Dimorphite-DL: an open-source program for enumerating the ionization states of drug-like small molecules

Elucidating the linagliptin and fibroblast activation protein binding mechanism through molecular dynamics and binding free energy analysis

Fibroblast activation protein (FAP) is highly expressed in solid tumors and may be a potential diagnostic and therapeutic target in solid cancers. Linagliptin inhibits FAP; however, the interaction mechanism between linagliptin and FAP remains unclear. In this study, the binding free energy for linagliptin with human FAP was estimated at -13.66 kcal/mol, and the dissociation constant was 243 nM based on surface plasmon resonance analyses. E203, E204, and Y656 formed hydrogen bonds with ammonium. Y625 formed an unstable hydrogen bond with the carbonyl group. W623 and Y541 interacted with the quinazoline and pyrimidine-2,4-dione rings, respectively, via π-π interactions. The butyne group formed hydrophobic interactions with residues V650, Y653, Y656, and Y660. ZINC000299754517 and ZINC000299754576 were identified as potential FAP inhibitors. The R1 and R4 regions of linagliptin could be optimized to increase its FAP binding affinity. These findings can guide linagliptin structural optimization to improve its FAP binding affinity.

Exploring Chemical Spaces in the Billion Range: Is Docking a Computational Alternative to DNA-Encoded Libraries?

The concept of DNA-encoded libraries (DELs) enables the experimental screening of billions of compounds simultaneously, offering an unprecedented boost in the coverage of chemical space. In parallel, however, dramatically increased access to supercomputers and a number of ultrahigh throughput virtual screening (uHTVS) tools have made screening of billion-membered virtual libraries available. Here, we investigate whether current, brute-force, or AI-enabled uHTVS approaches might constitute a computational alternative to DEL screening. While it is tempting to look at uHTVS as a computational analogue of DEL screening, we found specific advantages and limitations of both methodologies that suggest them being complementary rather than competitive.

Widespread Misinterpretation of pKa Terminology for Zwitterionic Compounds and Its Consequences

The acid dissociation constant (pKa), which quantifies the propensity for a solute to donate a proton to its solvent, is crucial for drug design and synthesis, environmental fate studies, chemical manufacturing, and many other fields. Unfortunately, the terminology used for describing acid-base phenomena is sometimes inconsistent, causing large potential for misinterpretation. In this work, we examine a systematic confusion underlying the definition of "acidic" and "basic" pKa values for zwitterionic compounds. Due to this confusion, some pKa data are misrepresented in data repositories, including the widely used and highly trusted ChEMBL database. Such datasets are frequently used to supply training data for pKa prediction models, and hence, confusion and errors in the data make the model performance worse. Herein, we discuss the intricacies of this issue. We make suggestions for describing acid-base phenomena, training pKa prediction models, and stewarding pKa datasets, given the high potential for confusion and potentially high impact in downstream applications.

Data-driven discovery of potent small molecule ice recrystallisation inhibitors

Controlling the formation and growth of ice is essential to successfully cryopreserve cells, tissues and biologics. Current efforts to identify materials capable of modulating ice growth are guided by iterative changes and human intuition, with a major focus on proteins and polymers. With limited data, the discovery pipeline is constrained by a poor understanding of the mechanisms and the underlying structure-activity relationships. In this work, this barrier is overcome by constructing machine learning models capable of predicting the ice recrystallisation inhibition activity of small molecules. We generate a new dataset via experimental measurements of ice growth, then harness predictive models combining state-of-the-art descriptors with domain-specific features derived from molecular simulations. The models accurately identify potent small molecule ice recrystallisation inhibitors within a commercial compound library. Identified hits can also mitigate cellular damage during transient warming events in cryopreserved red blood cells, demonstrating how data-driven approaches can be used to discover innovative cryoprotectants and enable next-generation cryopreservation solutions for the cold chain.

The changing scenario of drug discovery using AI to deep learning: Recent advancement, success stories, collaborations, and challenges

Due to the transformation of artificial intelligence (AI) tools and technologies, AI-driven drug discovery has come to the forefront. It reduces the time and expenditure. Due to these advantages, pharmaceutical industries are concentrating on AI-driven drug discovery. Several drug molecules have been discovered using AI-based techniques and tools, and several newly AI-discovered drug molecules have already entered clinical trials. In this review, we first present the data and their resources in the pharmaceutical sector for AI-driven drug discovery and illustrated some significant algorithms or techniques used for AI and ML which are used in this field. We gave an overview of the deep neural network (NN) models and compared them with artificial NNs. Then, we illustrate the recent advancement of the landscape of drug discovery using AI to deep learning, such as the identification of drug targets, prediction of their structure, estimation of drug-target interaction, estimation of drug-target binding affinity, design of de novo drug, prediction of drug toxicity, estimation of absorption, distribution, metabolism, excretion, toxicity; and estimation of drug-drug interaction. Moreover, we highlighted the success stories of AI-driven drug discovery and discussed several collaboration and the challenges in this area. The discussions in the article will enrich the pharmaceutical industry.

