Directory of useful decoys, enhanced (DUD-E): better ligands and decoys for better benchmarking

Harnessing computational tools for drug discovery: An integrated computational approach to identify potential BACE-1 inhibitors

The hallmark of Alzheimer's disease (AD), a progressive neurodegenerative condition, is the buildup of amyloid-beta (Aβ) plaque, which is mainly caused by β-secretase 1 (BACE-1) activity. BACE-1 inhibition is a potentially effective treatment strategy to lower the progression of AD. In order to find possible BACE-1 inhibitors using a drug repurposing technique, this study uses an integrated computational approach that includes pharmacophore modelling, virtual screening, molecular docking, MM-GBSA, molecular dynamics (MD) simulations, in-silico ADMET profiling, and PBPK modelling. A pharmacophore model, was created with known BACE-1 inhibitors to enable virtual screening of both novel and FDA-approved chemical libraries. Top candidates with good free energy scores and strong binding affinities were found using molecular docking and MM-GBSA calculations. The stability of shortlisted Hits inside the BACE-1 active site was further validated using MD simulations, which showed that some of the important interactions were maintained across a period of 50ns. ADMET and PBPK studies predicted favorable pharmacokinetic and safety profiles for the shortlisted hits, particularly for B2 and B9. These findings identify potential candidates for future experimental validation, offering an inexpensive approach for identification of compounds as potential BACE-1 inhibitors.

Discovery of natural MCL1 inhibitors using pharmacophore modelling, QSAR, docking, ADMET, molecular dynamics, and DFT analysis

Mcl-1, a member of the Bcl-2 family, is a crucial regulator of apoptosis, frequently overexpressed in various cancers, including lung, breast, pancreatic, cervical, ovarian cancers, leukemia, and lymphoma. Its anti-apoptotic function allows tumor cells to evade cell death and contributes to drug resistance, making it an essential target for anticancer drug development. This study aimed to discover potent antileukemic compounds targeting Mcl-1. We selected diverse molecules from the BindingDB database to construct a structure-based pharmacophore model, which facilitated the virtual screening of 407,270 compounds from the COCONUT database. An e-pharmacophore model was developed using the co-crystallized inhibitor, followed by QSAR modeling to estimate IC50 values and filter compounds with predicted values below the median. The top hits underwent molecular docking and MMGBSA binding energy calculations against Mcl-1, resulting in the selection of two promising candidates for further ADMET analysis. DFT calculations assessed their electronic properties, confirming favorable reactivity profiles of the screened compounds. Predictions for physicochemical and ADMET properties aligned with expected bioactivity and safety. Molecular dynamics simulations further validated their strong binding affinity and stability, positioning them as potential Mcl-1 inhibitors. Our comprehensive computational approach highlights these compounds as promising antileukemic agents, with future in vivo and in vitro validation recommended for further confirmation.

Exploring 4th generation EGFR inhibitors: A review of clinical outcomes and structural binding insights

Epidermal growth factor receptor (EGFR) is a potential target for anticancer therapies and plays a crucial role in cell growth, survival, and metastasis. EGFR gene mutations trigger aberrant signaling, leading to non-small cell lung cancer (NSCLC). Tyrosine kinase inhibitors (TKIs) effectively target these mutations to treat NSCLC. While the first three generations of EGFR TKIs have been proven effective, the emergence of the EGFR-C797S resistance mutation poses a new challenge. To address this, various synthetic EGFR TKIs have been developed. In this review, we have summarized the EGFR TKIs reported in the past five years, focusing on their clinical outcomes and structure-activity relationship analysis. We have also explored binding modes and interactions between the binding pocket and ligands to provide insights into the mechanisms of these inhibitors, which contribute to advancements in targeted cancer therapy. Additionally, artificial Intelligence-driven methods, including recursive neural networks and reinforcement learning, have revolutionized EGFR inhibitor design by facilitating rapid screening, predicting EGFR mutations, and novel compound generation.

In silico screening to search for selective sodium channel blockers: When size matters

Dravet Syndrome is a severe childhood drug-resistant epilepsy. The predominant etiology of this condition is related to de novo mutations within the SCN1A gene, which codes for the alpha subunit of the NaV1.1 sodium channels. This dysfunction leads to hypoexcitability of GABAergic interneurons. In turn, the loss of electrical excitability in GABAergic interneurons leads to an imbalance of excitation over inhibition in many neural circuits. Notably, exacerbation of symptoms is observed when non-selective sodium channel blockers are administered to patients with Dravet. Recent studies in animal models of Dravet have highlighted the potential of highly specific sodium channel blockers capable of blocking other sodium channel subtypes without inhibiting NaV1.1 current and selective activators of NaV1.1 as viable therapeutic strategies for alleviating Dravet Syndrome symptoms. Here, we describe the development and validation of ligand-based machine learning models to identify ligands with inhibitory effects on sodium channel isoforms NaV1.1 and NaV1.2. These models were built based on in-house open-source routines and Mordred molecular descriptors. First, linear classifiers were inferred using a combination of feature-bagging and Forward Stepwise selection. Secondly, ensemble learning was applied to build meta-classifiers with improved predictive ability, whose performance was tested in retrospective screening experiments. After in silico validation, the models were applied to screen for drug repurposing opportunities in the DrugBank and Drug Repurposing Hub databases, to identify selective blocking agents of NaV1.2 devoid of NaV1.1 blocking activity, as potential compounds for the treatment of Dravet Syndrome. Forty in silico hits were later identified in a prospective screening experiment. Four of them were acquired and submitted to experimental confirmation via patch clamp: three of these candidates, Eltrombopag, Sufugolix, and Glesatinib, showed blocking effects on NaV1.2 currents, although no subtype selectivity was observed. The different predictive abilities of the NaV1.1 and NaV1.2 models may be attributed to the different sizes of the datasets used to train and validate the respective models.

Computational discovery of novel PI3KC2α inhibitors using structure-based pharmacophore modeling, machine learning and molecular dynamic simulation

PI3KC2α is a lipid kinase associated with cancer metastasis and thrombosis. In this study, we present a novel computational workflow integrating structure-based pharmacophore modeling, machine learning (ML), and molecular dynamics (MD) simulations to discover new PI3KC2α inhibitors. Key innovations include the generation of diverse pharmacophores from both crystallographic and docking-derived complexes, coupled with data augmentation via ligand conformational sampling to enhance ML robustness. The optimal model, developed using XGBoost with genetic function algorithm (GFA) and Shapley additive explanations (SHAP), identified four critical pharmacophores and three descriptors governing bioactivity. Virtual screening of the NCI database using these pharmacophores yielded three hits, with H_1 (NCI: 725847) demonstrating MD-derived binding stability and affinity comparable to the potent inhibitor PITCOIN1 (IC50 = 95 nM). This study represents the first application of a conformation-augmented ML framework to PI3KC2α inhibition, offering a blueprint for targeting underexplored kinases with limited structural data.

Protein Sequence Analysis landscape: A Systematic Review of Task Types, Databases, Datasets, Word Embeddings Methods, and Language Models

Protein sequence analysis examines the order of amino acids within protein sequences to unlock diverse types of a wealth of knowledge about biological processes and genetic disorders. It helps in forecasting disease susceptibility by finding unique protein signatures, or biomarkers that are linked to particular disease states. Protein Sequence analysis through wet-lab experiments is expensive, time-consuming and error prone. To facilitate large-scale proteomics sequence analysis, the biological community is striving for utilizing AI competence for transitioning from wet-lab to computer aided applications. However, Proteomics and AI are two distinct fields and development of AI-driven protein sequence analysis applications requires knowledge of both domains. To bridge the gap between both fields, various review articles have been written. However, these articles focus revolves around few individual tasks or specific applications rather than providing a comprehensive overview about wide tasks and applications. Following the need of a comprehensive literature that presents a holistic view of wide array of tasks and applications, contributions of this manuscript are manifold: It bridges the gap between Proteomics and AI fields by presenting a comprehensive array of AI-driven applications for 63 distinct protein sequence analysis tasks. It equips AI researchers by facilitating biological foundations of 63 protein sequence analysis tasks. It enhances development of AI-driven protein sequence analysis applications by providing comprehensive details of 68 protein databases. It presents a rich data landscape, encompassing 627 benchmark datasets of 63 diverse protein sequence analysis tasks. It highlights the utilization of 25 unique word embedding methods and 13 language models in AI-driven protein sequence analysis applications. It accelerates the development of AI-driven applications by facilitating current state-of-the-art performances across 63 protein sequence analysis tasks.

