A new era of antibody discovery: an in-depth review of AI-driven approaches.

Given their high affinity and specificity for a range of macromolecules, antibodies are widely used in the treatment of autoimmune diseases, cancers, inflammatory diseases, and Alzheimer's disease (AD). Traditional experimental methods are time-consuming, expensive, and labor-intensive. Recent advances in artificial intelligence (AI) technologies provide complementary methods that can reduce the time and costs required for antibody design by minimizing failures and increasing the success rate of experimental tests. In this review, we scrutinize the plethora of AI-driven methodologies that have been deployed over the past 4 years for modeling antibody structures, predicting antibody-antigen interactions, optimizing antibody affinity, and generating novel antibody candidates. We also briefly address the challenges faced in integrating AI-based models with traditional antibody discovery pipelines and highlight the potential future directions in this burgeoning field.

Metabolism-guided development of Ko143 analogs as ABCG2 inhibitors.

ATP-binding cassette subfamily G member 2 (ABCG2), an efflux transporter, is involved in multiple pathological processes. Ko143 is a potent ABCG2 inhibitor; however, it is quickly metabolized through carboxylesterase 1-mediated hydrolysis of its t-butyl ester moiety. The current work aimed to develop more metabolically stable ABCG2 inhibitors. Novel Ko143 analogs were designed and synthesized by replacing the unstable t-butyl ester moiety in Ko143 with an amide group. The synthesized Ko143 analogs were evaluated for their ABCG2 inhibitory activity, binding mode with ABCG2, cytotoxicity, and metabolic stability. We found that the amide modification of Ko143 led to metabolically stable ABCG2 inhibitors. Among these Ko143 analogs, K2 and K34 are promising candidates with favorable oral pharmacokinetic profiles in mice. In summary, we synthesized novel Ko143 analogs with improved metabolic stability, which can potentially be used as lead compounds for the future development of ABCG2 inhibitors.

Targeting the endocannabinoid system: Structural determinants and molecular mechanism of allosteric modulation.

Although drugs targeting the orthosteric binding site of cannabinoid receptors (CBRs) have several therapeutic effects on human physiological and pathological conditions, they can also cause serious adverse effects. Only a few orthosteric ligands have successfully passed clinical trials. Recently, allosteric modulation has become a novel option for drug discovery, with fewer adverse effects and the potential to avoid drug overdose. In this review, we highlight novel findings related to the drug discovery of allosteric modulators (AMs) targeting CBRs. We summarize newly synthesized AMs and the reported/predicted allosteric binding sites. We also discuss the structural determinants of the AMs binding as well as the molecular mechanism of CBR allostery.

Unpolarized laser method for infrared spectrum calculation of amide I CO bonds in proteins using molecular dynamics simulation.

The investigation of the strong infrared (IR)-active amide I modes of peptides and proteins has received considerable attention because a wealth of detailed information on hydrogen bonding, dipole-dipole interactions, and the conformations of the peptide backbone can be derived from the amide I bands. The interpretation of experimental spectra typically requires substantial theoretical support, such as direct ab-initio molecular dynamics simulation or mixed quantum-classical description. However, considering the difficulties associated with these theoretical methods and their applications are limited in small peptides, it is highly desirable to develop a simple yet efficient approach for simulating the amide I modes of any large proteins in solution. In this work, we proposed a comprehensive computational method that extends the well-established molecular dynamics (MD) simulation method to include an unpolarized IR laser for exciting the CO bonds of proteins. We showed the amide I frequency corresponding to the frequency of the laser pulse which resonated with the CO bond vibration. At this frequency, the protein energy and the CO bond length fluctuation were maximized. Overall, the amide I bands of various single proteins and amyloids agreed well with experimental data. The method has been implemented into the AMBER simulation package, making it widely available to the scientific community. Additionally, the application of the method to simulate the transient amide I bands of amyloid fibrils during the IR laser-induced disassembly process was discussed in details.

Molecular Modeling Study of a Receptor-Orthosteric Ligand-Allosteric Modulator Signaling Complex.