Structure-Aided Computational Design of Triazole-Based Targeted Covalent Inhibitors of Cruzipain

Cruzipain (CZP), the major cysteine protease present in T. cruzi, the ethiological agent of Chagas disease, has attracted particular attention as a therapeutic target for the development of targeted covalent inhibitors (TCI). The vast chemical space associated with the enormous molecular diversity feasible to explore by means of modern synthetic approaches allows the design of CZP inhibitors capable of exhibiting not only an efficient enzyme inhibition but also an adequate translation to anti-T. cruzi activity. In this work, a computer-aided design strategy was developed to combinatorially construct and screen large libraries of 1,4-disubstituted 1,2,3-triazole analogues, further identifying a selected set of candidates for advancement towards synthetic and biological activity evaluation stages. In this way, a virtual molecular library comprising more than 75 thousand diverse and synthetically feasible analogues was studied by means of molecular docking and molecular dynamic simulations in the search of potential TCI of CZP, guiding the synthetic efforts towards a subset of 48 candidates. These were synthesized by applying a Cu(I)-catalyzed azide-alkyne cycloaddition (CuAAC) centered synthetic scheme, resulting in moderate to good yields and leading to the identification of 12 hits selectively inhibiting CZP activity with IC50 in the low micromolar range. Furthermore, four triazole derivatives showed good anti-T. cruzi inhibition when studied at 50 μM; and Ald-6 excelled for its high antitrypanocidal activity and low cytotoxicity, exhibiting complete in vitro biological activity translation from CZP to T. cruzi. Overall, not only Ald-6 merits further advancement to preclinical in vivo studies, but these findings also shed light on a valuable chemical space where molecular diversity might be explored in the search for efficient triazole-based antichagasic agents.

An artificial intelligence accelerated virtual screening platform for drug discovery

Structure-based virtual screening is a key tool in early drug discovery, with growing interest in the screening of multi-billion chemical compound libraries. However, the success of virtual screening crucially depends on the accuracy of the binding pose and binding affinity predicted by computational docking. Here we develop a highly accurate structure-based virtual screen method, RosettaVS, for predicting docking poses and binding affinities. Our approach outperforms other state-of-the-art methods on a wide range of benchmarks, partially due to our ability to model receptor flexibility. We incorporate this into a new open-source artificial intelligence accelerated virtual screening platform for drug discovery. Using this platform, we screen multi-billion compound libraries against two unrelated targets, a ubiquitin ligase target KLHDC2 and the human voltage-gated sodium channel NaV1.7. For both targets, we discover hit compounds, including seven hits (14% hit rate) to KLHDC2 and four hits (44% hit rate) to NaV1.7, all with single digit micromolar binding affinities. Screening in both cases is completed in less than seven days. Finally, a high resolution X-ray crystallographic structure validates the predicted docking pose for the KLHDC2 ligand complex, demonstrating the effectiveness of our method in lead discovery.

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted In Vivo and In Vitro Data

Drug-induced liver injury (DILI) has been a significant challenge in drug discovery, often leading to clinical trial failures and necessitating drug withdrawals. Over the last decade, the existing suite of in vitro proxy-DILI assays has generally improved at identifying compounds with hepatotoxicity. However, there is considerable interest in enhancing the in silico prediction of DILI because it allows for evaluating large sets of compounds more quickly and cost-effectively, particularly in the early stages of projects. In this study, we aim to study ML models for DILI prediction that first predict nine proxy-DILI labels and then use them as features in addition to chemical structural features to predict DILI. The features include in vitro (e.g., mitochondrial toxicity, bile salt export pump inhibition) data, in vivo (e.g., preclinical rat hepatotoxicity studies) data, pharmacokinetic parameters of maximum concentration, structural fingerprints, and physicochemical parameters. We trained DILI-prediction models on 888 compounds from the DILI data set (composed of DILIst and DILIrank) and tested them on a held-out external test set of 223 compounds from the DILI data set. The best model, DILIPredictor, attained an AUC-PR of 0.79. This model enabled the detection of the top 25 toxic compounds (2.68 LR+, positive likelihood ratio) compared to models using only structural features (1.65 LR+ score). Using feature interpretation from DILIPredictor, we identified the chemical substructures causing DILI and differentiated cases of DILI caused by compounds in animals but not in humans. For example, DILIPredictor correctly recognized 2-butoxyethanol as nontoxic in humans despite its hepatotoxicity in mice models. Overall, the DILIPredictor model improves the detection of compounds causing DILI with an improved differentiation between animal and human sensitivity and the potential for mechanism evaluation. DILIPredictor required only chemical structures as input for prediction and is publicly available at https://broad.io/DILIPredictor for use via web interface and with all code available for download.

ChemoDOTS: a web server to design chemistry-driven focused libraries

In drug discovery, the successful optimization of an initial hit compound into a lead molecule requires multiple cycles of chemical modification. Consequently, there is a need to efficiently generate synthesizable chemical libraries to navigate the chemical space surrounding the primary hit. To address this need, we introduce ChemoDOTS, an easy-to-use web server for hit-to-lead chemical optimization freely available at https://chemodots.marseille.inserm.fr/. With this tool, users enter an activated form of the initial hit molecule then choose from automatically detected reactive functions. The server proposes compatible chemical transformations via an ensemble of encoded chemical reactions widely used in the pharmaceutical industry during hit-to-lead optimization. After selection of the desired reactions, all compatible chemical building blocks are automatically coupled to the initial hit to generate a raw chemical library. Post-processing filters can be applied to extract a subset of compounds with specific physicochemical properties. Finally, explicit stereoisomers and tautomers are computed, and a 3D conformer is generated for each molecule. The resulting virtual library is compatible with most docking software for virtual screening campaigns. ChemoDOTS rapidly generates synthetically feasible, hit-focused, large, diverse chemical libraries with finely-tuned physicochemical properties via a user-friendly interface providing a powerful resource for researchers engaged in hit-to-lead optimization.