Harnessing virtual screening and MD simulations: a multistage approach to identifying potent and nontoxic agonists for protein kinase A

Obesity-induced insulin resistance impairs glucose tolerance and β-cell function, significantly contributing to the pathogenesis of type 2 diabetes (T2D). Protein kinase A (PKA), being one of the key effector molecules of the cyclic AMP (cAMP) pathway, increases insulin secretion via membrane activity, gene expression, and exocytosis of insulin granules. The previous studies were limited to either target cAMP analogs as PKA agonist or mostly flavonoids using In vivo and In vitro studies (Hameed in Int J Biol Macromol 119:149-156, 2018;Shahab in Biomed Pharmacother 177, 2024;Hameed in Eur J Pharmacol 820:245-255, 2018;Hameed in Eur J Pharmacol 858, 2019;Hafizur in Med Chem Res 27:1408-1418, 2018;). To speed up the process, this study aimed to identify potential PKA activators as therapeutic agents for restoring β-cell function in Type 2 Diabetes (T2D) using a multistage virtual screening approach. In the initial phase, a ligand-based pharmacophore model was constructed to screen an in-house small molecule database for potential PKA agonists. By targeting the essential pharmacophoric features necessary for interaction with the cyclic nucleotide-binding (CNB) domain of PKA, the goal was to identify compounds with strong binding affinities and therapeutic promise. To gain deeper insights into the molecular mechanisms of PKA activation and evaluate key interactions and dynamic stability, a subset of promising hits was subjected to all-atom molecular dynamics simulations. Simulations showed significant conformational changes in PKA complexes, with average backbone root mean square deviations (RMSD) of 0.37 ± 0.15 nm for Comp-03, 0.53 ± 0.18 nm for Comp-11, 0.31 ± 0.06 nm for Comp-17, 0.28 ± 0.03 nm for Comp-38, and 0.48 ± 0.13 nm for Comp-41. The N3A motif showed consistent fluctuations, suggesting increased flexibility. Binding free energy calculations showed binding free energies (ΔGbind) for cAMP, Comp-03, Comp-17, Comp-38, and Comp-41, with ΔGbind values of - 62.87 ± 10.04, - 68.57 ± 12.77, - 78.13 ± 16.36, - 62.67 ± 13.06, and - 80.87 ± 10.45 kcal/mol, respectively. To further probe the conformational stability of these complexes, multidimensional scaling and free energy profiling were carried out. This exhaustive research study, involving examination of stability dynamics, deviation patterns, interaction networks, conformational changes, and energy profiles, provides profound understanding about mechanisms that activate PKA. The findings highlight several promising lead compounds, notably Comp-03, Comp-17, Comp-38, and Comp-41, which exhibit superior potential to activate PKA compared to cAMP. These findings lay a strong foundation for the development of novel PKA activators as potential therapeutic agents for managing T2D.

Discovery of Trypanosoma cruzi Carbonic Anhydrase Inhibitors by a Combination of Ligand- and Structure-Based Virtual Screening

Trypanosoma cruzi carbonic anhydrase (TcCA) has emerged as a promising therapeutic target for the treatment of Chagas disease. In this study, a sequential virtual screening strategy was employed to identify potential TcCA inhibitors. The workflow consisted of ligand-based virtual screening applied to diverse chemical libraries, followed by target-based molecular docking to refine the selection of compounds. Six candidates were selected for their in vitro evaluation against both the enzyme and the parasite. All of them confirmed inhibitory activity against TcCA, with three exhibiting Kis in the nanomolar or submicromolar range. Among these, Nitrofurantoin demonstrated significant inhibitory activity, with a Ki of 93 nM against TcCA and EC50 of 14.82 μM against T. cruzi trypomastigotes. These findings suggest that Nitrofurantoin is a promising lead compound for further optimization and highlight the therapeutic potential of TcCA as a drug target in Chagas disease.

On the Difficulty to Rescore Hits from Ultralarge Docking Screens

Docking-based virtual screening tools customized to mine ultralarge chemical spaces are consistently reported to yield both higher hit rates and more potent ligands than that achieved by conventional docking of smaller million-sized compound libraries. This remarkable achievement is however counterbalanced by the absolute necessity to design an efficient postprocessing of the millions of potential virtual hits for selecting a few chemically diverse compounds for synthesis and biological evaluation. We here retrospectively analyzed ten successful ultralarge virtual screening hit lists that underwent in vitro binding assays, for binding affinity prediction using eight rescoring methods including simple empirical scoring functions, machine learning, molecular-mechanics and quantum-mechanics approaches. Although the best predictions usually rely on the most sophisticated methods, none of the tested rescoring methods could robustly distinguish known binders from inactive compounds, across all assays. Energy refinement of protein-ligand complexes, prior to rescoring, marginally helped molecular mechanics and quantum mechanics approaches but deteriorates predictions from empirical and machine learning scoring functions.

Real Way to Target Gram-Negative Pathogens: Discovery of a Novel Helicobacter pylori Antibiotic Class

In an era of escalating antibiotic resistance, there is a pressing need for innovative strategies to develop novel antibiotics. Gram-negative bacteria, characterized by their robust dual-membrane, are intrinsically resistant to a wide range of antibiotics and can readily develop new resistances. Members of this bacterial class comprise numerous pathogenic organisms, including the primary cause of gastric cancer, Helicobacter pylori. In this study, we used the Giga-sized collection of theoretical molecules inside Enamine's REAL Space to identify inhibitors for H. pylori glutamate racemase. These compounds displayed a diverse range of activity in preventing H. pylori growth, with our most potent hits capable of selective full growth inhibition for metronidazole and clarithromycin resistant H. pylori strains. Alongside the introduction of a novel antibiotic class for this carcinogenic pathogen, our unique implementation of REAL Space holds great promise for Gram-negative antibiotic development as a whole.

Pharmacophore Modeling of Janus Kinase Inhibitors: Tools for Drug Discovery and Exposition Prediction

Pesticides are essential in agriculture for protecting crops and boosting productivity, but their widespread use may pose significant health risks. Farmworkers face direct exposure through skin contact and inhalation, which may lead to hormonal imbalances, neurological disorders, and elevated cancer risks. Moreover, pesticide residues in food and water may affect surrounding communities. One of the lesser investigated issues is immunotoxicity, mostly because the chronic effects of compound exposure are very complex to study. As a case study, this work utilized pharmacophore modeling and virtual screening to identify pesticides that may inhibit Janus kinases (JAK1, JAK2, JAK3) and tyrosine kinase 2 (TYK2), which are pivotal in immune response regulation, and are associated with cancer development and increased infection susceptibility. We identified 64 potential pesticide candidates, 22 of which have previously been detected in the human body, as confirmed by the Human Metabolome Database. These results underscore the critical need for further research into potential immunotoxic and chronic impacts of the respective pesticides on human health.