Allosteric modulators (AMs) are considered as a perpetual hotspot in research for their higher selectivity and various effects on orthosteric ligands (OL). They are classified in terms of their functionalities as positive, negative, or silent allosteric modulators (PAM, NAM, or SAM, respectively). In the present work, 11 pairs of three-dimensional (3D) structures of receptor-orthosteric ligand and receptor-orthosteric ligand-allosteric modulator complexes have been collected for the studies, including three different systems: GPCR, enzyme, and ion channel. Molecular dynamics (MD) simulations are applied to quantify the dynamic interactions in both the orthosteric and allosteric binding pockets and the structural fluctuation of the involved proteins. Our results showed that MD simulations of moderately large molecules or peptides undergo insignificant changes compared to crystal structure results. Furthermore, we also studied the conformational changes of receptors that bound with PAM and NAM, as well as the different allosteric binding sites in a receptor. There should be no preference for the position of the allosteric binding pocket after comparing the allosteric binding pockets of these three systems. Finally, we aligned four distinct ß2 adrenoceptor structures and three N-methyl-d-aspartate receptor (NMDAR) structures to investigate conformational changes. In the ß2 adrenoceptor systems, the aligned results revealed that transmembrane (TM) helices 1, 5, and 6 gradually increased outward movement from an enhanced inactive state to an improved active state. TM6 endured the most significant conformational changes (around 11 Å). For NMDAR, the bottom section of NMDAR's ligand-binding domain (LBD) experienced an upward and outward shift during the gradually activating process. In conclusion, our research provides insight into receptor-orthosteric ligand-allosteric modulator studies and the design and development of allosteric modulator drugs using MD simulation.

Discovery of a pyrano[2,3-b]pyridine derivative YX-2102 as a cannabinoid receptor 2 agonist for alleviating lung fibrosis.

BACKGROUND: Pharmacological modulation of cannabinoid 2 receptor (CB2R) is a promising therapeutic strategy for pulmonary fibrosis (PF). Thus, to develop CB2R selective ligands with new chemical space has attracted much research interests. This work aims to discover a novel CB2R agonist from an in-house library, and to evaluate its therapeutic effects on PF model, as well as to disclose the pharmacological mechanism. METHODS: Virtual screening was used to identify the candidate ligand for CB2R from a newly established in-house library. Both in vivo experiments on PF rat model and in vitro experiments on cells were performed to investigate the therapeutic effects of the lead compound and underlying mechanism. RESULTS: A "natural product-like" pyrano[2,3-b]pyridine derivative, YX-2102 was identified that bound to CB2R with high affinity. Intraperitoneal YX-2102 injections significantly ameliorated lung injury, inflammation and fibrosis in a rat model of PF induced by bleomycin (BLM). On one hand, YX-2102 inhibited inflammatory response at least partially through modulating macrophages polarization thereby exerting protective effects. Whereas, on the other hand, YX-2102 significantly upregulated CB2R expression in alveolar epithelial cells in vivo. Its pretreatment inhibited lung alveolar epithelial-to-mesenchymal transition (EMT) in vitro and PF model induced by transforming growth factor beta-1 (TGF-ß1) via a CB2 receptor-dependent pathway. Further studies suggested that the Nrf2-Smad7 pathway might be involved in. CONCLUSION: These findings suggest that CB2R is a potential target for PF treatment and YX-2102 is a promising CB2R agonist with new chemical space.

Virus-CKB 2.0: Viral-Associated Disease-Specific Chemogenomics Knowledgebase.

Transmissible and infectious viruses can cause large-scale epidemics around the world. This is because the virus can constantly mutate and produce different variants and subvariants to counter existing treatments. Therefore, a variety of treatments are urgently needed to keep up with the mutation of the viruses. To facilitate the research of such treatment, we updated our Virus-CKB 1.0 to Virus-CKB 2.0, which contains 10 kinds of viruses, including enterovirus, dengue virus, hepatitis C virus, Zika virus, herpes simplex virus, Andes orthohantavirus, human immunodeficiency virus, Ebola virus, Lassa virus, influenza virus, coronavirus, and norovirus. To date, Virus-CKB 2.0 archived at least 65 antiviral drugs (such as remdesivir, telaprevir, acyclovir, boceprevir, and nelfinavir) in the market, 178 viral-related targets with 292 available 3D crystal or cryo-EM structures, and 3766 chemical agents reported for these target proteins. Virus-CKB 2.0 is integrated with established tools for target prediction and result visualization; these include HTDocking, TargetHunter, blood-brain barrier (BBB) predictor, Spider Plot, etc. The Virus-CKB 2.0 server is accessible at https://www.cbligand.org/g/virus-ckb. By using the established chemogenomic tools and algorithms and newly developed tools, we can screen FDA-approved drugs and chemical compounds that may bind to these proteins involved in viral-associated disease regulation. If the virus strain mutates and the vaccine loses its effect, we can still screen drugs that can be used to treat the mutated virus in a fleeting time. In some cases, we can even repurpose FDA-approved drugs through Virus-CKB 2.0.

