenges facing the evaluation of nano-bio interactions of assembled nanodrugs. We finally conclude with future perspectives on safe-by-design and application-driven-design of nano-assemblies. This review will highlight the dynamic biological and physicochemical parameters of nano-bio interactions and bridge discrete concepts to build a full spectrum understanding of the biological outcomes of nano-assemblies. These principles are expected to pave the way for future development and clinical translation of precise, safe and effective nanomedicines with intelligent theranostic features.

Computational exploration of maternal embryonic leucine zipper kinase (MELK) as a cancer drug target

Maternal embryonic leucine zipper kinase (MELK) is of vital importance due to its significant role in cancer development and its association with poor prognosis in different cancers. Here, we employed several computer aided drug design approaches to shortlist potential binding molecules of MELK. For virtual screening, asinex oncology library (containing 6334 drugs) and comprehensive marine natural products database (containing approximately 32,000 drugs) were used. The study identified two drug molecules: Top-2 and Top-3 as high affinity binding MELK molecules compared to the control co-crystalized Top-1 inhibitor. Both the shortlisted compounds and the control showed high stable binding free energy and high GOLD score. The compounds and control also reported stable dynamics with root mean square deviations (RMSD) value ∼ 2 Å in 500 ns. Similarly, the MELK active site residues were observed in good stability with the compounds. Further, it was noticed the compounds/control formed multiple hydrogen bonds with the MELK active pocket residues which is the main reason of high intermolecular stability. Atomic level binding free energies determined van der Waals and electrostatic energies to play vital role in stable complex formation. From drug likeness and pharmacokinetics perspective, the compounds are ideal molecules for further investigation. Overall, the results are promising and might be tested in in vivo and in vitro studies against MELK.

Multi-task convolutional neural networks for predicting in vitro clearance endpoints from molecular images

Optimization of compound metabolic stability is a highly topical issue in pharmaceutical research. Accordingly, application of predictive in silico models can potentially reduce the number of design-make-test-analyze iterations and consequently speed up the progression of novel candidate molecules. Herein, we have investigated the question if multiple in vitro clearance endpoints could be accurately predicted from image-based molecular representations. Thus, compound measurements for four commonly investigated clearance endpoints were curated from AstraZeneca internal sources, providing a sound basis for building multi-task convolutional neural network models. Application of several increasingly challenging data splitting strategies confirmed that convolutional neural network models were successful at capturing implicit chemical relationships contained in training and test data, similar to what is commonly observed for structural fingerprints. Furthermore, model benchmarking against state-of-the-art machine learning methods, including deep neural networks and graph convolutional neural networks, trained with structure- and graph-based representations, respectively, revealed on par or increased accuracy of convolutional neural networks with clear benefit of multi-task learning across all clearance endpoints. Our findings indicate that image-based molecular representations can be applied to predict multiple clearance endpoints, suggesting a potential follow-up to investigate model interpretability from molecular images.

Mesembryanthemum tortuosum L. alkaloids modify anxiety-like behaviour in a zebrafish model

ETHNOPHARMACOLOGICAL RELEVANCE

Mesembryanthemum tortuosum L. (previously known as Sceletium tortuosum (L.) N.E. Br.) is indigenous to South Africa and traditionally used to alleviate anxiety, stress and depression. Mesembrine and its alkaloid analogues such as mesembrenone, mesembrenol and mesembranol have been identified as the key compounds responsible for the reported effects on the central nervous system.

AIM OF THE STUDY

To investigate M. tortuosum alkaloids for possible anxiolytic-like effects in the 5-dpf in vivo zebrafish model by assessing thigmotaxis and locomotor activity.

MATERIALS AND METHODS

Locomotor activity and reverse-thigmotaxis, recognised anxiety-related behaviours in 5-days post fertilization zebrafish larvae, were analysed under simulated stressful conditions of alternating light-dark challenges. Cheminformatics screening and molecular docking were also performed to rationalize the inhibitory activity of the alkaloids on the serotonin reuptake transporter, the accepted primary mechanism of action of selective serotonin reuptake inhibitors. Mesembrine has been reported to have inhibitory effects on serotonin reuptake, with consequential anti-depressant and anxiolytic effects.