Allosteric modulation of Plasmodium falciparum Isoleucyl tRNA synthetase by South African natural compounds

Targeting Plasmodium falciparum (Pf) aminoacyl tRNA synthetases is a viable strategy to overcome malaria parasite multi-drug resistance. Here, we focused on Pf Isoleucyl tRNA synthetase (PfIleRS) to identify potential allosteric inhibitors from 1019 South African Natural Compounds (SANC). Eleven potential hits, which passed ADMET and PAINS, were selected based on their docking binding affinity which was higher for PfIleRS than for human IleRS. Molecular dynamics simulations revealed that the compounds, particularly SANC456, commonly induced considerable changes in the global conformation and dynamics of PfIleRS, suggesting potential allosteric modulatory effects. Importantly, all 11 SANC hits reduced the binding affinity of the nucleotide AMP molecule by at least 25%. Some SANC ligand-bound systems (SANC456, SANC1095, and SANC1104) significantly increased the distance between the AMP and Ile ligands. Possible explanations for these changes were explored using three dynamic residue network centrality metrics. Betweenness centrality identified a major allosteric pathway in holo PfIleRS spanning the entire protein length. In contrast, SANC382, SANC456, SANC522, SANC806 and SANC1095 ligand-bound systems exhibited delta BC pathways (SANC-protein minus holo-protein), induced by the ligands, extending from their respective pockets into the active site. Additionally, eigenvector centrality revealed two important residue clusters either side of the holo active site which became altered in the ligand-bound systems, indicating possible allosteric activity. Lastly, many SANC systems showed decreased closeness centrality of zinc finger and active site residues, including the HYGH and KMSKR motifs. We believe that the compounds identified in this study as potential allosteric inhibitors have strong translational potential and warrant further investigation through in vitro and in vivo experiments. Overall, they hold promise as starting points for the development of new and effective antimalarial therapies, particularly against multidrug-resistant Plasmodium parasites.

Generating diversity and securing completeness in algorithmic retrosynthesis

Chemical synthesis planning has considerably benefited from advances in the field of machine learning. Neural networks can reliably and accurately predict reactions leading to a given, possibly complex, molecule. In this work we focus on algorithms for assembling such predictions to a full synthesis plan that, starting from simple building blocks, produces a given target molecule, a procedure known as retrosynthesis. Objective functions for this task are hard to define and context-specific. In order to generate a diverse set of synthesis plans for chemists to select from, we capture the concept of diversity in a novel chemical diversity score (CDS). Our experiments show that our algorithm outperforms the algorithm predominantly employed in this domain, Monte-Carlo Tree Search, with respect to diversity in terms of our score as well as time efficiency. SCIENTIFIC CONTRIBUTION: We adapt Depth-First Proof-Number Search (DFPN) (Please refer to https://github.com/Bayer-Group/bayer-retrosynthesis-search for the accompanying source code.) and its variants, which have been applied to retrosynthesis before, to produce a set of solutions, with an explicit focus on diversity. We also make progress on understanding DFPN in terms of completeness, i.e., the ability to find a solution whenever there exists one. DFPN is known to be incomplete, for which we provide a much cleaner example, but we also show that it is complete when reinforced with a threshold-controlling routine from the literature.

Improving the Hit Rates of Virtual Screening by Active Learning from Bioactivity Feedback

Virtual screening has been widely used to identify potential hit candidates that can bind to the target protein in drug discovery. Contemporary screening methods typically rely on oversimplified scoring functions, frequently yielding one-digit hit rates (or even zero) among top-ranked candidates. The substantial cost of laboratory validation further constrains the exploration of candidate molecules. We find that test-time prediction refinement is almost blank in this area, which means bioactivity feedback in the wet-lab experiments is neglected. Here, we introduce an Active Learning from Bioactivity Feedback (ALBF) framework to enhance the weak hit rate of current virtual screening methods. ALBF spends the budget of wet-lab experiments iteratively and leverages the target-specific bioactivity insights from current wet-lab tests to refine the score results (i.e., rankings). Our framework consists of two components: a novel query strategy that considers the evaluation quality and its overall influence on other top-scored molecules; and an efficient score optimization strategy that propagates the bioactivity feedback to structurally similar molecules. We evaluated ALBF on diverse subsets of the well-known DUD-E and LIT-PCBA benchmarks. Our active learning protocol averagely enhances top-100 hit rates by 60% and 30% on DUD-E and LIT-PCBA with 50 to 200 bioactivity queries on the selected molecules that are deployed in ten rounds. The consistently superior performance demonstrates ALBF's potential to enhance both the accuracy and cost-effectiveness of active learning-based laboratory testing.

Enhancing PI3Kγ inhibitor discovery: a machine learning-based virtual screening approach integrating pharmacophores, docking, and molecular descriptors

PI3Kγ is a lipid kinase that is expressed primarily in leukocytes and plays a significant role in tumors, inflammation, and autoimmune diseases. Consequently, considerable attention has been given to the development of pharmacological inhibitors of PI3Kγ. Recently, machine learning-based virtual screening approaches have been increasingly applied in new drug discovery research, potentially providing innovative strategies for the development of PI3Kγ inhibitors. Thus, in this study, we developed a naïve Bayesian classification (NBC) model that integrates molecular descriptors, molecular fingerprints, molecular docking, and pharmacophore models for virtual screening of the PI3Kγ protein. The validation results indicated that the optimal model demonstrated significant potential for differentiating between active and inactive compounds, as well as a reliable ability to identify true PI3Kγ inhibitors with defined biological activity. Additionally, the optimal NBC model provided favorable and unfavorable fragments for PI3Kγ inhibitors, which will help guide the design and discovery of novel PI3Kγ inhibitors. Finally, the optimal NBC model was employed to perform virtual screening on the ChEMBL database, resulting in the identification of several compounds with high potential as PI3Kγ inhibitors. We anticipate that the developed machine learning-based virtual screening approach will offer valuable insights and guidance for the development of novel PI3Kγ inhibitors.

Discovery of Potential Scaffolds for Methionine Adenosyltransferase 2A (MAT2A) Inhibitors: Virtual Screening, Synthesis, and Biological Evaluation

The inhibition of methionine adenosyltransferase 2A (MAT2A) in cancers harboring deletions of the methylthioadenosine phosphorylase (MTAP) gene induces synthetic lethality, making it a highly compelling strategy in the pursuit of precision anticancer therapeutics. In this study, structure-based computing methods were employed to discover novel scaffolds as potential MAT2A inhibitors. The most potent compound, 17, demonstrated inhibition of MAT2A with an IC50 of 0.43 μM, and showed antitumor effects against MTAP-/- HCT116 cells with an IC50 of 1.4 μM. The identified compounds and their associated structural data could provide valuable insights for related drug discovery projects.

A Database for Large-Scale Docking and Experimental Results

The rapid expansion of readily accessible compounds over the past six years has transformed molecular docking, improving hit rates and affinities. While many millions of molecules may score well in a docking campaign, the results are rarely fully shared, hindering the benchmarking of machine learning and chemical space exploration methods that seek to explore the expanding chemical spaces. To address this gap, we develop a website providing access to recent large library campaigns, including poses, scores, and in vitro results for campaigns against 11 targets, with 6.3 billion molecules docked and 3729 compounds experimentally tested. In a simple proof-of-concept study that speaks to the new library's utility, we use the new database to train machine learning models to predict docking scores and to find the top 0.01% scoring molecules while evaluating only 1% of the library. Even in these proof-of-concept studies, some interesting trends emerge: unsurprisingly, as models train on larger sets, they perform better; less expectedly, models could achieve high correlations with docking scores and yet still fail to enrich the new docking-discovered ligands, or even the top 0.01% of docking-ranked molecules. It will be interesting to see how these trends develop for methods more sophisticated than the simple proof-of-concept studies undertaken here; the database is openly available at lsd.docking.org.

E3Docker: a docking server for potential E3 binder discovery

Targeted protein degradation (TPD) has emerged as a promising therapeutic strategy for modulating protein levels in cells. Proteolysis-targeting chimeras and molecular glues facilitate the formation of a complex between the protein of interest (POI) and a specific E3 ligase, leading to POI ubiquitination and subsequent degradation by the proteasome. Considering over 600 E3s in the human genome, it is of great potential to find novel E3 binders and recruit new E3 ligase for TPD related drug discovery. Here we introduce E3Docker, an online computational tool for E3 binder discovery. A total of 1075 Homo sapiens E3 ligases are collected from databases and literature, and 4474 three-dimensional structures of these E3 ligases, in either apo or complex forms, are integrated into the web server. The druggable pockets for each E3 ligase are defined by experimentally bound ligand from PDB or predicted by using DeepPocket. CoDock-Ligand is employed as docking engine for potential E3 binder estimation. With a user-friendly interface, E3Docker facilitates the generation of binding poses and affinity scores for compounds with over 1000 kinds of E3 ligases and may benefit for novel E3 binder discovery. The E3Docker server and tutorials are freely available at https://e3docker.schanglab.org.cn/.