MolModa: accessible and secure molecular docking in a web browser

Molecular docking advances early-stage drug discovery by predicting the geometries and affinities of small-molecule compounds bound to drug-target receptors, predictions that researchers can leverage in prioritizing drug candidates for experimental testing. Unfortunately, existing docking tools often suffer from poor usability, data security, and maintainability, limiting broader adoption. Additionally, the complexity of the docking process, which requires users to execute a series of specialized steps, often poses a substantial barrier for non-expert users. Here, we introduce MolModa, a secure, accessible environment where users can perform molecular docking entirely in their web browsers. We provide two case studies that illustrate how MolModa provides valuable biological insights. We further compare MolModa to other docking tools to highlight its strengths and limitations. MolModa is available free of charge for academic and commercial use, without login or registration, at https://durrantlab.com/molmoda.

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted in vivo and in vitro Data

Drug-induced liver injury (DILI) has been significant challenge in drug discovery, often leading to clinical trial failures and necessitating drug withdrawals. The existing suite of in vitro proxy-DILI assays is generally effective at identifying compounds with hepatotoxicity. However, there is considerable interest in enhancing in silico prediction of DILI because it allows for the evaluation of large sets of compounds more quickly and cost-effectively, particularly in the early stages of projects. In this study, we aim to study ML models for DILI prediction that first predicts nine proxy-DILI labels and then uses them as features in addition to chemical structural features to predict DILI. The features include in vitro (e.g., mitochondrial toxicity, bile salt export pump inhibition) data, in vivo (e.g., preclinical rat hepatotoxicity studies) data, pharmacokinetic parameters of maximum concentration, structural fingerprints, and physicochemical parameters. We trained DILI-prediction models on 888 compounds from the DILIst dataset and tested on a held-out external test set of 223 compounds from DILIst dataset. The best model, DILIPredictor, attained an AUC-ROC of 0.79. This model enabled the detection of top 25 toxic compounds compared to models using only structural features (2.68 LR+ score). Using feature interpretation from DILIPredictor, we were able to identify the chemical substructures causing DILI as well as differentiate cases DILI is caused by compounds in animals but not in humans. For example, DILIPredictor correctly recognized 2-butoxyethanol as non-toxic in humans despite its hepatotoxicity in mice models. Overall, the DILIPredictor model improves the detection of compounds causing DILI with an improved differentiation between animal and human sensitivity as well as the potential for mechanism evaluation. DILIPredictor is publicly available at https://broad.io/DILIPredictor for use via web interface and with all code available for download and local implementation via https://pypi.org/project/dilipred/.

Deep reinforcement learning enables better bias control in benchmark for virtual screening

Virtual screening (VS) has been incorporated into the paradigm of modern drug discovery. This field is now undergoing a new wave of revolution driven by artificial intelligence and more specifically, machine learning (ML). In terms of those out-of-the-box datasets for model training or benchmarking, their data volume and applicability domain are limited. They are suffering from the biases constantly reported in the ML application. To address these issues, we present a novel benchmark named MUBDsyn. The utilization of synthetic decoys (i.e., presumed inactives) is the main feature of MUBDsyn, where deep reinforcement learning was leveraged for bias control during decoy generation. Then, we carried out extensive validations on this new benchmark. First, we confirmed that MUBDsyn was superior to the classical benchmarks in control of domain bias, artificial enrichment bias and analogue bias. Moreover, we found that the assessment of ML models based on MUBDsyn was less biased as revealed by the analysis of asymmetric validation embedding bias. In addition, MUBDsyn showed better setting of benchmarking challenge for deep learning models compared with NRLiSt-BDB. Overall, we have proven that MUBDsyn is the close-to-ideal benchmark for VS. The computational tool is publicly available for the easy extension of MUBDsyn.

Predicting pharmacodynamic effects through early drug discovery with artificial intelligence-physiologically based pharmacokinetic (AI-PBPK) modelling

A mechanism-based pharmacokinetic/pharmacodynamic (PK/PD) model links the concentration-time profile of a drug with its therapeutic effects based on the underlying biological or physiological processes. Clinical endpoints play a pivotal role in drug development. Despite the substantial time and effort invested in screening drugs for favourable pharmacokinetic (PK) properties, they may not consistently yield optimal clinical outcomes. Furthermore, in the virtual compound screening phase, researchers cannot observe clinical outcomes in humans directly. These uncertainties prolong the process of drug development. As incorporation of Artificial Intelligence (AI) into the physiologically based pharmacokinetic/pharmacodynamic (PBPK) model can assist in forecasting pharmacodynamic (PD) effects within the human body, we introduce a methodology for utilizing the AI-PBPK platform to predict the PK and PD outcomes of target compounds in the early drug discovery stage. In this integrated platform, machine learning is used to predict the parameters for the model, and the mechanism-based PD model is used to predict the PD outcome through the PK results. This platform enables researchers to align the PK profile of a drug with desired PD effects at the early drug discovery stage. Case studies are presented to assess and compare five potassium-competitive acid blocker (P-CAB) compounds, after calibration and verification using vonoprazan and revaprazan.