RESULTS

All four alkaloids assessed decreased the anxiety-related behaviour of zebrafish larvae exposed to the light-dark challenge. Significant increases in the percentage of time spent in the central arena during the dark phase were also observed when larvae were exposed to the pure alkaloids (mesembrenone, mesembrenol, mesembrine and mesembrenol) compared to the control. However, mesembrenone and mesembranol demonstrated a greater anxiolytic-like effect than the other alkaloids. In addition to favourable pharmacokinetic and physicochemical properties revealed via in silico predictions, high-affinity interactions characterized the binding of the alkaloids with the serotonin transporter.

CONCLUSIONS

M. tortuosum alkaloids demonstrated an anxiolytic-like effect in zebrafish larvae providing evidence for its traditional and modern day use as an anxiolytic.

Computational targeting of allosteric site of MEK1 by quinoline-based molecules

MEK1 is an attractive target due to its role in selective extracellular-signal-regulated kinase phosphorylation, which plays a pivotal role in regulating cell proliferation. Another benefit of targeting the MEK protein is its unique hydrophobic pocket that can accommodate highly selective allosteric inhibitors. To date, various MEK1 inhibitors have reached clinical trials against several cancers, but they were discarded due to their severe toxicity and low efficacy. Thus, the development of allosteric inhibitors for MEK1 is the demand of the hour. In this in-silico study, molecular docking, long-term molecular dynamics (5 µs), and molecular mechanics Poisson-Boltzmann surface area analysis were undertaken to address the potential of quinolines as allosteric inhibitors. We selected four reference MEK1 inhibitors for the comparative analysis. The drug-likeness and toxicity of these molecules were also examined based on their ADMET and Toxicity Prediction by Komputer Assisted Technology profiles. The outcome of the analysis revealed that the quinolines (4m, 4o, 4s, and 4n) exhibited better stability and binding affinity while being nontoxic compared to reference inhibitors. We have reached the conclusion that these quinoline molecules could be checked by experimental studies to validate their use as allosteric inhibitors against MEK1.

Application of Physiologically Based Pharmacokinetic Modeling in Preclinical Studies: A Feasible Strategy to Practice the Principles of 3Rs

Pharmacokinetic characterization plays a vital role in drug discovery and development. Although involving numerous laboratory animals with error-prone, labor-intensive, and time-consuming procedures, pharmacokinetic profiling is still irreplaceable in preclinical studies. With physiologically based pharmacokinetic (PBPK) modeling, the in vivo profiles of drug absorption, distribution, metabolism, and excretion can be predicted. To evaluate the application of such an approach in preclinical investigations, the plasma pharmacokinetic profiles of seven commonly used probe substrates of microsomal enzymes, including phenacetin, tolbutamide, omeprazole, metoprolol, chlorzoxazone, nifedipine, and baicalein, were predicted in rats using bottom-up PBPK models built with in vitro data alone. The prediction's reliability was assessed by comparison with in vivo pharmacokinetic data reported in the literature. The overall predicted accuracy of PBPK models was good with most fold errors within 2, and the coefficient of determination (R2) between the predicted concentration data and the observed ones was more than 0.8. Moreover, most of the observation dots were within the prediction span of the sensitivity analysis. We conclude that PBPK modeling with acceptable accuracy may be incorporated into preclinical studies to refine in vivo investigations, and PBPK modeling is a feasible strategy to practice the principles of 3Rs.

Current status and future directions for a neurotoxicity hazard assessment framework that integrates in silico approaches

Neurotoxicology is the study of adverse effects on the structure or function of the developing or mature adult nervous system following exposure to chemical, biological, or physical agents. The development of more informative alternative methods to assess developmental (DNT) and adult (NT) neurotoxicity induced by xenobiotics is critically needed. The use of such alternative methods including in silico approaches that predict DNT or NT from chemical structure (e.g., statistical-based and expert rule-based systems) is ideally based on a comprehensive understanding of the relevant biological mechanisms. This paper discusses known mechanisms alongside the current state of the art in DNT/NT testing. In silico approaches available today that support the assessment of neurotoxicity based on knowledge of chemical structure are reviewed, and a conceptual framework for the integration of in silico methods with experimental information is presented. Establishing this framework is essential for the development of protocols, namely standardized approaches, to ensure that assessments of NT and DNT based on chemical structures are generated in a transparent, consistent, and defendable manner.