An Effective Reaction-Based Virtual Screening Method to Discover New CDK8 Ligands

Cyclin Dependent Kinase 8 (CDK8) is a valuable drug target for cancer suppression. Through an effective reaction-based virtual screening method consisting of pharmacophore modeling, Scifinder database searches, and energy evaluations, a number of new type II CDK8 ligands were discovered with comparable or better binding free energies than the ones reported in literature. In this method, a pharmacophore model, derived from seven crystal structures of CDK8 and type II ligands, was able to catch the key interactions for ligands binding at the ATP binding site of CDK8. This model then was used to screen the results from Scifinder database searches that apply chemical reaction rules in the search cycles, and the output compounds were evaluated and ranked first by a fast energy estimation method and then by the VM2 free energy calculation method. Among the top discovered ligands, three have lower Kd values against CDK8 than a potent reference ligand, and compound F (3-[3-tert-butyl-1-(4-methylphenyl)-1H-pyrazol-5-yl]-1-(8-hydroxynaphthalen-1-yl)urea) is the most potent one with an Kd value of 7.5 nM. Compound F has a relatively small molecular structure, receives both strong van der Waals energy and optimized overall electrostatic energy when binding with CDK8, and deserves to be a promising lead compound for further development. This effective virtual screening method and the novel compounds found in this work have implications for CDK8 drug discovery.

Discovery of ASGR1 and HMGCR dual-target inhibitors based on supervised learning, molecular docking, molecular dynamics simulations, and biological evaluation

3-Hydroxy-3-Methylglutaryl-CoA Reductase (HMGCR) and Asialoglycoprotein Receptor 1 (ASGR1) are potential therapeutic targets for atherosclerotic cardiovascular disease (ASCVD). In this study, we employed an innovative approach that combined ligand-based supervised learning, molecular docking, molecular dynamics simulations, and various in-silico techniques. The objective was to effectively screen the Chemdiv and SPECS molecule databases to discover potential inhibitors targeting both HMGCR and ASGR1, resulting in a dual inhibition effect. Compound 8006-6092, K007-0721, and D011-1471 exhibited inhibition rates of 41.48 %, 61.48 %, and 49.63 %, respectively, at a concentration of 10 μM against HMGCR. In addition, they demonstrated significant binding to ASGR1, with dissociation constants (Kd) of 461.33 μM, 67.63 μM, and 695.50 μM, respectively. These findings suggest that these dual inhibitors, 8006-6092, K007-0721, and D011-1471, present promising outcomes, potentially warranting further optimization as lead compounds for the treatment of ASCVD.

Discovery of therapeutic promising natural products to target Kv1.3 channel, a transmembrane protein regulating immune disorders, through multidimensional virtual screening, molecular dynamics simulations and biological validation

Kv1.3 voltage-gated potassium channel, is a transmembrane protein that facilitates K+ movement through cell membranes via its intrinsic pores, regulating the cell signaling cascades, especially in immune disorders. In this paper, we employed multidimensional virtual screening to identify 24 potential Kv1.3 inhibitors from a library of 27,637 compounds, with electrophysiological assays confirming 8 active inhibitors (33.33 % hit rate). Structure-activity relationship (SAR) analysis demonstrated that 4-methylpentyl group in side chain and furan ring in Furanocoumarins skeleton are crucial to the bioactivity of target compounds. Orthogonal projection to latent structures model reveals that increasing the QPlogPo/w of the compound can increase activity. Molecular dynamics simulations revealed key roles of residues (VAL469 and ILE472) as active binding sites of Kv1.3 for binding of specific compound. Notopterol (Z4), the most potent Kv1.3 inhibitor (IC50 = 311.90 ± 1.24 nM), significantly suppressed IFN-γ release from CD4+ T cells, whereas, Kv1.3 inactive compound Z20 at 5 μM showed no significant difference in IFN-γ release from CD4+ T cells. In atopic dermatitis rat model, Notopterol reduced epidermal thickening, IgE, Kv1.3, IL-1β production, and infiltration of CD4+ T cells and mast cells. These findings establish Notopterol as a promising Kv1.3 inhibitor for therapeutic applications in immune disorders.

Structure-Based Discovery Targeting GSK-3α Reveals Potent Nanomolar Selective 4-Phenyl-1H-benzofuro[3,2-b]pyrazolo[4,3-e]pyridine Inhibitor with Promising Glioblastoma and CNS-Active Potential in Cellular Models

Glycogen synthase kinase-3 (GSK-3) is linked with multiple CNS conditions, including glioblastoma (GBM). Compared to the GSK-3β isoform, structure-based inhibitor design targeting GSK-3α is limited. Virtual screening was employed to identify GSK-3α inhibitors with CNS-active potential. Using a GSK-3α homology model, an optimized protocol with three-dimensional (3D)-pharmacophore filtering and Glide-SP docking was used to screen the ZINC20 biogenic subset. From 14 compounds selected for binding assay validation, three novel hit compounds were identified, with 1 (4-phenyl-1H-benzofuro[3,2-b]pyrazolo[4,3-e]pyridine scaffold) exhibiting nanomolar activity against GSK-3α/β (IC50s ∼ 0.26 μM). Selectivity profiling (12 homologous kinases) revealed selectivity for GSK-3α/β and protein kinase A (PKA). Compound 1 was more potent against three GBM cell lines (cell viability IC50s = 3-6 μM at 72 h) compared to benchmark GSK-3 inhibitor, 4-benzyl-2-methyl-1,2,4-thiadiazolidine-3,5-dione (TDZD-8), and nontoxic to human astrocytes. It demonstrated CNS-active potential in an all-human in vitro blood-brain barrier GBM model, good in vitro metabolic stability, excellent predicted oral bioavailability and represents a promising lead compound for development.

Discovery and Characterization of Novel Receptor-Interacting Protein Kinase 1 Inhibitors Using Deep Learning and Virtual Screening

Receptor-interacting protein kinase 1 (RIPK1) serves as a critical mediator of cell necroptosis and represents a promising therapeutic target for various human neurodegenerative diseases and inflammatory diseases. Nonetheless, the RIPK1 inhibitors currently reported are inadequate for clinical research due to suboptimal inhibitory activities or lack of selectivity. Consequently, there is a need for the discovery of novel RIPK1 kinase inhibitors. In this study, we integrated a deep learning model, specifically the fingerprint graph attention network (FP-GAT), with molecular docking-based virtual screening to identify potential RIPK1 inhibitors from a library comprising 13 million compounds. Out of 43 compounds procured, two compounds (designated as 24 and 41) demonstrated enzyme inhibition activity exceeding 50% at a concentration of 10 μM against RIPK1. The half-maximal inhibitory concentrations (IC50) for compounds 24 and 41 were determined to be 2.01 and 2.95 μM, respectively. Furthermore, these compounds exhibited protective effects in an HT-29 cell model of TSZ-induced necroptosis, with half-maximal effective concentrations (EC50) of 6.77 μM for compound 24 and 68.70 μM for compound 41. Finally, molecular dynamics simulations and binding free energy calculations were conducted to elucidate the molecular mechanism of compounds 24 and 41 binding to RIPK1. The results show that Met92, Met95, Ala155, and Asp156 are key residues for novel RIPK1 inhibitors. In summary, this work discovered two hit compounds targeting RIPK1, which can be further structurally modified to become promising lead compounds.

Enhancing MM/P(G)BSA Methods: Integration of Formulaic Entropy for Improved Binding Free Energy Calculations

Balancing computational efficiency and precision, MM/P(G)BSA methods have been widely employed in the estimation of binding free energies within biological systems. However, the entropy contribution to the binding free energy is often neglected in MM/P(G)BSA calculations, due to the computational cost of conventional methods such as normal mode analysis (NMA). In this work, the entropy effect using a formulaic entropy can be computed from one single structure according to variations in the polar and non-polar solvents accessible surface areas and the count of rotatable bonds in ligands. It was incorporated into MM/P(G)BSA methods to enhance their performance. Extensive benchmarking reveals that the integration of formulaic entropy systematically elevates the performance of both MM/PBSA and MM/GBSA without incurring additional computational expenses. In addition, we found the inclusion of dispersion can deteriorate the correlation performance (Rp) but reduce the root mean square error (RMSE) of both MM/PBSA and MM/GBSA. Notably, MM/PBSA_S, including the formulaic entropy but excluding the dispersion, surpasses all other MM/P(G)BSA methods across a spectrum of datasets. Our investigation furnishes a valuable and practical MM/P(G)BSA method, optimizing binding free energy calculations for a variety of biological systems.