AutoMolDesigner for Antibiotic Discovery: An AI-Based Open-Source Software for Automated Design of Small-Molecule Antibiotics

Discovery of small-molecule antibiotics with novel chemotypes serves as one of the essential strategies to address antibiotic resistance. Although a considerable number of computational tools committed to molecular design have been reported, there is a deficit in holistic and efficient tools specifically developed for small-molecule antibiotic discovery. To address this issue, we report AutoMolDesigner, a computational modeling software dedicated to small-molecule antibiotic design. It is a generalized framework comprising two functional modules, i.e., generative-deep-learning-enabled molecular generation and automated machine-learning-based antibacterial activity/property prediction, wherein individually trained models and curated datasets are out-of-the-box for whole-cell-based antibiotic screening and design. It is open-source, thus allowing for the incorporation of new features for flexible use. Unlike most software programs based on Linux and command lines, this application equipped with a Qt-based graphical user interface can be run on personal computers with multiple operating systems, making it much easier to use for experimental scientists. The software and related materials are freely available at GitHub (https://github.com/taoshen99/AutoMolDesigner) and Zenodo (https://zenodo.org/record/10097899).

QuantumScents: Quantum-Mechanical Properties for 3.5k Olfactory Molecules

Quantitative structure-odor relationships are critically important for studies related to the function of olfaction. Current literature data sets contain expert-labeled molecules but lack feature data. This paper introduces QuantumScents, a quantum mechanics augmented derivative of the Leffingwell data set. QuantumScents contains 3.5k structurally and chemically diverse molecules ranging from 2 to 30 heavy atoms (CNOS) and their corresponding 3D coordinates, total PBE0 energy, molecular dipole moment, and per-atom Hirshfeld charges, dipoles, and ratios. The authors demonstrate that Hirshfeld charges and ratios contain sufficient information to perform molecular classification by training a Message Passing Neural Network with chemprop (Heid, E.; et al. ChemRxiv, 2023, DOI: 10.26434/chemrxiv-2023-3zcfl) to predict scent labels. The QuantumScents data set is freely available on Zenodo along with the authors' code, example models, and data set generation workflow (https://zenodo.org/doi/10.5281/zenodo.8239853).

Energy-entropy multiscale cell correlation method to predict toluene-water log P in the SAMPL9 challenge

The energy-entropy multiscale cell correlation (EE-MCC) method is used to calculate toluene-water log P values of 16 drug molecules in the SAMPL9 physical properties challenge. EE-MCC calculates the free energy, energy and entropy from molecular dynamics (MD) simulations of the water and toluene solutions. Specifically, MCC evaluates entropy by partitioning the system into cells of correlated atoms at multiple length scales and further partitioning the local coordinates into energy wells, yielding vibrational and topographical terms from the energy-well sizes and probabilities. The log P values calculated by EE-MCC using three 200 ns MD simulations have a mean average error of 0.82 and standard error of the mean of 0.97 versus experiment, which is comparable with the best methods entered in SAMPL9. The main contribution to log P is from energy. Less polar drugs have more favourable energies of transfer. The entropy of transfer consists of increased solute vibrational and conformational terms in toluene due to weaker interactions, fewer solute positions in the larger-molecule solvent, reduced water vibrational entropy, negligible change in toluene vibrational entropy, and gains in solvent orientational entropy. The solvent entropy contributions here may be slightly underestimated because software limitations and statistical fluctuations meant that only the first shell could be included while averaged over the whole solution. Nonetheless, such issues will be addressed in future software to offer a general method to calculate entropy directly from MD simulation and to provide molecular understanding or guide system design.

Rapid and Automated Ab Initio Metabolite Collisional Cross Section Prediction from SMILES Input

We implemented an ab initio CCS prediction workflow which incrementally refines generated structures using molecular mechanics, a deep learning potential, conformational clustering, and quantum mechanics (QM). Automating intermediate steps for a high performance computing (HPC) environment allows users to input the SMILES structure of small organic molecules and obtain a Boltzmann averaged collisional cross section (CCS) value as output. The CCS of a molecular species is a metric measured by ion mobility spectrometry (IMS) which can improve annotation of untargeted metabolomics experiments. We report only a minor drop in accuracy when we expedite the CCS calculation by replacing the QM geometry refinement step with a single-point energy calculation. Even though the workflow involves stochastic steps (i.e., conformation generation and clustering), the final CCS value was highly reproducible for multiple iterations on L-carnosine. Finally, we illustrate that the gas phase ensembles modeled for the workflow are intermediate files which can be used for the prediction of other properties such as aqueous phase nuclear magnetic resonance chemical shift prediction. The software is available at the following link: https://github.com/DasSusanta/snakemake_ccs.

