Journal of Chemical Information and Modeling 最新文献

{"title":"FlexENN: A Graph\u0000Neural Network for Binding Energy\u0000Prediction of Globular and Intrinsically Disordered Proteins","authors":"Maryum Irshad,Kassandra Ori-McKenney,Ruxandra I. Dima","doi":"10.1021/acs.jcim.6c01545","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c01545","url":null,"abstract":"Abstract Intrinsically disordered proteins (IDPs) drive a large fraction of cellular signaling, transcription, and assembly through interfaces that often lack a defined geometry, making the prediction of their binding energies challenging for methods developed primarily for folded complexes. Here, we introduce FlexENN, an interface-focused message-passing graph neural network framework comprising two independently trained models: FlexENN-F for folded protein complexes and FlexENN-I for complexes containing a disordered partner. The FlexENN architecture constructs a local interface graph in which each residue node integrates the relative geometry, flexibility-related structural descriptors, and sequence information, allowing the network to learn structure-affinity relationships from a single resolved bound conformation. We also construct an IDP-containing dataset for training the FlexENN-I model. Benchmarked across folded complexes and IDP-containing systems, including tubulin–tubulin interfaces within microtubules (MTs) and MTs in complex with microtubule-associated proteins (MAPs), with reference binding affinity values extracted from our experiments, the models reproduce physically reasonable energetic profiles, including the stronger interdimer relative to the lateral MT interactions, while feature-ablation analyses reveal system-dependent contributions of selected descriptors to the predictions. More broadly, this work delivers an accurate, structure-aware approach to predicting binding free energies for complexes of folded protein and complexes that contain ordered or partially ordered IDP regions for which a sufficiently well-defined bound structure is available.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"7 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149151217","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sequence Dependence\u0000of Folding and Self-Assembly of\u0000Glycine-X Repeat Peptides on Graphite","authors":"Eduardo Cubillos-Llantén,Julius A. Neumann,Horacio Poblete,Jeffrey Comer","doi":"10.1021/acs.jcim.6c02272","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02272","url":null,"abstract":"Abstract Cyclic peptides consisting of glycine-X repeats have been shown in simulation and experiment to fold at the graphite–water interface into beta-hairpins, which then self-assemble by intermolecular hydrogen bonding. In this contribution, we investigate the role of the hairpin turn sequence, strand length, and the X amino acid on folding and self-assembly. Simulations of relatively large systems with 9 initially unfolded peptides allow us to observe spontaneous folding, which, together with free energy profiles as a function of RMSD, enable comparison of folding thermodynamics and kinetics among sequences and identification of metastable misfolded conformations. Likewise, spontaneous assembly also occurs in these simulations, which, together with absolute binding free energy calculations, enable comparison of assembly kinetics and preferences for different intermolecular alignments. We find that the GPGG sequence best enforces folding with the desired hairpin turn compared to GGGG and GDGG sequences. While our previous peptide design was head-to-tail cyclized, here, we demonstrate that disulfide cyclization is a viable alternative. Peptides having strands with GT, GA, GG, GS, and GD repeats, as well as those with charge-complementary GE/GZ pairs (where Z is diaminobutyric acid), can fold correctly, but vary in their folding free energy, tendency to adopt misfolded conformations, folding cooperativity, and preference for different intermolecular alignments. GT-repeat strands most favor folding into the desired beta-strands; however, they yield the weakest intermolecular assembly. GG-repeats yield the strongest intermolecular assembly, but peptides with all GG repeats show a greater tendency to misfold. GA-repeat strands exhibit strong folding cooperativity─folding becomes more favorable in the presence of other folded peptides. A peptide with GE repeats on one strand and GZ on the other yields assemblies with alternating negative and positive strands.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"32 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149151221","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Nonadditive Conformational\u0000Effects of tRNA Core Modifications","authors":"Teddy E. Amissah,Lev Levintov,Harish Vashisth","doi":"10.1021/acs.jcim.6c02045","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02045","url":null,"abstract":"Abstract Post-transcriptional modifications are known to regulate the structure and function of tRNA molecules. Compared to modifications in the anticodon region, the mechanistic effects of modifications in the core region of tRNA remain poorly understood. Using all-atom molecular dynamics (MD) simulations on yeast tRNAPhe, we investigated the conformational effects of incorporating chemical modifications in the core region (termed hereafter “core modifications”) of tRNA. Specifically, we focused on three core