Integrating machine learning driven virtual screening and molecular dynamics simulations to identify potential inhibitors targeting PARP1 against prostate cancer

Prostate cancer (PC) is one of the most common types of malignancies in men, with a noteworthy increase in newly diagnosed cases in recent years. PARP1 is a ubiquitous nuclear enzyme involved in DNA repair, nuclear transport, ribosome synthesis, and epigenetic bookmarking. In this study, a library of 9000 phytochemicals was screened, with a focus on those with high drug efficacy and potential PARP1 inhibition. Different machine learning models were generated and assessed using various statistical measures. The RF model outperformed all other models in terms of accuracy (0.9489), specificity (0.9171), and area under the curve (AUC = 0.9846). Following this, a library of 9510 phytochemicals was screened, yielding 181 compounds predicted to be active. These compounds were subsequently assessed using Lipinski's Rule of Five, yielding 40 interesting candidates. Molecular docking experiments demonstrated that compound ZINC2356684563, ZINC2356558598, and ZINC14584870, had strong affinity for the PARP1 active site. Further molecular dynamics simulations and MM-PBSA validated the stability of the ligand-protein complexes, with ZINC14584870 and ZINC43120769 demonstrating the most stable interaction, as seen by low RMSD and RMSF levels. Our findings emphasize the potential of these phytochemical inhibitors as novel therapeutic agents against PARP1 in prostate cancer treatment, paving the path for further experimental validation and clinical investigations. These results open new possibilities for developing treatments to benefit prostate cancer patients.

InertDB as a generative AI-expanded resource of biologically inactive small molecules from PubChem

The development of robust artificial intelligence (AI)-driven predictive models relies on high-quality, diverse chemical datasets. However, the scarcity of negative data and a publication bias toward positive results often hinder accurate biological activity prediction. To address this challenge, we introduce InertDB, a comprehensive database comprising 3,205 curated inactive compounds (CICs) identified through rigorous review of over 4.6 million compound records in PubChem. CIC selection prioritized bioassay diversity, determined using natural language processing (NLP)-based clustering metrics, while ensuring minimal biological activity across all evaluated bioassays. Notably, 97.2% of CICs adhere to the Rule of Five, a proportion significantly higher than that of overall PubChem dataset. To further expand the chemical space, InertDB also features 64,368 generated inactive compounds (GICs) produced using a deep generative AI model trained on the CIC dataset. Compared to conventional approaches such as random sampling or property-matched decoys, InertDB significantly improves predictive AI performance, particularly for phenotypic activity prediction by providing reliable inactive compound sets.Scientific contributionsInertDB addresses a critical gap in AI-driven drug discovery by providing a comprehensive repository of biologically inactive compounds, effectively resolving the scarcity of negative data that limits prediction accuracy and model reliability. By leveraging language model-based bioassay diversity metrics and generative AI, InertDB integrates rigorously curated inactive compounds with an expanded chemical space. InertDB serves as a valuable alternative to random sampling and decoy generation, offering improved training datasets and enhancing the accuracy of phenotypic pharmacological activity prediction.

Spectra-descriptor-based machine learning for predicting protein-ligand interactions

Machine learning models have emerged as powerful tools for drug discovery of lead compounds. Nevertheless, despite notable advances in model architectures, research on more reliable and physicochemical-based descriptors for molecules and proteins remains limited. To address this gap, we introduce the Fragment Integral Spectrum Descriptor (FISD), aimed at utilizing the spatial configuration and electronic structure information of molecules and proteins, as a novel physicochemical descriptor for virtual screening models. Validation demonstrates that the combination of FISD and a classical neural network model achieves performance comparable to that of complex models paired with conventional structural descriptors. Furthermore, we successfully predict and screen potential binding ligands for two given protein targets, showcasing the broad applicability and practicality of FISD. This research enriches the molecular and protein representation strategies of machine learning and accelerates the process of drug discovery.

Discovery of farnesoid X receptor antagonists from Salvia miltiorrhiza based on virtual screening and activity verification

The farnesoid X receptor (FXR) is a promising therapeutic target for the treatment of non-alcoholic fatty liver disease (NAFLD). Salvia miltiorrhiza, a traditional Chinese medicine, has demonstrated significant efficacy in the prevention and treatment of liver diseases. Consequently, investigating the potential effects of Salvia miltiorrhiza on FXR could provide new insights for NAFLD treatment. This study explores whether active ingredients from Salvia miltiorrhiza can target FXR and serve as therapeutic agents for treating NAFLD. The findings revealed that cynaroside and lithospermic acid displayed strong FXR antagonistic activity, with IC50 values of 5.41 ± 1.08 μM and 16.92 ± 2.68 μM, respectively. Salvianolic acid A also showed moderate activity (IC50 = 56.35 ± 4.54 μM). MTT assays demonstrated that these three compounds were non-toxic to HepG2 and LO2 cells at a concentration of 200 μM. Molecular dynamics simulations were conducted to elucidate the interaction mechanisms of cynaroside and lithospermic acid with FXR. These results suggest that cynaroside and lithospermic acid from Salvia miltiorrhiza may be potential candidates for targeting FXR in treating NAFLD.

Rapid traversal of vast chemical space using machine learning-guided docking screens

The accelerating growth of make-on-demand chemical libraries provides unprecedented opportunities to identify starting points for drug discovery with virtual screening. However, these multi-billion-scale libraries are challenging to screen, even for the fastest structure-based docking methods. Here we explore a strategy that combines machine learning and molecular docking to enable rapid virtual screening of databases containing billions of compounds. In our workflow, a classification algorithm is trained to identify top-scoring compounds based on molecular docking of 1 million compounds to the target protein. The conformal prediction framework is then used to make selections from the multi-billion-scale library, reducing the number of compounds to be scored by docking. The CatBoost classifier showed an optimal balance between speed and accuracy and was used to adapt the workflow for screens of ultralarge libraries. Application to a library of 3.5 billion compounds demonstrated that our protocol can reduce the computational cost of structure-based virtual screening by more than 1,000-fold. Experimental testing of predictions identified ligands of G protein-coupled receptors and demonstrated that our approach enables discovery of compounds with multi-target activity tailored for therapeutic effect.

T6496 targeting EGFR mediated by T790M or C797S mutant: machine learning, virtual screening and bioactivity evaluation study

Acquired resistance to EGFR is a major impediment in lung cancer treatment, highlighting the urgent need to discover novel compounds to overcome EGFR drug resistance. In this study, we utilized in silico methods and bioactivity evaluation for drug discovery to identify novel active anticancer agents targeting EGFRT790M/L858R and EGFRT790M/C797S/L858R. Firstly, we employed ROC-guided machine learning to retrieve nearly 7,765 compounds from a collection of three libraries (comprising over 220,000 compounds). Next, virtual screening, cluster analysis, and binding model analysis were employed to identify six potential compounds. Additionally, the kinase assay revealed that these six compounds demonstrated higher sensitivity to EGFR than c-Met. Among these compounds, T6496 inhibited both EGFRT790M/L858R and EGFRT790M/C797S/L858R kinases, with an IC50 of 3.30 and 8.72 μM. Furthermore, we evaluated the antitumor effects of the six selected compounds, and compound T6496 exhibited the strongest anticancer activity against H1975 cell lines, with an IC50 value of 2.7 μM. These results suggest that T6496 may mitigate EGFR resistance caused by T790M or C797S mutations. Moreover, the AO staining assay, JC-1 staining, ROS experiment and hemolytic toxicity evaluation revealed that T6496 could induce apoptosis in H1975 cell lines in a time-dependent and concentration-dependent manner, and is a potential compound for further structural optimization.