Toward α-1,3/4 fucosyltransferases targeted drug discovery: In silico uncovering of promising natural inhibitors of fucosyltransferase 6

Sialyl Lewis X (sLex ) antigen is a fucosylated cell-surface glycan that is normally involved in cell-cell interactions. The enhanced expression of sLex on cell surface glycans, which is attributed to the upregulation of fucosyltransferase 6 (FUT6), has been implicated in facilitating metastasis in human colorectal, lung, prostate, and oral cancers. The role that the upregulated FUT6 plays in the progression of tumor to malignancy, with reduced survival rates, makes it a potential target for anticancer drugs. Unfortunately, the lack of experimental structures for FUT6 has hampered the design and development of its inhibitors. In this study, we used in silico techniques to identify potential FUT6 inhibitors. We first modeled the three-dimensional structure of human FUT6 using AlphaFold. Then, we screened the natural compound libraries from the COCONUT database to sort out potential natural products (NPs) with best affinity toward the FUT6 model. As a result of these simulations, we identified three NPs for which we predicted binding affinities and interaction patterns quite similar to those we calculated for two experimentally tested FUT6 inhibitors, that is, fucose mimetic-1 and a GDP-triazole derived compound. We also performed molecular dynamics (MD) simulations for the FUT6 complexes with identified NPs, to investigate their stability. Analysis of the MD simulations showed that the identified NPs establish stable contacts with FUT6 under dynamics conditions. On these grounds, the three screened compounds appear as promising natural alternatives to experimentally tested FUT6 synthetic inhibitors, with expected comparable binding affinity. This envisages good prospects for future experimental validation toward FUT6 inhibition.

Identification of dietary compounds that interact with the circadian clock machinery: Molecular docking and structural similarity analysis

The molecular clock is vital for regulating circadian rhythms in various physiological processes, and its dysregulation is associated with multiple diseases. As such, the use of small molecule modulators to regulate the molecular clock presents a promising therapeutic approach. In this study, we generated a homology model of the human circadian locomotor output cycles kaput (CLOCK) protein to evaluate its ligand binding sites. Using molecular docking, we obtained further insights into the binding mode of the control compound CLK8 and explored a selection of dietary compounds. Our investigation of dietary compounds was guided by their potential interactions with the retinoic acid-related orphan receptors RORα/γ, which are involved in circadian regulation. Through the molecular similarity and docking analyses, we identified oleanolic acid demethyl, 3-epi-lupeol, and taraxasterol as potential ROR-interacting compounds. These compounds may exert therapeutic effects through their modulation of RORα/γ activity and subsequently influence the molecular clock. Overall, our study highlights the potential of small molecule modulators in regulating the molecular clock and the importance of exploring dietary compounds as a source of such modulators. Our findings also provide insights into the binding mechanisms of CLK8 and shed light on potential compounds that can interact with RORs to regulate the molecular clock. Future investigations could focus on validating the efficacy of these compounds in modulating the molecular clock and their potential use as therapeutic agents.

Design and Synthesis of New Acyl Urea Analogs as Potential σ1R Ligands

In search of synthetically accessible open-ring analogs of PD144418 or 5-(1-propyl-1,2,5,6-tetrahydropyridin-3-yl)-3-(p-tolyl)isoxazole, a highly potent sigma-1 receptor (σ1R) ligand, we herein report the design and synthesis of sixteen arylated acyl urea derivatives. Design aspects included modeling the target compounds for drug-likeness, docking at σ1R crystal structure 5HK1, and contrasting the lower energy molecular conformers with that of the receptor-embedded PD144418-a molecule we opined that our compounds could mimic pharmacologically. Synthesis of our acyl urea target compounds was achieved in two facile steps which involved first generating the N-(phenoxycarbonyl) benzamide intermediate and then coupling it with the appropriate amines weakly to strongly nucleophilic amines. Two potential leads (compounds 10 and 12, with respective in vitro σ1R binding affinities of 2.18 and 9.54 μM) emerged from this series. These leads will undergo further structure optimization with the ultimate goal of developing novel σ1R ligands for testing in neurodegeneration models of Alzheimer's disease (AD).

Automatic and accurate ligand structure determination guided by cryo-electron microscopy maps

Advances in cryo-electron microscopy (cryoEM) and deep-learning guided protein structure prediction have expedited structural studies of protein complexes. However, methods for accurately determining ligand conformations are lacking. In this manuscript, we develop EMERALD, a tool for automatically determining ligand structures guided by medium-resolution cryoEM density. We show this method is robust at predicting ligands along with surrounding side chains in maps as low as 4.5 Å local resolution. Combining this with a measure of placement confidence and running on all protein/ligand structures in the EMDB, we show that 57% of ligands replicate the deposited model, 16% confidently find alternate conformations, 22% have ambiguous density where multiple conformations might be present, and 5% are incorrectly placed. For five cases where our approach finds an alternate conformation with high confidence, high-resolution crystal structures validate our placement. EMERALD and the resulting analysis should prove critical in using cryoEM to solve protein-ligand complexes.

How Does the Study MD of pH-Dependent Exposure of Nanoparticles Affect Cellular Uptake of Anticancer Drugs?