Application of Artificial Intelligence and Machine Learning in Drug Discovery

Machine Learning (ML) and Deep Learning (DL) are two subclasses of Artificial Intelligence (AI), that, in this day and age of big data provides significant opportunities to pharmaceutical discovery research and development by translating data to information and ultimately to knowledge. Machine Learning or AI is not really new but over last few years, application of better methods have emerged and they have been successfully applied for drug discovery and development. This chapter would provide an overview of these methods and how they have been applied across various work streams, e.g., generative chemistry, ADMET prediction, retrosynthetic analysis, etc. within drug discovery process. This chapter would also attempt to provide caution and pit falls in utilizing these methods blindly while summarizing challenges and limitations.

Efficient Hit-to-Lead Searching of Kinase Inhibitor Chemical Space via Computational Fragment Merging

In early-stage drug discovery, the hit-to-lead optimization (or "hit expansion") stage entails starting from a newly identified active compound and improving its potency or other properties. Traditionally, this process relies on synthesizing and evaluating a series of analogues to build up structure-activity relationships. Here, we describe a computational strategy focused on kinase inhibitors, intended to expedite the process of identifying analogues with improved potency. Our protocol begins from an inhibitor of the target kinase and generalizes the synthetic route used to access it. By searching for commercially available replacements for the individual building blocks used to make the parent inhibitor, we compile an enumerated library of compounds that can be accessed using the same chemical transformations; these huge libraries can exceed many millions─or billions─of compounds. Because the resulting libraries are much too large for explicit virtual screening, we instead consider alternate approaches to identify the top-scoring compounds. We find that contributions from individual substituents are well described by a pairwise additivity approximation, provided that the corresponding fragments position their shared core in precisely the same way relative to the binding site. This key insight allows us to determine which fragments are suitable for merging into single new compounds and which are not. Further, the use of pairwise approximation allows interaction energies to be assigned to each compound in the library without the need for any further structure-based modeling: interaction energies instead can be reliably estimated from the energies of the component fragments, and the reduced computational requirements allow for flexible energy minimizations that allow the kinase to respond to each substitution. We demonstrate this protocol using libraries built from six representative kinase inhibitors drawn from the literature, which target five different kinases: CDK9, CHK1, CDK2, EGFR^T790M, and ACK1. In each example, the enumerated library includes additional analogues reported by the original study to have activity, and these analogues are successfully prioritized within the library. We envision that the insights from this work can facilitate the rapid assembly and screening of increasingly large libraries for focused hit-to-lead optimization. To enable adoption of these methods and to encourage further analyses, we disseminate the computational tools needed to deploy this protocol.

Synthesis and potential vasorelaxant effect of a novel ruthenium-based nitro complex

This study aimed to investigate the synthesis and potential vasodilator effect of a novel ruthenium complex, cis-[Ru(bpy)₂(2-MIM)(NO₂)]PF₆ (bpy = 2,2'-bipyridine and 2-MIM = 2-methylimidazole) (FOR711A), containing an imidazole derivative via an in silico molecular docking model using β1 H-NOX (Heme-nitric oxide/oxygen binding) domain proteins of reduced and oxidized soluble guanylate cyclase (sGC). In addition, pharmacokinetic properties in the human organism were predicted through computational simulations and the potential for acute irritation of FOR711A was also investigated in vitro using the hen's egg chorioallantoic membrane (HET-CAM). FOR711A interacted with sites of the β1 H-NOX domain of reduced and oxidized sGC, demonstrating shorter bond distances to several residues and negative values of total energy. The predictive study revealed molar refractivity (RM): 127.65; Log Po/w = 1.29; topological polar surface area (TPSA): 86.26 Å²; molar mass (MM) = 541.55 g/mol; low solubility, high unsaturation index, high gastrointestinal absorption; toxicity class 4; failure to cross the blood-brain barrier and to react with cytochrome P450 (CYP) enzymes CYP1A2, CYP2C19, CYP2C9, CYP2D6 and CYP3A4. After the HET-CAM assay, the FOR711A complex was classified as non-irritant (N.I.) and its vasodilator effect was confirmed through greater evidence of blood vessels after the administration and ending of the observation period of 5 min. These results suggest that FOR711A presented a potential stimulator/activator effect of sGC via NO/sGC/cGMP. However, results indicate it needs a vehicle for oral administration.