Discover the Power of Lithospermic Acid as Human Carbonic Anhydrase VA and Pancreatic Lipase Inhibitor Through In Silico and In Vitro Studies

Obesity remains a significant global health concern, with limited pharmacological options that balance efficacy and safety. In this study, we identified lithospermic acid (LTS0059529) from Salvia miltiorrhiza as a potential dual inhibitor of pancreatic lipase (PL) and human carbonic anhydrase VA (hCA VA), two key enzymes in lipid metabolism. Using molecular docking and dynamics simulations, we observed that lithospermic acid interacts with Zn²⁺ in hCA VA via its benzofuran carboxylate moiety and forms stable complexes with PL through hydrogen bonding with ASP 205 and π-stacking interactions with PHE 77 and PHE 215. Experimental validation confirmed its inhibitory activity, with Ki values of 33.1 ± 1.6 μM for PL and 0.69 ± 0.01 μM for hCA VA. While its inhibition of hCA VA is not isoform-specific, lithospermic acid demonstrates significant potential as a dual inhibitor, targeting complementary pathways in obesity management. This study is the first to explore its dual action on PL and hCA VA, highlighting a promising strategy for future antiobesity therapies. Further research will focus on optimizing selectivity and potency to develop safer and more effective treatments.

Identification of novel covalent JAK3 inhibitors through consensus scoring virtual screening: integration of common feature pharmacophore and covalent docking

Accumulated research strongly indicates that Janus kinase 3 (JAK3) is intricately involved in the initiation and advancement of a diverse range of human diseases, underscoring JAK3 as a promising target for therapeutic intervention. However, JAK3 shows significant homology with other JAK family isoforms, posing substantial challenges in the development of JAK3 inhibitors. To address these limitations, one strategy is to design selective covalent JAK3 inhibitors. Therefore, this study introduces a virtual screening approach that combines common feature pharmacophore modeling, covalent docking, and consensus scoring to identify novel inhibitors for JAK3. First, common feature pharmacophore models were constructed based on a selection of representative covalent JAK3 inhibitors. The optimal qualitative pharmacophore model proved highly effective in distinguishing active and inactive compounds. Second, 14 crystal structures of the JAK3-covalent inhibitor complex were chosen for the covalent docking studies. Following validation of the screening performance, 5TTU was identified as the most suitable candidate for screening potential JAK3 inhibitors due to its higher predictive accuracy. Finally, a virtual screening protocol based on consensus scoring was conducted, integrating pharmacophore mapping and covalent docking. This approach resulted in the discovery of multiple compounds with notable potential as effective JAK3 inhibitors. We hope that the developed virtual screening strategy will provide valuable guidance in the discovery of novel covalent JAK3 inhibitors.

CPI-GGS: A deep learning model for predicting compound-protein interaction based on graphs and sequences

BACKGROUND

Compound-protein interaction (CPI) is essential to drug discovery and design, where traditional methods are often costly and have low success rates. Recently, the integration of machine learning and deep learning in CPI research has shown potential to reduce costs and enhance discovery efficiency by improving protein target identification accuracy. Additionally, with an urgent need for novel therapies against complex diseases, CPI investigation could lead to the identification of effective new drugs. Since drug-target interactions involve complex biological processes, refined models are necessary for precise feature extraction and analysis. Nevertheless, current CPI prediction methods still face significant limitations: predictions lack sufficient accuracy, models require improved generalization ability, and further validation across diverse datasets remains essential.

RESULTS

To address some issues at the current stage, this paper proposes a combined deep learning method, CPI-GGS, for predicting and analyzing compound-protein interactions. The source code is available on GitHub at https://github.com/xingjie321/CPI-GGS.

CONCLUSIONS

The experimental results demonstrate improved accuracy in predicting compound-protein interactions and enhance the understanding of how compounds and proteins interact, providing a valuable new tool for drug discovery and development.

Folding-Based End-To-End Chemical Drug Design with Uncertainty Estimation: Tackling Hallucination in the Post-GPT Era

In the post-GPT era, Llama-Gram represents a promising advancement in AI-driven chemical drug discovery, grounded in the chemical principle that molecular structure determines properties. This folding-based end-to-end framework seeks to address the hallucination issues of traditional large language models by integrating protein folding embeddings, graph-based molecular representations, and uncertainty estimation to better capture the structural complexities of protein-ligand interactions. By leveraging the frozen-gradient ESMFold model and a Graph Transformer variant, Llama-Gram aims to enhance predictive accuracy and reliability through grouped-query attention and a Gram layer inspired by support points theory. By incorporating protein folding information, the model demonstrates competitive performance against state-of-the-art approaches such as Transformer CPI 2.0 and Graph-DTA, offering improvements in compound-target interaction. Llama-Gram provides a scalable and innovative chemical theory that could contribute to accelerating the chemical drug discovery process.

Identification of Small-Molecule Inhibitors for Enterovirus A71 IRES by Structure-Based Virtual Screening

Structured RNAs play a crucial role in regulating gene expression, which includes both protein synthesis and RNA processing. Dysregulation of these processes is associated with various conditions, including viral and bacterial infections, as well as cancer. The unique tertiary structures of structured RNAs provide an opportunity for small molecules to directly modulate such processes, making them promising targets for drug discovery. Although small-molecule inhibitors targeting RNA have shown early success, in silico strategies like structure-based virtual screening remain underutilized for RNA-targeted drug discovery. In this study, we developed a virtual screening scheme targeting the structural ensemble of EV-A71 IRES SL II, a noncoding viral RNA element essential for viral replication. We subsequently optimized the experimentally validated hit compound IRE-03 from virtual screening through an "analog-by-catalog" search. This led to the identification of a more potent IRES inhibitor, IRE-03-3, validated through biochemical and functional assays with an EC50 value of 11.96 μM against viral proliferation. Our findings demonstrate that structure-based virtual screening can be effectively applied to RNA targets, providing exciting new opportunities for future antiviral drug discovery.

iScore: A ML-Based Scoring Function for De Novo Drug Discovery

In the quest for accelerating de novo drug discovery, the development of efficient and accurate scoring functions represents a fundamental challenge. This study introduces iScore, a novel machine learning (ML)-based scoring function designed to predict the binding affinity of protein-ligand complexes with remarkable speed and precision. Uniquely, iScore circumvents the conventional reliance on explicit knowledge of protein-ligand interactions and a full picture of atomic contacts, instead leveraging a set of ligand and binding pocket descriptors to directly evaluate binding affinity. This approach enables skipping the inefficient and slow conformational sampling stage, thereby enabling the rapid screening of ultrahuge molecular libraries, a crucial advancement given the practically infinite dimensions of chemical space. iScore was rigorously trained and validated using the PDBbind 2020 refined set, CASF 2016, CSAR NRC-HiQ Set1/2, DUD-E, and target fishing data sets, employing three distinct ML methodologies: Deep neural network (iScore-DNN), random forest (iScore-RF), and eXtreme gradient boosting (iScore-XGB). A hybrid model, iScore-Hybrid, was subsequently developed to incorporate the strengths of these individual base learners. The hybrid model demonstrated a Pearson correlation coefficient (R) of 0.78 and a root-mean-square error (RMSE) of 1.23 in cross-validation, outperforming the individual base learners and establishing new benchmarks for scoring power (R = 0.814, RMSE = 1.34), ranking power (ρ = 0.705), and screening power (success rate at top 10% = 73.7%). Moreover, iScore-Hybrid demonstrated great performance in the target fishing benchmarking study.