The lack of knowledge about the uptake of NPs by biological cells poses a significant problem for drug delivery. For this reason, designing an appropriate model is the main challenge for modelers. To address this problem, molecular modeling studies that can describe the mechanism of cellular uptake of drug-loaded nanoparticles have been conducted in recent decades. In this context, we developed three different models for the amphipathic nature of drug-loaded nanoparticles (MTX-SS-γ-PGA), whose cellular uptake mechanism was predicted by molecular dynamics studies. Many factors affect nanoparticle uptake, including nanoparticle physicochemical properties, protein-particle interactions, and subsequent agglomeration, diffusion, and sedimentation. Therefore, the scientific community needs to understand how these factors can be controlled and the NP uptake of nanoparticles. Based on these considerations, in this study, we investigated for the first time the effects of the selected physicochemical properties of the anticancer drug methotrexate (MTX) grafted with hydrophilic-γ-polyglutamic acid (MTX-SS-γ-PGA) on its cellular uptake at different pH values. To answer this question, we developed three theoretical models describing drug-loaded nanoparticles (MTX-SS-γ-PGA) at three different pH values, such as (1) pH 7.0 (the so-called neutral pH model), (2) pH 6.4 (the so-called tumor pH model), and (3) pH 2.0 (the so-called stomach pH model). Exceptionally, the electron density profile shows that the tumor model interacts more strongly with the head groups of the lipid bilayer than the other models due to charge fluctuations. Hydrogen bonding and RDF analyses provide information about the solution of the NPs with water and their interaction with the lipid bilayer. Finally, dipole moment and HOMO-LUMO analysis showed the free energy of the solution in the water phase and chemical reactivity, which are particularly useful for determining the cellular uptake of the NPs. The proposed study provides fundamental insights into molecular dynamics (MD) that will allow researchers to determine the influence of pH, structure, charge, and energetics of NPs on the cellular uptake of anticancer drugs. We believe that our current study will be useful in developing a new model for drug delivery to cancer cells with a much more efficient and less time-consuming model.

pIChemiSt ─ Free Tool for the Calculation of Isoelectric Points of Modified Peptides

The isoelectric point (pI) is a fundamental physicochemical property of peptides and proteins. It is widely used to steer design away from low solubility and aggregation and guide peptide separation and purification. Experimental measurements of pI can be replaced by calculations knowing the ionizable groups of peptides and their corresponding pK_a values. Different pK_a sets are published in the literature for natural amino acids, however, they are insufficient to describe synthetically modified peptides, complex peptides of natural origin, and peptides conjugated with structures of other modalities. Noncanonical modifications (nCAAs) are ignored in the conventional sequence-based pI calculations, therefore producing large errors in their pI predictions. In this work, we describe a pI calculation method that uses the chemical structure as an input, automatically identifies ionizable groups of nCAAs and other fragments, and performs pK_a predictions for them. The method is validated on a curated set of experimental measures on 29 modified and 119093 natural peptides, providing an improvement of R² from 0.74 to 0.95 and 0.96 against the conventional sequence-based approach for modified peptides for the two studied pK_a prediction tools, ACDlabs and pKaMatcher, correspondingly. The method is available in the form of an open source Python library at https://github.com/AstraZeneca/peptide-tools, which can be integrated into other proprietary and free software packages. We anticipate that the pI calculation tool may facilitate optimization and purification activities across various application domains of peptides, including the development of biopharmaceuticals.

Preliminary Structure-Activity Relationship Study of the MMV Pathogen Box Compound MMV675968 (2,4-Diaminoquinazoline) Unveils Novel Inhibitors of Trypanosoma brucei brucei

New drugs are urgently needed for the treatment of human African trypanosomiasis (HAT). In line with our quest for novel inhibitors of trypanosomes, a small library of analogs of the antitrypanosomal hit (MMV675968) available at MMV as solid materials was screened for antitrypanosomal activity. In silico exploration of two potent antitrypanosomal structural analogs (7-MMV1578647 and 10-MMV1578445) as inhibitors of dihydrofolate reductase (DHFR) was achieved, together with elucidation of other antitrypanosomal modes of action. In addition, they were assessed in vitro for tentative inhibition of DHFR in a crude trypanosome extract. Their ADMET properties were also predicted using dedicated software. Overall, the two diaminoquinazoline analogs displayed approximately 40-fold and 60-fold more potency and selectivity in vitro than the parent hit, respectively (MMV1578445 (10): IC₅₀ = 0.045 µM, SI = 1737; MMV1578467 (7): IC₅₀ = 0.06 µM; SI = 412). Analogs 7 and 10 were also strong binders of the DHFR enzyme in silico, in all their accessible protonation states, and interacted with key DHFR ligand recognition residues Val32, Asp54, and Ile160. They also exhibited significant activity against trypanosome protein isolate. MMV1578445 (10) portrayed fast and irreversible trypanosome growth arrest between 4–72 h at IC₉₉. Analogs 7 and 10 induced in vitro ferric iron reduction and DNA fragmentation or apoptosis induction, respectively. The two potent analogs endowed with predicted suitable physicochemical and ADMET properties are good candidates for further deciphering their potential as starting points for new drug development for HAT.