Enhancing preclinical drug discovery with artificial intelligence

Artificial intelligence (AI) is becoming an integral part of drug discovery. It has the potential to deliver across the drug discovery and development value chain, starting from target identification and reaching through clinical development. In this review, we provide an overview of current AI technologies and a glimpse of how AI is reimagining preclinical drug discovery by highlighting examples where AI has made a real impact. Considering the excitement and hyperbole surrounding AI in drug discovery, we aim to present a realistic view by discussing both opportunities and challenges in adopting AI in drug discovery.

Computational screening of phytochemicals from three medicinal plants as inhibitors of transmembrane protease serine 2 implicated in SARS-CoV-2 infection

Background

SARS-CoV-2 infection or COVID-19 is a major global public health issue that requires urgent attention in terms of drug development. Transmembrane Protease Serine 2 (TMPRSS2) is a good drug target against SARS-CoV-2 because of the role it plays during the viral entry into the cell. Virtual screening of phytochemicals as potential inhibitors of TMPRSS2 can lead to the discovery of drug candidates for the treatment of COVID-19.

Purpose

The study was designed to screen 132 phytochemicals from three medicinal plants traditionally used as antivirals; Zingiber officinalis Roscoe (Zingiberaceae), Artemisia annua L. (Asteraceae), and Moringa oleifera Lam. (Moringaceae), as potential inhibitors of TMPRSS2 for the purpose of finding therapeutic options to treat COVID-19.

Methods

Homology model of TMPRSS2 was built using the ProMod3 3.1.1 program of the SWISS-MODEL. Binding affinities and interaction between compounds and TMPRSS2 model was examined using molecular docking and molecular dynamics simulation. The drug-likeness and ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties of potential inhibitors of TMPRSS2 were also assessed using admetSAR web tool.

Results

Three compounds, namely, niazirin, quercetin, and moringyne from M. oleifera demonstrated better molecular interactions with binding affinities ranging from -7.1 to -8.0 kcal/mol compared to -7.0 kcal/mol obtained for camostat mesylate (a known TMPRSS2 inhibitor), which served as a control. All the three compounds exhibited good drug-like properties by not violating the Lipinski rule of 5. Niazirin and moringyne possessed good ADMET properties and were stable in their interactions with the TMPRSS2 based on the molecular dynamics simulation. However, the ADMET tool predicted the potential hepatotoxic and mutagenic effects of quercetin.

Conclusion

This study demonstrated the potentials of niazirin, quercetin, and moringyne from M. oleifera, to inhibit the activities of human TMPRSS2, thus probably being good candidates for further development as new drugs for the treatment or management of COVID-19.

A Novel Method for Predicting the Human Inherent Clearance and Its Application in the Study of the Pharmacokinetics and Drug-Drug Interaction between Azidothymidine and Fluconazole Mediated by UGT Enzyme

In order to improve the benefit–risk ratio of pharmacokinetic (PK) research in the early development of new drugs, in silico and in vitro methods were constructed and improved. Models of intrinsic clearance rate (CL_int) were constructed based on the quantitative structure–activity relationship (QSAR) of 7882 collected compounds. Moreover, a novel in vitro metabolic method, the Bio-PK dynamic metabolic system, was constructed and combined with a physiology-based pharmacokinetic model (PBPK) model to predict the metabolism and the drug–drug interaction (DDI) of azidothymidine (AZT) and fluconazole (FCZ) mediated by the phase II metabolic enzyme UDP-glycosyltransferase (UGT) in humans. Compared with the QSAR models reported previously, the goodness of fit of our CL_int model was slightly improved (determination coefficient (R²) = 0.58 vs. 0.25–0.45). Meanwhile, compared with the predicted clearance of 61.96 L/h (fold error: 2.95–3.13) using CL_int (8 µL/min/mg) from traditional microsomal experiment, the predicted clearance using CL_int (25 μL/min/mg) from Bio-PK system was increased to 143.26 L/h (fold error: 1.27–1.36). The predicted C_max and AUC (the area under the concentration–time curve) ratio were 1.32 and 1.84 (fold error: 1.36 and 1.05) in a DDI study with an inhibition coefficient (Ki) of 13.97 μM from the Bio-PK system. The results indicate that the Bio-PK system more truly reflects the dynamic metabolism and DDI of AZT and FCZ in the body. In summary, the novel in silico and in vitro method may provide new ideas for the optimization of drug metabolism and DDI research methods in early drug development.