Differential performance of RoseTTAFold in antibody modeling.

Antibodies are essential to life, and knowing their structures can facilitate the understanding of antibody-antigen recognition mechanisms. Precise antibody structure prediction has been a core challenge for a prolonged period, especially the accuracy of H3 loop prediction. Despite recent progress, existing methods cannot achieve atomic accuracy, especially when the homologous structures required for these methods are not available. Recently, RoseTTAFold, a deep learning-based algorithm, has shown remarkable breakthroughs in predicting the 3D structures of proteins. To assess the antibody modeling ability of RoseTTAFold, we first retrieved the sequences of 30 antibodies as the test set and used RoseTTAFold to model their 3D structures. We then compared the models constructed by RoseTTAFold with those of SWISS-MODEL in a different way, in which we stratified Global Model Quality Estimate (GMQE) into three different ranges. The results indicated that RoseTTAFold could achieve results similar to SWISS-MODEL in modeling most CDR loops, especially the templates with a GMQE score under 0.8. In addition, we also compared the structures modeled by RoseTTAFold, SWISS-MODEL and ABodyBuilder. In brief, RoseTTAFold could accurately predict 3D structures of antibodies, but its accuracy was not as good as the other two methods. However, RoseTTAFold exhibited better accuracy for modeling H3 loop than ABodyBuilder and was comparable to SWISS-MODEL. Finally, we discussed the limitations and potential improvements of the current RoseTTAFold, which may help to further the accuracy of RoseTTAFold's antibody modeling.

How Do Modulators Affect the Orthosteric and Allosteric Binding Pockets?

Allosteric modulators (AMs) that bind allosteric sites can exhibit greater selectivity than the orthosteric ligands and can either enhance agonist-induced receptor activity (termed positive allosteric modulator or PAM), inhibit agonist-induced activity (negative AM or NAM), or have no effect on activity (silent AM or SAM). Until now, it is not clear what the exact effects of AMs are on the orthosteric active site or the allosteric binding pocket(s). In the present work, we collected both the three-dimensional (3D) structures of receptor-orthosteric ligand and receptor-orthosteric ligand-AM complexes of a specific target protein. Using our novel algorithm toolset, molecular complex characterizing system (MCCS), we were able to quantify the key residues in both the orthosteric and allosteric binding sites along with potential changes of the binding pockets. After analyzing 21 pairs of 3D crystal or cryo-electron microscopy (cryo-EM) complexes, including 4 pairs of GPCRs, 5 pairs of ion channels, 11 pairs of enzymes, and 1 pair of transcription factors, we found that the binding of AMs had little impact on both the orthosteric and allosteric binding pockets. In return, given the accurately predicted allosteric binding pocket(s) of a drug target of medicinal interest, we can confidently conduct the virtual screening or lead optimization without concern that the huge conformational change of the pocket could lead to the low accuracy of virtual screening.

Artificial Intelligent Deep Learning Molecular Generative Modeling of Scaffold-Focused and Cannabinoid CB2 Target-Specific Small-Molecule Sublibraries.

Design and generation of high-quality target- and scaffold-specific small molecules is an important strategy for the discovery of unique and potent bioactive drug molecules. To achieve this goal, authors have developed the deep-learning molecule generation model (DeepMGM) and applied it for the de novo molecular generation of scaffold-focused small-molecule libraries. In this study, a recurrent neural network (RNN) using long short-term memory (LSTM) units was trained with drug-like molecules to result in a general model (g-DeepMGM). Sampling practices on indole and purine scaffolds illustrate the feasibility of creating scaffold-focused chemical libraries based on machine intelligence. Subsequently, a target-specific model (t-DeepMGM) for cannabinoid receptor 2 (CB2) was constructed following the transfer learning process of known CB2 ligands. Sampling outcomes can present similar properties to the reported active molecules. Finally, a discriminator was trained and attached to the DeepMGM to result in an in silico molecular design-test circle. Medicinal chemistry synthesis and biological validation was performed to further investigate the generation outcome, showing that XIE9137 was identified as a potential allosteric modulator of CB2. This study demonstrates how recent progress in deep learning intelligence can benefit drug discovery, especially in de novo molecular design and chemical library generation.