Machine learning methods for pKa prediction of small molecules: Advances and challenges

The acid-base dissociation constant (pK_a) is a fundamental property influencing many ADMET properties of small molecules. However, rapid and accurate pK_a prediction remains a great challenge. In this review, we outline the current advances in machine-learning-based QSAR models for pK_a prediction, including descriptor-based and graph-based approaches, and summarize their pros and cons. Moreover, we highlight the current challenges and future directions regarding experimental data, crucial factors influencing pK_a and in silico prediction tools. We hope that this review can provide a practical guidance for the follow-up studies.

A single natural RNA modification can destabilize a U•A-T-rich RNA•DNA-DNA triple helix

Recent studies suggest noncoding RNAs interact with genomic DNA, forming RNA•DNA-DNA triple helices, as a mechanism to regulate transcription. One way cells could regulate the formation of these triple helices is through RNA modifications. With over 140 naturally occurring RNA modifications, we hypothesize that some modifications stabilize RNA•DNA-DNA triple helices while others destabilize them. Here, we focus on a pyrimidine-motif triple helix composed of canonical U•A-T and C•G-C base triples. We employed electrophoretic mobility shift assays and microscale thermophoresis to examine how 11 different RNA modifications at a single position in an RNA•DNA-DNA triple helix affect stability: 5-methylcytidine (m⁵C), 5-methyluridine (m⁵U or rT), 3-methyluridine (m³U), pseudouridine (Ψ), 4-thiouridine (s⁴U), N ⁶-methyladenosine (m⁶A), inosine (I), and each nucleobase with 2'-O-methylation (Nm). Compared to the unmodified U•A-T base triple, some modifications have no significant change in stability (Um•A-T), some have ∼2.5-fold decreases in stability (m⁵U•A-T, Ψ•A-T, and s⁴U•A-T), and some completely disrupt triple helix formation (m³U•A-T). To identify potential biological examples of RNA•DNA-DNA triple helices controlled by an RNA modification, we searched RMVar, a database for RNA modifications mapped at single-nucleotide resolution, for lncRNAs containing an RNA modification within a pyrimidine-rich sequence. Using electrophoretic mobility shift assays, the binding of DNA-DNA to a 22-mer segment of human lncRNA Al157886.1 was destabilized by ∼1.7-fold with the substitution of m⁵C at known m⁵C sites. Therefore, the formation and stability of cellular RNA•DNA-DNA triple helices could be influenced by RNA modifications.

Informing geometric deep learning with electronic interactions to accelerate quantum chemistry

Predicting electronic energies, densities, and related chemical properties can facilitate the discovery of novel catalysts, medicines, and battery materials. However, existing machine learning techniques are challenged by the scarcity of training data when exploring unknown chemical spaces. We overcome this barrier by systematically incorporating knowledge of molecular electronic structure into deep learning. By developing a physics-inspired equivariant neural network, we introduce a method to learn molecular representations based on the electronic interactions among atomic orbitals. Our method, OrbNet-Equi, leverages efficient tight-binding simulations and learned mappings to recover high-fidelity physical quantities. OrbNet-Equi accurately models a wide spectrum of target properties while being several orders of magnitude faster than density functional theory. Despite only using training samples collected from readily available small-molecule libraries, OrbNet-Equi outperforms traditional semiempirical and machine learning-based methods on comprehensive downstream benchmarks that encompass diverse main-group chemical processes. Our method also describes interactions in challenging charge-transfer complexes and open-shell systems. We anticipate that the strategy presented here will help to expand opportunities for studies in chemistry and materials science, where the acquisition of experimental or reference training data is costly.

Galaxy workflows for fragment-based virtual screening: a case study on the SARS-CoV-2 main protease

We present several workflows for protein-ligand docking and free energy calculation for use in the workflow management system Galaxy. The workflows are composed of several widely used open-source tools, including rDock and GROMACS, and can be executed on public infrastructure using either Galaxy’s graphical interface or the command line. We demonstrate the utility of the workflows by running a high-throughput virtual screening of around 50000 compounds against the SARS-CoV-2 main protease, a system which has been the subject of intense study in the last year.

OrbNet Denali: A machine learning potential for biological and organic chemistry with semi-empirical cost and DFT accuracy

We present OrbNet Denali, a machine learning model for an electronic structure that is designed as a drop-in replacement for ground-state density functional theory (DFT) energy calculations. The model is a message-passing graph neural network that uses symmetry-adapted atomic orbital features from a low-cost quantum calculation to predict the energy of a molecule. OrbNet Denali is trained on a vast dataset of 2.3 × 10⁶ DFT calculations on molecules and geometries. This dataset covers the most common elements in biochemistry and organic chemistry (H, Li, B, C, N, O, F, Na, Mg, Si, P, S, Cl, K, Ca, Br, and I) and charged molecules. OrbNet Denali is demonstrated on several well-established benchmark datasets, and we find that it provides accuracy that is on par with modern DFT methods while offering a speedup of up to three orders of magnitude. For the GMTKN55 benchmark set, OrbNet Denali achieves WTMAD-1 and WTMAD-2 scores of 7.19 and 9.84, on par with modern DFT functionals. For several GMTKN55 subsets, which contain chemical problems that are not present in the training set, OrbNet Denali produces a mean absolute error comparable to those of DFT methods. For the Hutchison conformer benchmark set, OrbNet Denali has a median correlation coefficient of R² = 0.90 compared to the reference DLPNO-CCSD(T) calculation and R² = 0.97 compared to the method used to generate the training data (ωB97X-D3/def2-TZVP), exceeding the performance of any other method with a similar cost. Similarly, the model reaches chemical accuracy for non-covalent interactions in the S66x10 dataset. For torsional profiles, OrbNet Denali reproduces the torsion profiles of ωB97X-D3/def2-TZVP with an average mean absolute error of 0.12 kcal/mol for the potential energy surfaces of the diverse fragments in the TorsionNet500 dataset.