Synergy conformal prediction applied to large-scale bioactivity datasets and in federated learning

Confidence predictors can deliver predictions with the associated confidence required for decision making and can play an important role in drug discovery and toxicity predictions. In this work we investigate a recently introduced version of conformal prediction, synergy conformal prediction, focusing on the predictive performance when applied to bioactivity data. We compare the performance to other variants of conformal predictors for multiple partitioned datasets and demonstrate the utility of synergy conformal predictors for federated learning where data cannot be pooled in one location. Our results show that synergy conformal predictors based on training data randomly sampled with replacement can compete with other conformal setups, while using completely separate training sets often results in worse performance. However, in a federated setup where no method has access to all the data, synergy conformal prediction is shown to give promising results. Based on our study, we conclude that synergy conformal predictors are a valuable addition to the conformal prediction toolbox.

Application of Deep Neural Network Models in Drug Discovery Programs

In silico driven optimization of compound properties related to pharmacokinetics, pharmacodynamics, and safety is a key requirement in modern drug discovery. Nowadays, large and harmonized datasets allow to implement deep neural networks (DNNs) as a framework for leveraging predictive models. Nevertheless, various available model architectures differ in their global applicability and performance in lead optimization projects, such as stability over time and interpretability of the results. Here, we describe and compare the value of established DNN-based methods for the prediction of key ADME property trends and biological activity in an industrial drug discovery environment, represented by microsomal lability, CYP3A4 inhibition and factor Xa inhibition. Three architectures are exemplified, our earlier described multilayer perceptron approach (MLP), graph convolutional network-based models (GCN) and a vector representation approach, Mol2Vec. From a statistical perspective, MLP and GCN were found to perform superior over Mol2Vec, when applied to external validation sets. Interestingly, GCN-based predictions are most stable over a longer period in a time series validation study. Apart from those statistical observations, DNN prove of value to guide local SAR. To illustrate this important aspect in pharmaceutical research projects, we discuss challenging applications in medicinal chemistry towards a more realistic picture of artificial intelligence in drug discovery.

Designing small molecules for therapeutic success: A contemporary perspective

Successful small-molecule drug design requires a molecular target with inherent therapeutic potential and a molecule with the right properties to unlock its potential. Present-day drug design strategies have evolved to leave little room for improvement in drug-like properties. As a result, inadequate safety or efficacy associated with molecular targets now constitutes the primary cause of attrition in preclinical development through Phase II. This finding has led to a deeper focus on target selection. In this current reality, design tactics that enable rapid identification of risk-balanced clinical candidates, translation of clinical experience into meaningful differentiation strategies, and expansion of the druggable proteome represent significant levers by which drug designers can accelerate the discovery of the next generation of medicines.

How can SHAP values help to shape metabolic stability of chemical compounds?

Background

Computational methods support nowadays each stage of drug design campaigns. They assist not only in the process of identification of new active compounds towards particular biological target, but also help in the evaluation and optimization of their physicochemical and pharmacokinetic properties. Such features are not less important in terms of the possible turn of a compound into a future drug than its desired affinity profile towards considered proteins. In the study, we focus on metabolic stability, which determines the time that the compound can act in the organism and play its role as a drug. Due to great complexity of xenobiotic transformation pathways in the living organisms, evaluation and optimization of metabolic stability remains a big challenge.

Results

Here, we present a novel methodology for the evaluation and analysis of structural features influencing metabolic stability. To this end, we use a well-established explainability method called SHAP. We built several predictive models and analyse their predictions with the SHAP values to reveal how particular compound substructures influence the model’s prediction. The method can be widely applied by users thanks to the web service, which accompanies the article. It allows a detailed analysis of SHAP values obtained for compounds from the ChEMBL database, as well as their determination and analysis for any compound submitted by a user. Moreover, the service enables manual analysis of the possible structural modifications via the provision of analogous analysis for the most similar compound from the ChEMBL dataset.

Conclusions

To our knowledge, this is the first attempt to employ SHAP to reveal which substructural features are utilized by machine learning models when evaluating compound metabolic stability. The accompanying web service for metabolic stability evaluation can be of great help for medicinal chemists. Its significant usefulness is related not only to the possibility of assessing compound stability, but also to the provision of information about substructures influencing this parameter. It can assist in the design of new ligands with improved metabolic stability, helping in the detection of privileged and unfavourable chemical moieties during stability optimization. The tool is available at https://metstab-shap.matinf.uj.edu.pl/.