Chemical Space Exploration and Machine Learning-Based Screening of PDE7A Inhibitors

Background/Objectives: Phosphodiesterase 7 (PDE7), a member of the PDE superfamily, selectively catalyzes the hydrolysis of cyclic adenosine 3',5'-monophosphate (cAMP), thereby regulating the intracellular levels of this second messenger and influencing various physiological functions and processes. There are two subtypes of PDE7, PDE7A and PDE7B, which are encoded by distinct genes. PDE7 inhibitors have been shown to exert therapeutic effects on neurological and respiratory diseases. However, FDA-approved drugs based on the PDE7A inhibitor are still absent, highlighting the need for novel compounds to advance PDE7A inhibitor development. Methods: To address this urgent and important issue, we conducted a comprehensive cheminformatics analysis of compounds with potential for PDE7A inhibition using a curated database to elucidate the chemical characteristics of the highly active PDE7A inhibitors. The specific substructures that significantly enhance the activity of PDE7A inhibitors, including benzenesulfonamido, acylamino, and phenoxyl, were identified by an interpretable machine learning analysis. Subsequently, a machine learning model employing the Random Forest-Morgan pattern was constructed for the qualitative and quantitative prediction of PDE7A inhibitors. Results: As a result, six compounds with potential PDE7A inhibitory activity were screened out from the SPECS compound library. These identified compounds exhibited favorable molecular properties and potent binding affinities with the target protein, holding promise as candidates for further exploration in the development of potent PDE7A inhibitors. Conclusions: The results of the present study would advance the exploration of innovative PDE7A inhibitors and provide valuable insights for future endeavors in the discovery of novel PDE inhibitors.

A Hybrid Energy-Based and AI-Based Screening Approach for the Discovery of Novel Inhibitors of AXL

AXL, part of the TAM receptor tyrosine kinase family, plays a significant role in the growth and survival of various tissues and tumors, making it a critical target for cancer therapy. This study introduces a novel high-throughput virtual screening (HTVS) methodology that merges an AI-enhanced graph neural network, PLANET, with a geometric deep learning algorithm, DeepDock. Using this approach, we identified potent AXL inhibitors from our database. Notably, compound 9, with an IC50 of 9.378 nM, showed excellent inhibitory activity, suggesting its potential as a candidate for further research. We also performed molecular dynamics simulations to explore the interactions between compound 9 and AXL, providing insights for future enhancements. This hybrid screening method proves effective in finding promising AXL inhibitors, and advancing the development of new cancer therapies.

The Need for Continuing Blinded Pose- and Activity Prediction Benchmarks

Computational tools for structure-based drug design (SBDD) are widely used in drug discovery and can provide valuable insights to advance projects in an efficient and cost-effective manner. However, despite the importance of SBDD to the field, the underlying methodologies and techniques have many limitations. In particular, binding pose and activity predictions (P-AP) are still not consistently reliable. We strongly believe that a limiting factor is the lack of a widely accepted and established community benchmarking process that independently assesses the performance and drives the development of methods, similar to the CASP benchmarking challenge for protein structure prediction. Here, we provide an overview of P-AP, unblinded benchmarking data sets, and blinded benchmarking initiatives (concluded and ongoing) and offer a perspective on learnings and the future of the field. To accelerate a breakthrough on the development of novel P-AP methods, it is necessary for the community to establish and support a long-term benchmark challenge that provides nonbiased training/test/validation sets, a systematic independent validation, and a forum for scientific discussions.

Natural Language Processing Methods for the Study of Protein-Ligand Interactions

Natural Language Processing (NLP) has revolutionized the way computers are used to study and interact with human languages and is increasingly influential in the study of protein and ligand binding, which is critical for drug discovery and development. This review examines how NLP techniques have been adapted to decode the "language" of proteins and small molecule ligands to predict protein-ligand interactions (PLIs). We discuss how methods such as long short-term memory (LSTM) networks, transformers, and attention mechanisms can leverage different protein and ligand data types to identify potential interaction patterns. Significant challenges are highlighted including the scarcity of high-quality negative data, difficulties in interpreting model decisions, and sampling biases in existing data sets. We argue that focusing on improving data quality, enhancing model robustness, and fostering both collaboration and competition could catalyze future advances in machine-learning-based predictions of PLIs.

Pharmacophore screening, molecular docking, and MD simulations for identification of VEGFR-2 and c-Met potential dual inhibitors

Introduction

The vascular endothelial growth factor receptor 2 (VEGFR-2) and the mesenchymal-epithelial transition factor (c-Met) are critical in the pathogenesis and progression of various cancers by synergistically contributing to angiogenesis and tumor progression. The development of dual-target inhibitors for VEGFR-2 and c-Met holds promise for more effective cancer therapies that could overcome tumor cell resistance, a limitation often observed with inhibitors targeting a single receptor.

Methods

In this study, a computational virtual screening approach involving drug likeness evaluation, pharmacophore modeling and molecular docking was employed to identify VEGFR-2/c-Met dual-target inhibitors from ChemDiv database. Subsequent molecular dynamics (MD) simulations and MM/PBSA calculations were conducted to assess the stability of the protein-ligand interactions.

Results

From the virtual screening process, 18 hit compounds were identified to exhibit potential inhibitory activity against VEGFR-2 and c-Met. Among them, compound17924 and compound4312 possessed the best inhibitory potential according to our screening criteria.

Discussion

The analysis of the MD simulation results indicated that compound17924 and compound4312 showed superior binding free energies to both VEGFR-2 and c-Met when compared to the positive ligands. These findings suggested that both compounds were promising candidates for further drug development and could potentially serve as improved alternatives of cancer therapeutics.

Knowledge-guided diffusion model for 3D ligand-pharmacophore mapping

Pharmacophores are abstractions of essential chemical interaction patterns, holding an irreplaceable position in drug discovery. Despite the availability of many pharmacophore tools, the adoption of deep learning for pharmacophore-guided drug discovery remains relatively rare. We herein propose a knowledge-guided diffusion framework for 'on-the-fly' 3D ligand-pharmacophore mapping, named DiffPhore. It leverages ligand-pharmacophore matching knowledge to guide ligand conformation generation, meanwhile utilizing calibrated sampling to mitigate the exposure bias of the iterative conformation search process. By training on two self-established datasets of 3D ligand-pharmacophore pairs, DiffPhore achieves state-of-the-art performance in predicting ligand binding conformations, surpassing traditional pharmacophore tools and several advanced docking methods. It also manifests superior virtual screening power for lead discovery and target fishing. Using DiffPhore, we successfully identify structurally distinct inhibitors for human glutaminyl cyclases, and their binding modes are further validated through co-crystallographic analysis. We believe this work will advance the AI-enabled pharmacophore-guided drug discovery techniques.

Repurposing FDA-approved drugs to target G-quadruplexes in breast cancer

Breast cancer, a leading cause of cancer-related mortality in women, is characterized by genomic instability and aberrant gene expression, often influenced by noncanonical nucleic acid structures such as G-quadruplexes (G4s). These structures, commonly found in the promoter regions and 5'-untranslated RNA sequences of several oncogenes, play crucial roles in regulating transcription and translation. Stabilizing these G4 structures offers a promising therapeutic strategy for targeting key oncogenic pathways. In this study, we employed a drug repurposing approach to identify FDA-approved drugs capable of binding and stabilizing G4s in breast cancer-related genes. Using ligand-based virtual screening and biophysical methods, we identified several promising compounds, such as azelastine, belotecan, and irinotecan, as effective G4 binders, with significant antiproliferative effects in breast cancer cell lines. Notably, belotecan and irinotecan exhibited a synergistic mechanism, combining G4 stabilization with their established topoisomerase I inhibition activity to enhance cytotoxicity in cancer cells. Our findings support the therapeutic potential of G4 stabilization in breast cancer, validate drug repurposing as an efficient strategy to identify G4-targeting drugs, and highlight how combining G4 stabilization with other established drug activities may improve anticancer efficacy.

Increase Docking Score Screening Power by Simple Fusion With CNNscore

Scoring functions (SFs) of molecular docking is a vital component of structure-based virtual screening (SBVS). Traditional SFs yield their inherent shortage for idealized approximations and simplifications predicting the binding affinity. Complementarily, SFs based on deep learning (DL) have emerged as powerful tools for capturing intricate features within protein-ligand (PL) interactions. We here present a docking-score fusion strategy that integrates pose scores derived from GNINA's convolutional neural network (CNN) with traditional docking scores. Extensive validation on diverse datasets has shown that by means of multiplying Watvina docking score by CNNscore demonstrates state-of-the-art screening power. Furthermore, in a reverse practice, our docking-score fusion technique was incorporated into the virtual screening (VS) workflow aimed at identifying inhibitors of the challenging target TYK2. Two promising hits with IC50 9.99 μM and 13.76 μM in vitro were identified from nearly 12 billion molecules.