Effective Use of Empirical Data for Virtual Screening against APJR GPCR Receptor

Alzheimer's disease is a neurodegenerative disorder incompatible with normal daily activity, affecting one in nine people. One of its potential targets is the apelin receptor (APJR), a G-protein coupled receptor, which presents considerably high expression levels in the central nervous system. In silico studies of APJR drug-like molecule binding are in small numbers while high throughput screenings (HTS) are already sufficiently many to devise efficient drug design strategies. This presents itself as an opportunity to optimize different steps in future large scale virtual screening endeavours. Here, we ran a first stage docking simulation against a library of 95 known binders and 3829 generated decoys in an effort to improve the rescoring stage. We then analyzed receptor binding site structure and ligands binding poses to describe their interactions. As a result, we devised a simple and straightforward virtual screening Stage II filtering score based on search space extension followed by a geometric estimation of the ligand-binding site fitness. Having this score, we used an ensemble of receptors generated by Hamiltonian Monte Carlo simulation and reported the results. The improvements shown herein prove that our ensemble docking protocol is suited for APJR and can be easily extrapolated to other GPCRs.

MolGpka: A Web Server for Small Molecule pKa Prediction Using a Graph-Convolutional Neural Network

pK_a is an important property in the lead optimization process since the charge state of a molecule in physiologic pH plays a critical role in its biological activity, solubility, membrane permeability, metabolism, and toxicity. Accurate and fast estimation of small molecule pK_a is vital during the drug discovery process. We present MolGpKa, a web server for pK_a prediction using a graph-convolutional neural network model. The model works by learning pK_a related chemical patterns automatically and building reliable predictors with learned features. ACD/pK_a data for 1.6 million compounds from the ChEMBL database was used for model training. We found that the performance of the model is better than machine learning models built with human-engineered fingerprints. Detailed analysis shows that the substitution effect on pK_a is well learned by the model. MolGpKa is a handy tool for the rapid estimation of pK_a during the ligand design process. The MolGpKa server is freely available to researchers and can be accessed at https://xundrug.cn/molgpka.

AutoGrow4: an open-source genetic algorithm for de novo drug design and lead optimization

We here present AutoGrow4, an open-source program for semi-automated computer-aided drug discovery. AutoGrow4 uses a genetic algorithm to evolve predicted ligands on demand and so is not limited to a virtual library of pre-enumerated compounds. It is a useful tool for generating entirely novel drug-like molecules and for optimizing preexisting ligands. By leveraging recent computational and cheminformatics advancements, AutoGrow4 is faster, more stable, and more modular than previous versions. It implements new docking-program compatibility, chemical filters, multithreading options, and selection methods to support a wide range of user needs. To illustrate both de novo design and lead optimization, we here apply AutoGrow4 to the catalytic domain of poly(ADP-ribose) polymerase 1 (PARP-1), a well characterized DNA-damage-recognition protein. AutoGrow4 produces drug-like compounds with better predicted binding affinities than FDA-approved PARP-1 inhibitors (positive controls). The predicted binding modes of the AutoGrow4 compounds mimic those of the known inhibitors, even when AutoGrow4 is seeded with random small molecules. AutoGrow4 is available under the terms of the Apache License, Version 2.0. A copy can be downloaded free of charge from http://durrantlab.com/autogrow4.

Gypsum-DL: an open-source program for preparing small-molecule libraries for structure-based virtual screening

Computational techniques such as structure-based virtual screening require carefully prepared 3D models of potential small-molecule ligands. Though powerful, existing commercial programs for virtual-library preparation have restrictive and/or expensive licenses. Freely available alternatives, though often effective, do not fully account for all possible ionization, tautomeric, and ring-conformational variants. We here present Gypsum-DL, a free, robust open-source program that addresses these challenges. As input, Gypsum-DL accepts virtual compound libraries in SMILES or flat SDF formats. For each molecule in the virtual library, it enumerates appropriate ionization, tautomeric, chiral, cis/trans isomeric, and ring-conformational forms. As output, Gypsum-DL produces an SDF file containing each molecular form, with 3D coordinates assigned. To demonstrate its utility, we processed 1558 molecules taken from the NCI Diversity Set VI and 56,608 molecules taken from a Distributed Drug Discovery (D3) combinatorial virtual library. We also used 4463 high-quality protein–ligand complexes from the PDBBind database to show that Gypsum-DL processing can improve virtual-screening pose prediction. Gypsum-DL is available free of charge under the terms of the Apache License, Version 2.0.