modifications, namely, 7-methylguanosine (m7G), 5-methylcytosine (m5C), and 1-methyladenosine (m1A). We discovered that individual modifications increased the structural stability of tRNA, but their combinations produced multimodal free energy landscapes with multiple distinct basins. We identified m1A58 as a bifunctional regulator that locally disrupts structure while globally coordinating anticorrelated arm movement. The combination of m1A58 and m5C49 exhibited the lowest structural deviation from the crystal structure, while the triply modified variant incorporating all three modifications produced the most integrated residue community network, reflecting nonadditive allosteric coupling that cannot be predicted from any pairwise combination. Critically, core modifications established extensive allosteric pathways that prevent the consistent inward displacement of the anticodon stem loop otherwise observed in all simulations of chemically unmodified tRNA. Our results demonstrate that core modifications function as regulators of conformational ensembles rather than simple stabilizers.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"32 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149151220","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"COMPLAX: A Python\u0000Tool Package for Automated Microsolvation\u0000of Organic Molecules","authors":"Federica Lauria,Andrea Maranzana","doi":"10.1021/acs.jcim.6c02105","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02105","url":null,"abstract":"Abstract Since the solvent choice drives organic reactivity, accurate computational modeling of explicit solute–solvent interactions remains crucial but preparation-intensive. We present COMPLAX, an open-source Python tool that automates the generation and optimization of targeted microsolvated clusters around highly localized active sites. Unlike unconstrained global optimizers, COMPLAX employs a rigid-core approach to preserve delicate solute geometries. To ensure unbiased spatial sampling, it couples a deterministic Fibonacci spherical lattice with three-dimensional rotational randomizations. Generated structures are refined via a two-stage xTB geometry optimization, utilizing a constrained pre-relaxation step to prevent unphysical solvent detachment. By automating conformer ranking and energy evaluations via a simple command-line interface, COMPLAX provides a reproducible, out-of-the-box solution for explicit microsolvation workflows.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"12 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149151222","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"State-Constrained SHAP Distills Polymer Property Models into Transferable Structural Rules for Fixed-Interval Screening.","authors":"Weihao Wang, Haoou Ruan, Keith T Butler, Akiko Kumada, Masahiro Sato","doi":"10.1021/acs.jcim.6c02120","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02120","url":null,"abstract":"<p><p>Polymer informatics has accelerated data-driven prediction of polymer properties, yet accurate prediction does not directly enable discovery. Here, we introduce state-constrained SHAP (SC-SHAP), an attribution-to-rule distillation framework for fixed-interval polymer screening. Using XGBoost as the teacher, SC-SHAP reorganizes SHAP and SHAP-interaction attributions on chemically interpretable MACCS fingerprints into support-qualified structural states defined by required-present and required-absent substructures. These states provide auditable property ordering, executable structural conditions, and candidate-level rule traces, while the learned library is reused across target windows with interpolation-only boundary calibration. Evaluation on experimental PoLyInfo datasets for glass-transition temperature (Tg), tensile stress at break (TSS), and electrical conductivity (EC) yields Spearman correlations between state scores and mean measured properties of 0.908 ± 0.008, 0.817 ± 0.069, and 0.835 ± 0.025, respectively. Benchmarks include direct XGBoost, MACCS-GCN, MACCS-Transformer, and rule-based screening. In the fixed-interval screening benchmark, SC-SHAP achieves higher mean scores than direct XGBoost, MACCS-GCN, MACCS-Transformer, and WRAcc on the smaller TSS and EC datasets. Multiple seeds and alternative structural splits further assess transfer to structurally distinct polymers. By unifying nonlinear knowledge distillation, chemically explicit rules, and cross-window reuse, SC-SHAP provides an interpretable screening framework spanning thermal, mechanical, and electrical polymer properties.</p>","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":" ","pages":""},"PeriodicalIF":6.4,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149147460","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"FragDockRL: A Reinforcement\u0000Learning Method for Fragment-Based\u0000Ligand Design via Building-Block Assembly and Tethered Docking","authors":"Seung Hwan Hong,Hyunsoo Kim,SeJin Kim,Soosung Kang","doi":"10.1021/acs.jcim.6c02851","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02851","url":null,"abstract":"Abstract Efficient exploration of combinatorial chemical space under synthetic constraints remains a central challenge in computational molecular design. Here, we present FragDock, a molecular design framework that combines building-block (BB)-based virtual synthesis with tethered docking guided by a predefined core structure. FragDock