DockEM: an enhanced method for atomic-scale protein-ligand docking refinement leveraging low-to-medium resolution cryo-EM density maps

Protein-ligand docking plays a pivotal role in virtual drug screening, and recent advancements in cryo-electron microscopy (cryo-EM) technology have significantly accelerated the progress of structure-based drug discovery. However, the majority of cryo-EM density maps are of medium to low resolution (3-10 Å), which presents challenges in effectively integrating cryo-EM data into molecular docking workflows. In this study, we present an updated protein-ligand docking method, DockEM, which leverages local cryo-EM density maps and physical energy refinement to precisely dock ligands into specific protein binding sites. Tested on a dataset of 121 protein-ligand compound, our results demonstrate that DockEM outperforms other advanced docking methods. The strength of DockEM lies in its ability to incorporate cryo-EM density map information, effectively leveraging the structural information of ligands embedded within these maps. This advancement enhances the use of cryo-EM density maps in virtual drug screening, offering a more reliable framework for drug discovery.

LeScore: a scoring function incorporating hydrogen bonding penalty for protein-ligand docking

CONTEXT

Molecular docking is vital for structure-based virtual screening and heavily depends on accurate and robust scoring functions. Scoring functions often inadequately account for the breakage of solvent hydrogen bonds, hindering the accuracy of predicting binding energy. Here, we introduce LeScore, a novel scoring function that specifically incorporates the hydrogen bonding penalty (HBP) in an aqueous environment, aiming to penalize unfavorable polar interactions when hydrogen bonds with water are broken but the energy loss is not fully compensated by newly formed protein-ligand interactions. LeScore was optimized for descriptor combinations and subsequently validated using a testing data set, achieving a Pearson correlation coefficient (rp) of 0.53 in the training set and 0.52 in the testing set. To evaluate its screening capability, a subset of the Directory of Useful Decoys: Enhanced (DUD-E) was used. And LeScore achieved an AUC of 0.71 for specific targets, outperforming models without HBP and enhancing the ranking and classification of active compounds. Overall, LeScore provides a robust tool for improving virtual screening, especially in cases where hydrogen bonding is crucial for ligand binding.

METHOD

LeScore is formulated as a linear combination of descriptors, including van der Waals interactions, hydrogen bond energy, ligand strain energy, and newly integrated HBP. The function was optimized using multiple linear regression (MLR) on the PDBbind 2019 dataset. Evaluation metrics, such as Pearson and Spearman correlation coefficients were utilized to assess the performance of 12 descriptor combinations. Additionally, the study employed datasets from the DUD-E to evaluate LeScore's ability to distinguish active ligands from decoys across multiple target proteins.

GNINA 1.3: the next increment in molecular docking with deep learning

Computer-aided drug design has the potential to significantly reduce the astronomical costs of drug development, and molecular docking plays a prominent role in this process. Molecular docking is an in silico technique that predicts the bound 3D conformations of two molecules, a necessary step for other structure-based methods. Here, we describe version 1.3 of the open-source molecular docking software GNINA. This release updates the underlying deep learning framework to PyTorch, resulting in more computationally efficient docking and paving the way for seamless integration of other deep learning methods into the docking pipeline. We retrained our CNN scoring functions on the updated CrossDocked2020 v1.3 dataset and introduce knowledge-distilled CNN scoring functions to facilitate high-throughput virtual screening with GNINA. Furthermore, we add functionality for covalent docking, where an atom of the ligand is covalently bound to an atom of the receptor. This update expands the scope of docking with GNINA and further positions GNINA as a user-friendly, open-source molecular docking framework. GNINA is available at https://github.com/gnina/gnina .Scientific contributions: GNINA 1.3 is an open source a molecular docking tool with enhanced support for covalent docking and updated deep learning models for more effective docking and screening.

Docking guidance with experimental ligand structural density improves docking pose prediction and virtual screening performance

Recent advances in structural biology have led to the publication of a wealth of high-resolution x-ray crystallography (XRC) and cryo-EM macromolecule structures, including many complexes with small molecules of interest for drug design. While it is common to incorporate information from the atomic coordinates of these complexes into docking (e.g., pharmacophore models or scaffold hopping), there are limited methods to directly leverage the underlying density information. This is desirable because it does not rely on the determination of relevant coordinates, which may require expert intervention, but instead interprets all density as indicative of regions to which a ligand may be bound. To do so, we have developed CryoXKit, a tool to incorporate experimental densities from either cryo-EM or XRC as a biasing potential on heavy atoms during docking. Using this structural density guidance with AutoDock-GPU, we found significant improvements in re-docking and cross-docking, important pose prediction tasks, compared with the unmodified AutoDock4 force field. Failures in cross-docking tasks are additionally reflective of changes in the positioning of pharmacophores in the site, suggesting it is a fundamental limitation of transferring information between complexes. We additionally found, against a set of targets selected from the LIT-PCBA dataset, that rescoring of these improved poses leads to better discriminatory power in virtual screenings for selected targets. Overall, CryoXKit provides a user-friendly method for improving docking performance with experimental data while requiring no a priori pharmacophore definition and at virtually no computational expense. Map-modification code available at: https://github.com/forlilab/CryoXKit.

A transferability-guided protein-ligand interaction prediction method

Accurate prediction of protein-ligand interaction (PLI) is crucial for drug discovery and development. However, existing methods often struggle with effectively integrating heterogeneous protein and ligand data modalities and optimizing knowledge transfer from pretraining to the target task. This paper proposes a novel transferability-guided PLI prediction method that maximizes knowledge transfer by deeply integrating protein and ligand representations through a cross-attention mechanism and incorporating transferability metrics to guide fine-tuning. The cross-attention mechanism facilitates interactive information exchange between modalities, enabling the model to capture intricate interdependencies. Meanwhile, the transferability-guided strategy quantifies transferability from pretraining tasks and incorporates it into the training objective, ensuring the effective utilization of beneficial knowledge while mitigating negative transfer. Extensive experiments demonstrate significant and consistent improvements over traditional fine-tuning, validated by statistical tests. Ablation studies highlight the pivotal role of cross-attention, and quantitative analysis reveals the method's ability to reduce harmful transfer. Our guided strategy provides a paradigm for more comprehensive utilization of pretraining knowledge, offering prospects for enhancing other PLI prediction approaches. This method advances PLI prediction via innovative modality fusion and guided knowledge transfer, paving the way for accelerated drug discovery pipelines. Code and data are freely available at https://github.com/brian-zZZ/Guided-PLI.

Identification of Potential PBP2a Inhibitors Against Methicillin-Resistant Staphylococcus aureus via Drug Repurposing and Combination Therapy

Methicillin-resistant Staphylococcus aureus (MRSA) achieves high-level resistance against β-lactam antibiotics through the expression of penicillin-binding protein 2a (PBP2a), which features a closed active site that impedes antibiotic binding. Herein, we implemented a strategy that combines drug repurposing with synergistic therapy to identify potential inhibitors targeting PBP2a's allosteric site from an FDA-approved drug database. Initially, retrospective verifications were conducted, employing different Glide docking methods (HTVS, SP, and XP) and two representative PBP2a structures. The combination of Glide SP and one representative PBP2a conformation showed the highest efficacy in identifying active compounds. The optimized parameters were then utilized to screen FDA-approved drugs, and 15 compounds were shortlisted for potential combination therapy with cefazolin, an ineffective cephalosporin against MRSA. Through biological assays-checkerboard, time-kill assays, and live/dead bacterial staining-we discovered that four compounds exhibited robust bactericidal activity (FICI < 0.5) compared to both untreated control and monotherapy with cefazolin alone. Scanning electron microscopy (SEM) confirmed that while cefazolin alone did not cause visible damage to MRSA cells, the combination treatment markedly induced cell lysis. Additional MM-GBSA studies underscored the strong binding affinity of mitoxantrone to the allosteric site. These findings introduce a combination therapy approach that potentially restores MRSA's susceptibility to β-lactam antibiotics.