In Silico Prediction and Validation of CB2 Allosteric Binding Sites to Aid the Design of Allosteric Modulators.

Although the 3D structures of active and inactive cannabinoid receptors type 2 (CB2) are available, neither the X-ray crystal nor the cryo-EM structure of CB2-orthosteric ligand-modulator has been resolved, prohibiting the drug discovery and development of CB2 allosteric modulators (AMs). In the present work, we mainly focused on investigating the potential allosteric binding site(s) of CB2. We applied different algorithms or tools to predict the potential allosteric binding sites of CB2 with the existing agonists. Seven potential allosteric sites can be observed for either CB2-CP55940 or CB2-WIN 55,212-2 complex, among which sites B, C, G and K are supported by the reported 3D structures of Class A GPCRs coupled with AMs. Applying our novel algorithm toolset-MCCS, we docked three known AMs of CB2 including Ec2la (C-2), trans-ß-caryophyllene (TBC) and cannabidiol (CBD) to each site for further comparisons and quantified the potential binding residues in each allosteric binding site. Sequentially, we selected the most promising binding pose of C-2 in five allosteric sites to conduct the molecular dynamics (MD) simulations. Based on the results of docking studies and MD simulations, we suggest that site H is the most promising allosteric binding site. We plan to conduct bio-assay validations in the future.

Sequestsome-1/p62-targeted small molecules for pancreatic cancer therapy.

Pancreatic ductal adenocarcinoma (PDAC) is characterized by heightened autophagy and systemic immune dysfunction. Modest improvements in clinical outcomes have been demonstrated in completed clinical trials targeting autophagy with combination hydroxychloroquine (HCQ) and chemotherapy. Recent mechanistic insights into the role of autophagy-dependent immune evasion have prompted the need for more precise and druggable targets of autophagy inhibition. Sequestosome-1 (SQSTM-1) is a multidomain scaffold protein with well-established roles in autophagy, tumor necrosis factor alpha (TNFα)- and NF-κB-related signaling pathways. SQSTM1 overexpression is frequently observed in PDAC, correlating with clinical stage and outcome. Given the unique molecular structure of SQSTM-1 and its diverse activity, identifying means of limiting SQSTM-1-dependent autophagy to promote an effective immune response in PDAC could be a promising treatment strategy.

Enhanced self-renewal of human long-term hematopoietic stem cells by a sulfamoyl benzoate derivative targeting p18INK4C.

The use of umbilical cord blood transplant has been substantially limited by the finite number of hematopoietic stem and progenitor cells in a single umbilical cord blood unit. Small molecules that not only quantitatively but also qualitatively stimulate enhancement of hematopoietic stem cell (HSC) self-renewal ex vivo should facilitate the clinical use of HSC transplantation and gene therapy. Recent evidence has suggested that the cyclin-dependent kinase inhibitor, p18INK4C (p18), is a critical regulator of mice HSC self-renewal. The role of p18 in human HSCs and the effect of p18 inhibitor on human HSC expansion ex vivo need further studies. Here we report that knockdown of p18 allowed for an increase in long-term colony-forming cells in vitro. We then identified an optimized small molecule inhibitor of p18, 005A, to induce ex vivo expansion of HSCs that was capable of reconstituting human hematopoiesis for at least 4 months in immunocompromised mice, and hence, similarly reconstituted secondary recipients for at least 4 more months, indicating that cells exposed to 005A were still competent in secondary recipients. Mechanistic studies showed that 005A might delay cell division and activate both the Notch signaling pathway and expression of transcription factor HoxB4, leading to enhancement of the self-renewal of long-term engrafting HSCs and the pool of progenitor cells. Taken together, these observations support a role for p18 in human HSC maintenance and that the p18 inhibitor 005A can enhance the self-renewal of long-term HSCs.

Integrated Multi-Class Classification and Prediction of GPCR Allosteric Modulators by Machine Learning Intelligence.