defines a structured search space by assembling molecules from synthetically accessible BBs through known chemical reactions and evaluating candidates using tethered docking with a restrained core-binding pose. Within this framework, we introduce FragDockRL, a reinforcement-learning-based search method that uses rewards derived from rDock-based tethered docking followed by SMINA scoring and a modified Deep Q-Network (DQN) to guide stepwise molecular growth. We evaluated FragDockRL on three therapeutically important protein targets─the receptor tyrosine kinases CSF1R and VEGFR2 and the coagulation protease FA10─selected to represent distinct biological functions and target classes. We assessed its performance using training-cycle analysis and benchmark comparisons with One-Step Reaction, Random Search, Beam Search, and Monte Carlo Tree Search. FragDockRL progressively enriched molecules with favorable docking scores during learning and generated more cutoff-passing unique molecules than Random Search across all three targets, supporting the benefit of learning-guided prioritization. However, the best-performing search strategy was target-dependent: One-Step Reaction, FragDockRL, and Beam Search each showed advantages in different cases. Representative molecular case studies showed that selected compounds retained reference-like binding poses while introducing structural variation into peripheral regions. The reaction schemes used commercially available BBs and well-established medicinal chemistry transformations, which supported the synthetic plausibility of the selected compounds. Overall, FragDock provides a flexible framework for synthetically constrained, structure-guided molecular exploration, and FragDockRL offers a learning-guided search mode for productive candidate prioritization under limited-generation budgets.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"114 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149151224","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Temporally Validated\u0000Models for Target “Tclin-likeness”","authors":"Mohammed Quazi,Suman Sirimulla,Cristian G. Bologa,Alexei Pushechnikov,Bill Farley,Nikolay Savchuk,Tudor I. Oprea","doi":"10.1021/acs.jcim.6c03088","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c03088","url":null,"abstract":"Abstract Identifying which human proteins are likely to become validated drug targets remains a challenge in drug discovery. We build on the “Tclin” concept, which encompasses proteins for which at least one approved drug exerts its mode of action (MoA); these comprise only ∼3% of the human proteome. We developed disease-agnostic, knowledge-graph machine learning (KGML) models that estimate “Tclin-likeness”: how closely a non-Tclin protein resembles established MoA drug targets in knowledge-graph feature space. Single-class support vector machines (1SVM) and XGBoost classifiers were trained on Pharos data current as of December 2017, using static (Gene Ontology, Reactome, the Kyoto encyclopedia for genes and genomes, STRING, HPA) and dynamic (genotype-tissue expression, LINCS, DISEASES, cancer cell line encyclopedia) features. To prevent data leakage, 115 proteins that acquired Tclin status after January 2018 were withheld entirely from training. Sensitivity on this external set was 88%–95% across individual models. Because non-Tclin proteins are unlabeled rather than confirmed negatives, specificity and precision are reported as bounds: balanced accuracy is 54%–75% for individual models and 73.7% for the eight-model consensus, which flags 5019 non-Tclin proteins (27.4%). The 1SVM models also identified a set of 1290 consensus negatives for XGBoost training. Benchmarked against two KGML druggability models, DrugnomeAI and PINNED, a consensus of 1133 proteins emerged. The Tclin-like consensus recovered 145 of 181 non-Tclin proteins in Phase 1–3 clinical trials (80.1%) and 23 of 28 proteins that reached Tclin status in 2023–2025 (82.1%), each approximately 3-fold enriched over background; it is statistically indistinguishable from PINNED (McNemar p = 0.42) while flagging 40% fewer proteins and outperforms DrugnomeAI (p = 0.0015). The three-model consensus places the druggable genome between 1860 and 5745 genes. Tclin-likeness is a prior, not a proof of tractability: these features encode nothing about binding pockets, ligandability or therapeutic modality. The “Tclin-like” consensus score is best used as an independent filter alongside disease-specific evidence and modality-specific tractability assessment.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"21 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149091815","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Unified Multitask\u0000Model for Aquatic Toxicity Prediction\u0000across Organism Groups and Effect End Points: A Comparison of Molecular\u0000Representations","authors":"Xiushan Wu,Baochuan Hu,Dongliang Chen,Jinrong Yang,Xiao He","doi":"10.1021/acs.jcim.6c02487","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02487","url":null,"abstract":"Abstract Computational prediction of aquatic toxicity provides a rapid and cost-efficient alternative to resource-intensive in vivo ecotoxicological testing. However, most existing models adopt a fragmented modeling strategy, developing separate models for each organism group and toxicity end point. Such task-specific models restrict knowledge transfer across related biological