Development and implementation of an enterprise-wide predictive model for early absorption, distribution, metabolism and excretion properties

Background: Accurate prediction of absorption, distribution, metabolism and excretion (ADME) properties can facilitate the identification of promising drug candidates. Methodology & Results: The authors present the Janssen generic Target Product Profile (gTPP) model, which predicts 18 early ADME properties, employs a graph convolutional neural network algorithm and was trained on between 1000-10,000 internal data points per predicted parameter. gTPP demonstrated stronger predictive power than pretrained commercial ADME models and automatic model builders. Through a novel logging method, the authors report gTPP usage for more than 200 Janssen drug discovery scientists. Conclusion: The investigators successfully enabled the rapid and systematic implementation of predictive ML tools across a drug discovery pipeline in all therapeutic areas. This experience provides useful guidance for other large-scale AI/ML deployment efforts.

Prediction of Molecular Properties Using Molecular Topographic Map

Prediction of molecular properties plays a critical role towards rational drug design. In this study, the Molecular Topographic Map (MTM) is proposed, which is a two-dimensional (2D) map that can be used to represent a molecule. An MTM is generated from the atomic features set of a molecule using generative topographic mapping and is then used as input data for analyzing structure-property/activity relationships. In the visualization and classification of 20 amino acids, differences of the amino acids can be visually confirmed from and revealed by hierarchical clustering with a similarity matrix of their MTMs. The prediction of molecular properties was performed on the basis of convolutional neural networks using MTMs as input data. The performance of the predictive models using MTM was found to be equal to or better than that using Morgan fingerprint or MACCS keys. Furthermore, data augmentation of MTMs using mixup has improved the prediction performance. Since molecules converted to MTMs can be treated like 2D images, they can be easily used with existing neural networks for image recognition and related technologies. MTM can be effectively utilized to predict molecular properties of small molecules to aid drug discovery research.

Design, synthesis and pharmacological screening of novel renoprotective methionine-based peptidomimetics: Amelioration of cisplatin-induced nephrotoxicity

Cisplatin (CP) is an effective chemotherapeutic agent for treatment of various types of cancer, however efforts are needed to reduce its toxic side effect. Previous studies revealed promising effect of peptides in decreasing CP induced nephrotoxicity. Herein, novel Met-based peptidomimetics were synthesized using N-acylbenzotriazole as acylating agent in high yield. Evaluation of renoprotective effect of the synthesized targets on CP treated kidney cell line (LLC-PK1) revealed that pretreatment with 1/3 IC₅₀ of targets II, IIIa-g attenuated CP induced cell death where the IC₅₀ of CP was raised from 3.28 µM to 9.25-41.1 µM. The most potent compounds IIIg, II and IIIb exhibited antioxidative stress in CP-treated LLC-PK1 cells as confirmed by raising GSH/GSSG ratio and SOD concentration as well as decreasing ROS and MDA. Additionally, in vivo experiments using Sprague Dawley rats showed renoprotective effect of IIIg against CP-induced nephrotoxicity as evidenced by improved results of renal function tests and attenuated CP-induced renal structural injury. Moreover, antioxidant activity of IIIg was demonstrated via its ability to reduce renal MDA level and up-regulate renal antioxidant element GSH level. Further, immunohistochemistry of renal specimens showed the ability of IIIg to restore CP-induced suppression of Nrf2. Interestingly, in vivo and in vitro studies demonstrated that IIIg had no effect on CP antiproliferative activity. An assessment of the ADMET properties revealed that targets IIIg, II and IIIb showed good drug-likeness in terms of their physicochemical, pharmacokinetic properties. The findings presented here showcase that IIIg is a promising renoprotective candidate with antioxidative stress potential.

A computational approach for rational discovery of inhibitors for non-structural protein 1 of SARS-CoV-2

Background

Non-structural protein 1 (Nsp1), a virulence agent of SARS-CoV-2, has emerged as an important target for drug discovery. Nsp1 shuts down the host gene function by associating with the 40S ribosomal subunit.