G-protein-coupled receptors (GPCRs) are the largest and most diverse group of cell surface receptors that respond to various extracellular signals. The allosteric modulation of GPCRs has emerged in recent years as a promising approach for developing target-selective therapies. Moreover, the discovery of new GPCR allosteric modulators can greatly benefit the further understanding of GPCR cell signaling mechanisms. It is critical but also challenging to make an accurate distinction of modulators for different GPCR groups in an efficient and effective manner. In this study, we focus on an 11-class classification task with 10 GPCR subtype classes and a random compounds class. We used a dataset containing 34,434 compounds with allosteric modulators collected from classical GPCR families A, B, and C, as well as random drug-like compounds. Six types of machine learning models, including support vector machine, naïve Bayes, decision tree, random forest, logistic regression, and multilayer perceptron, were trained using different combinations of features including molecular descriptors, Atom-pair fingerprints, MACCS fingerprints, and ECFP6 fingerprints. The performances of trained machine learning models with different feature combinations were closely investigated and discussed. To the best of our knowledge, this is the first work on the multi-class classification of GPCR allosteric modulators. We believe that the classification models developed in this study can be used as simple and accurate tools for the discovery and development of GPCR allosteric modulators.

Structural and in Vitro Functional Characterization of a Menthyl TRPM8 Antagonist Indicates Species-Dependent Regulation.

TRPM8 antagonists derived from its cognate ligand, (-)-menthol, are underrepresented. We determine the absolute stereochemistry of a well-known TRPM8 antagonist, (-)-menthyl 1, using VCD and 2D NMR. We explore 1 for its antagonist effects of the human TRPM8 (hTRPM8) orthologue to uncover species-dependent inhibition versus rat channels. (-)-Menthyl 1 inhibits menthol- and icilin-evoked Ca2+ responses at hTRPM8 with IC50 values of 805 ± 200 nM and 1.8 ± 0.6 µM, respectively, while more potently inhibiting agonist responses at the rat orthologue (rTRPM8 IC50 (menthol) = 117 ± 18 nM, IC50 (icilin) = 521 ± 20 nM). Whole-cell patch-clamp recordings of hTRPM8 confirm the 1 inhibition of menthol-stimulated currents, with an IC50 of 700 ± 200 nM. We demonstrate that 1 possesses ≥400-fold selectivity for hTRPM8 versus hTRPA1/hTRPV1. (-)-menthyl 1 can be used as a novel chemical tool to study hTRPM8 pharmacology and differences in species commonly used in drug discovery.

Drug-Drug Interaction Between Oxycodone and Diazepam by a Combined in Silico Pharmacokinetic and Pharmacodynamic Modeling Approach.

Opioids and benzodiazepines have complex drug-drug interactions (DDIs), which serve as an important source of adverse drug effects. In this work, we predicted the DDI between oxycodone (OXY) and diazepam (DZP) in the human body by applying in silico pharmacokinetic (PK) and pharmacodynamic (PD) modeling and simulation. First, we studied the PK interaction between OXY and DZP with a physiologically based pharmacokinetic (PBPK) model. Second, we applied molecular modeling techniques including molecular docking, molecular dynamics (MD) simulation, and the molecular mechanics/Poisson-Boltzmann surface area (MM-PBSA) free energy method to predict the PD-DDI between these two drugs. The PK interaction between OXY and DZP predicted by the PBPK model was not obvious. No significant interaction was observed between the two drugs at normal doses, though very high doses of DZP demonstrated a non-negligible inhibitory effect on OXY metabolism. On the contrary, the molecular modeling study shows that DZP has potential to compete with OXY at the same binding pocket of the active µ-opioid receptor (MOR) and κ-opioid receptor (KOR). MD simulation and MM-PBSA calculation results demonstrated that there is likely a synergetic effect between OXY and DZP binding to opioid receptors, as OXY is likely to target the active MOR while DZP selectively binds to the active KOR. Thus, pharmacokinetics contributes slightly to the DDI between OXY and DZP although an overdose of DZP has been brought to attention. Pharmacodynamics is likely to play a more important role than pharmacokinetics in revealing the mechanism of DDI between OXY and DZP.

IsAb: a computational protocol for antibody design.