contexts and fail to account for the biological ordering between multiple effect concentrations (EC50/EC10). Here, we propose a unified multitask framework to simultaneously predict effect concentrations for three aquatic organism groups (fish, crustaceans, algae) and two end points (EC50/EC10) within a single model. This framework integrates a molecular encoder with a Multigate Mixture-of-Experts (MMoE) network, a soft end point-ordering term that encourages the biological relation EC50 ≥ EC10, and a heteroscedastic regression output that provides per-prediction uncertainty estimates. We evaluated a multirepresentation fusion encoder combining physicochemical descriptors, molecular fingerprints, and a 2D molecular graph neural network (GNN), together with 3D-pretrained Uni-Mol and TRIDENT baselines. Under a common global scaffold/acyclic-cluster 5-fold pooled out-of-fold (OOF) protocol, the multirepresentation fusion model demonstrated robust predictive performance across all six organism–end point tasks, with lower prediction errors than TRIDENT across all six tasks and Uni-Mol in four tasks. Attention-based interpretation further highlighted chemically meaningful atoms and substructures that were consistent with plausible toxicity-relevant structural motifs. Together, these findings demonstrate that integrating multitask learning, biologically informed end point constraints, and complementary molecular representations provides a promising foundation for broadly applicable and uncertainty-aware models of aquatic toxicity.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"22 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149091813","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Causal-Chemprop:\u0000Causal Machine Learning for Molecular\u0000Property Prediction and Design","authors":"Christian Natajaya,Lucas Attia,Jackson W. Burns,Patrick S. Doyle","doi":"10.1021/acs.jcim.6c01429","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c01429","url":null,"abstract":"Abstract A priori estimation of molecular properties has long been a topic of immense interest to the pharmaceutical sciences for molecular design. While neural network-based models have achieved high predictive accuracy, they have limited utility in molecular design, particularly in small and out-of-distribution data regimes. Current neural network-based models lack mechanisms to explicitly incorporate prior knowledge from existing data. Herein, we introduce a causal machine learning framework built on the Chemprop and DAGMA architectures for molecular property prediction called Causal-Chemprop. Its capacity for counterfactual reasoning supports human-in-the-loop optimization of molecular structure, which we demonstrate by improving the rank-ordering of out-of-distribution kinase inhibitor pIC50 values─most where standard encoders extrapolate poorly─both within and across scaffold series, in a lead-optimization setting in which the test scaffold is known. We further integrate Causal-Chemprop with the molecular optimization algorithm EvoMol as a scoring function, where it tends to yield more synthesizable molecules that more closely resemble the sought-after active compound than Chemprop. To our knowledge, this is the first application bridging neural network-learned molecular representations with structural causal models for molecular property prediction and optimization via counterfactual reasoning, suggesting several promising new avenues for exploration in molecular property prediction and design.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"21 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149091816","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SAEKP: A Site Attention\u0000Enzyme Kinetics Prediction\u0000Method to Facilitate Enzyme Engineering","authors":"Jia-He Qiu,Zongying Lin,Ke-Wei Chen,Qiang Wang,Tian-Yu Sun,Li Yuan,Xian Zhang,Yonghong Tian,Yun-Dong Wu","doi":"10.1021/acs.jcim.6c01516","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c01516","url":null,"abstract":"Abstract While the determination of enzyme kinetic parameters traditionally relies on prolonged and costly experimental methods, machine learning models have recently demonstrated promise as predictive tools. These models still have limitations, such as limited training data and insufficient sensitivity to mutations at important sites. Here, we introduce SAEKP (Site Attention Enzyme Kinetics Prediction Method), which incorporates representations of enzyme sequence, substrate, and an additional weighted sequence representation that highlights functionally important sites. Trained on wild-type and mutant enzyme data from BRENDA and SABIO-RK, SAEKP outperforms existing models in predicting enzyme kinetic parameters, including turnover number (kcat), Michaelis constant (Km), and inhibition constant (Ki). The selected multimodal protein and substrate representation framework provides strong overall performance, while the site attention module further improves performance in predicting mutations at or near important sites through prior-knowledge-guided residue reweighting. SAEKP offers a robust and practical tool for predicting enzyme kinetic parameters, facilitating enzyme engineering and subsequent industrial applications.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"82 1 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149151223","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}