Methods

Molecular interactions, drug-likeness, physiochemical property predictions, and robust molecular dynamics (MD) simulations were employed to discover novel Nsp1 inhibitors. In this study, we evaluated a series of molecules based on the plant (Cedrus deodara) derived α,β,γ-Himachalenes scaffolds.

Results

The results obtained from estimated affinity and ligand efficiency suggested that BCH10, BCH15, BCH16, and BCH17 could act as potential inhibitors of Nsp1. Moreover, MD simulations comprising various MD driven time-dependent analyses and thermodynamic free energy calculations also suggested stable protein-ligand complexes and strong interactions with the binding site. Furthermore, the selected molecules passed drug likeliness parameters and the physiochemical property analysis showed acceptable bioactivity scores.

Conclusion

The structural parameters of dynamic simulations revealed that the reported molecules could act as lead compounds against SARS-CoV-2 Nsp1 protein.

Direct Comparison of the Prediction of the Unbound Brain-to-Plasma Partitioning Utilizing Machine Learning Approach and Mechanistic Neuropharmacokinetic Model

The mechanistic neuropharmacokinetic (neuroPK) model was established to predict unbound brain-to-plasma partitioning (K_p,uu,brain) by considering in vitro efflux activities of multiple drug resistance 1 (MDR1) and breast cancer resistance protein (BCRP). Herein, we directly compare this model to a computational machine learning approach utilizing physicochemical descriptors and efflux ratios of MDR1 and BCRP-expressing cells for predicting K_p,uu,brain in rats. Two different types of machine learning techniques, Gaussian processes (GP) and random forest regression (RF), were assessed by the time and cluster-split validation methods using 640 internal compounds. The predictivity of machine learning models based on only molecular descriptors in the time-split dataset performed worse than the cluster-split dataset, whereas the models incorporating MDR1 and BCRP efflux ratios showed similar predictivity between time and cluster-split datasets. The GP incorporating MDR1 and BCRP in the time-split dataset achieved the highest correlation (R² = 0.602). These results suggested that incorporation of MDR1 and BCRP in machine learning is beneficial for robust and accurate prediction. K_p,uu,brain prediction utilizing the neuroPK model was significantly worse compared to machine learning approaches for the same dataset. We also investigated the predictivity of K_p,uu,brain using an external independent test set of 34 marketed drugs. Compared to machine learning models, the neuroPK model showed better predictive performance with R² of 0.577. This work demonstrates that the machine learning model for K_p,uu,brain achieves maximum predictive performance within the chemical applicability domain, whereas the neuroPK model is applicable more widely beyond the chemical space covered in the training dataset.

Rational design of potent anti-COVID-19 main protease drugs: An extensive multi-spectrum in silico approach

The emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) as a novel coronavirus and the etiological agent of global pandemic coronavirus disease (COVID-19) requires quick development of potential therapeutic strategies. Computer aided drug design approaches are highly efficient in identifying promising drug candidates among an available pool of biological active antivirals with safe pharmacokinetics. The main protease (M^Pro) enzyme of SARS-CoV-2 is considered key in virus production and its crystal structures are available at excellent resolution. This marks the enzyme as a good starting receptor to conduct an extensive structure-based virtual screening (SBVS) of ASINEX antiviral library for the purpose of uncovering valuable hits against SARS-CoV-2 M^Pro. A compound hit (BBB_26580140) was stand out in the screening process, as opposed to the control, as a potential inhibitor of SARS-CoV-2 M^Pro based on a combined approach of SBVS, drug likeness and lead likeness annotations, pharmacokinetics, molecular dynamics (MD) simulations, and end point MM-PBSA binding free energy methods. The lead was further used in ligand-based similarity search (LBSS) that found 33 similar compounds from the ChEMBL database. A set of three compounds (SCHEMBL12616233, SCHEMBL18616095, and SCHEMBL20148701), based on their binding affinity for M^Pro, was selected and analyzed using extensive MD simulation, hydrogen bond profiling, MM-PBSA, and WaterSwap binding free energy techniques. The compounds conformation with M^Pro show good stability after initial within active cavity moves, a rich intermolecular network of chemical interactions, and reliable relative and absolute binding free energies. Findings of the study suggested the use of BBB_26580140 lead and its similar analogs to be explored in vivo which might pave the path for rational drug discovery against SARS-CoV-2 M^Pro.