The design of therapeutic antibodies has attracted a large amount of attention over the years. Antibodies are widely used to treat many diseases due to their high efficiency and low risk of adverse events. However, the experimental methods of antibody design are time-consuming and expensive. Although computational antibody design techniques have had significant advances in the past years, there are still some challenges that need to be solved, such as the flexibility of antigen structure, the lack of antibody structural data and the absence of standard antibody design protocol. In the present work, we elaborated on an in silico antibody design protocol for users to easily perform computer-aided antibody design. First, the Rosetta web server will be applied to generate the 3D structure of query antibodies if there is no structural information available. Then, two-step docking will be used to identify the binding pose of an antibody-antigen complex when the binding information is unknown. ClusPro is the first method to be used to conduct the global docking, and SnugDock is applied for the local docking. Sequentially, based on the predicted binding poses, in silico alanine scanning will be used to predict the potential hotspots (or key residues). Finally, computational affinity maturation protocol will be used to modify the structure of antibodies to theoretically increase their affinity and stability, which will be further validated by the bioassays in the future. As a proof of concept, we redesigned antibody D44.1 and compared it with previously reported data in order to validate IsAb protocol. To further illustrate our proposed protocol, we used cemiplimab antibody, a PD-1 checkpoint inhibitor, as an example to showcase a step-by-step tutorial.

Binding Characterization of Agonists and Antagonists by MCCS: A Case Study from Adenosine A2A Receptor.

Characterizing the structural basis of ligand recognition of adenosine A2A receptor (AA2AR) will facilitate its rational design and development of small molecules with high affinity and selectivity, as well as optimal therapeutic effects for pain, cancers, drug abuse disorders, etc. In the present work, we applied our reported algorithm, molecular complex characterizing system (MCCS), to characterize the binding features of AA2AR based on its reported 3D structures of protein-ligand complexes. First, we compared the binding score to the reported experimental binding affinities of each compound. Then, we calculated an output example of residue energy contribution using MCCS and compared the results with data obtained from MM/GBSA. The consistency in results indicated that MCCS is a powerful, fast, and accurate method. Sequentially, using a receptor-ligand data set of 57 crystallized structures of AA2ARs, we characterized the binding features of the binding pockets in AA2AR, summarized the key residues that distinguish antagonist from agonist, produced heatmaps of residue energy contribution for clustering various statuses of AA2ARs, explored the selectivity between AA2AR and AA1AR, etc. All the information provided new insights into the protein features of AA2AR and will facilitate its rational drug design.

Generative chemistry: drug discovery with deep learning generative models.

The de novo design of molecular structures using deep learning generative models introduces an encouraging solution to drug discovery in the face of the continuously increased cost of new drug development. From the generation of original texts, images, and videos, to the scratching of novel molecular structures the creativity of deep learning generative models exhibits the height machine intelligence can achieve. The purpose of this paper is to review the latest advances in generative chemistry which relies on generative modeling to expedite the drug discovery process. This review starts with a brief history of artificial intelligence in drug discovery to outline this emerging paradigm. Commonly used chemical databases, molecular representations, and tools in cheminformatics and machine learning are covered as the infrastructure for generative chemistry. The detailed discussions on utilizing cutting-edge generative architectures, including recurrent neural network, variational autoencoder, adversarial autoencoder, and generative adversarial network for compound generation are focused. Challenges and future perspectives follow.

Determination of van der Waals Parameters Using a Double Exponential Potential for Nonbonded Divalent Metal Cations in TIP3P Solvent.

A double exponential (DE) functional form for Lennard-Jones (LJ) interactions, proposed in our previous study, has many advantages over LJ potentials including a natural softcore characteristic for the convenience of the pathway-based free-energy calculations, fast convergence, and flexibility in use. In this work, we put the first step on the application of the DE functional form by identifying a DE potential, coined DE-TIP3P, for molecular simulations using the TIP3P water model. The developed DE-TIP3 potential was better than LJ potential in reproducing the experimental water properties. Afterward, we developed the nonbonded models of 15 divalent metal ions, which frequently appear and play vital roles in biological systems, to be consistent with the DE-TIP3P potential and TIP3P water model. Our nonbonded models were as good as the complicated nonbonded dummy cationic models by Jiang et al. and the nonbonded 12-6-4 LJ models by Li and Merz in reproducing the experimental properties of those ions. Moreover, our nonbonded models achieved a better performance than the compromise (CM) LJ models and 12-6-4 LJ models, developed by Li and Merz, in reproducing the properties of MgCl2 in aqueous solution.