US 20040204862 A1
A computer readable medium holding data of a molecular model of a ligand-gated ion channel receptor and/or a computer system for modeling said receptor are provided by the instant invention. The molecular model can be used to design novel compounds having activity as non-competitive inhibitors of the ion channel. A preferred embodiment of the invention relates to nicotinic acetylcholine receptors. Compounds having activity as non-competitive inhibitors of ligand-gated ion channel receptors and methods for inhibiting the receptor and treating diseases or disorders mediated by function of the receptor are also disclosed.
1. A computer system comprising:
i) a memory storing positional data of the atomic coordinates of the transmembrane portion of at least one subunit of a ligand-gated neurotransmitter receptor protein; and
ii) a processor generating a molecular model having a three dimensional shape representative of the pore portion of the ligand-gated neurotransmitter receptor based the positional data.
2. The computer system of
3. The computer system of
4. The computer system of
5. The computer system of
6. The computer system of
7. The computer system of
8. A method for screening a compound for activity as a non-competitive inhibitor of a ligand-gated neurotransmitter receptor comprising; identifying as a compound having activity as a non-competitive inhibitor as one having a ΔG less than −6 kcal/mol by docking a model of the compound in a model of the three-dimensional shape of the pore portion of the ligand-gated neurotransmitter receptor.
9. The method of
10. The method of
11. The method of
12. A method for making a non-competitive inhibitor of a ligand-gated neurotransmitter receptor comprising:
A method for screening a compound for activity as a non-competitive inhibitor of a ligand-gated neurotransmitter receptor comprising:
i) identifying as a compound having activity as a non-competitive inhibitor as one having a ΔG less than −6 kcal/mol by docking a model of the compound in a model of the three-dimensional shape of the pore portion of the ligand-gated neurotransmitter receptor;
ii) testing the compound for binding to the ligand-gated neurotransmitter receptor and/or inhibiting the ion-channel activity thereof;
iii) obtaining as a non-competitive inhibitor of the ligand-gated neurotransmitter receptor the compound that binds to the ligand-gated neurotransmitter receptor with a k′ value greater than 8 and being displaced in chromatographic experiments by mecamylamine and/or inhibits the ion channel activity of the ligand-gated neurotransmitter receptor in nicotine stimulated 86Rb+ efflux with IC50 lower than 100 μM.
13. The method of
14. The method of
15. A compound comprising: a bulky hydrophobic moiety and primary, secondary or tertiary amino group located 5 to 10 Å from said hydrophobic moiety, said compound having activity of inhibiting the ion-channel activity of a ligand-gated neurotransmitter receptor in an assay of nicotine stimulated 86Rb+ efflux with an IC50 lower than 50 μM.
16. The compound of
17. The compound of
18. The compound of
19. The compound of
20. The compound of
21. A compound comprising a bulky hydrophobic moiety and a primary, secondary or tertiary amino group placed 5 to 10 Å from said hydrophobic moiety, said compound having activity of binding to the non-competitive inhibitor site of a ligand-gated neurotransmitter receptor with a ΔG of at least −6 kcal/mol.
22. The compound of
23. The compound of
24. The compound of
25. A kit comprising a computer readable medium having stored therein positional data of the atomic coordinates of the transmembrane portion of at least one subunit of a ligand-gated neurotransmitter receptor protein; and
a composition comprising the ligand-gated neurotransmitter receptor protein.
26. The kit of
27. The kit of
28. A computer readable medium having stored therein positional data of the atomic coordinates of the transmembrane portion of at least one subunit of a ligand-gated neurotransmitter receptor protein.
29. The computer readable medium of
30. The computer readable medium of
31. The computer readable medium of
32. The computer readable medium of
33. A method for non-competitively inhibiting a ligand-gated ion channel receptor comprising contacting the ligand-gated ion channel receptor with a compound of
34. A method for treating Tourette's syndrome, or a cognitive disorder, pain, anxiety, depression, neurodegeneration or an addiction caused by an overactive ligand-gated ion channel receptor, comprising administering to a subject an amount of a compound of
35. The method of
36. A method for evaluating cardiovascular toxicity or GI spasming or diarrheal side effects of a compound comprising testing a compound for activity in the method of
 The present application includes an appended Sequence Listing of 15 amino acid sequences and Appendices 1 to 3 providing computer programming scripts and parameter files.
 The present invention relates to a computer system for generating molecular models of ligand-gated ion channels and in particular, molecular models of the inner lumen of a ligand-gated ion channel and associated binding pockets. The present invention further relates to a computer system simulating interaction of the computer-based model of the ligand-gated channel and non-competitive inhibitor compounds for identification and characterization of non-competitive inhibitors and to inhibitors compounds so discovered. The present invention also relates to methods for treating various disorders related to ligand-gated ion channel receptor function.
 Ligand-gated ion channels are currently a popular target for drug discovery in the pharmaceutical industry. The Ligand-Gated Ion Channel (LGIC) superfamily is separated into the nicotinic receptor superfamily (muscular and neuronal nicotinic, GABA-A and C, glycine and 5-HT3 receptors), the excitatory amino acid superfamily (glutamate, aspartate and kainate receptors) and the ATP purinergic ligand-gated ion channels. These families only differ in the number of transmembrane domains found in each subunit (nAChRs have 5 transmembrane domains), excitatory amino acid receptors have 4 transmembrane domains and ATP purinergic LGICs have 3 transmembrane domains).
 The nicotinic acetylcholine receptor (nAChR) is presently the best characterized member of the ligand-gated ion channel superfamily. The nicotinic receptors are of great therapeutic importance. The subunits assemble combinatorily to form a variety of pentameric transmembrane protein subtypes.
 Neuronal nicotinic acetylcholine receptors (nAChRs) are the class of ligand-gated ion channels of the central and peripheral nervous system that regulate synaptic activity. The basic structure of the nAChR is shown in FIG. 1. nAChR consists of five transmembrane subunits 1, 2, 3, 4, 5 oriented around a central pore 6 permeable to cations. Cations flow through the pore is regulated by ligand binding. The subunits in nAChR are typically a subunits and β subunits.
 At present, 12 different homologous subunits have been identified in neuronal nAChRs, 9 α subunits (α2-α10) and 3 β subunits (α2-α4). The major difference between α and β subunits is the presence and location of the disulfide bond formed by two adjacent cysteines in the α systems, the absence of this feature distinguishes non-α subunits. This disulfide bond located on the extracellular domain plays an important role in neurotransmitter binding as well as the mechanism of channel opening. These subunits combine to form multiple nAChR subtypes and predominant stoichiometry is (α)2(β)3, however pentamers containing only α subunit are also known e.g., (α7)5. In case of muscular nAChR the stoichiometry is more complicated, the muscular nAChR receptor is predominantly described as (α)2βδγ.
 The nAChRs are very complex systems with dozens of potential different binding domains for different classes of compounds of both endo- and exogenous origin (Arias H. R., (1997) Topology of ligand binding sites on the nicotinic acetylcholine receptor. Brain Res. Rev. 25: 133-91). Two primary cholinergic binding sites are located on the extracellular side 7 (approximately 30-35 Å above the membrane) in the pocket at the interface between the α and β subunits. The nAChR contains several other classes of binding sites at which non-competitive inhibitors (NCIs) bind (Arias H. R. (1998) Binding sites for exogenous and endogenous non-competitive inhibitors of the nicotinic acetylcholine receptor. Biochim. Biophys. Act. 1376: 173-220). One, so-called “luminal high affinity” NCI binding domain is located on the surface of the internal lumen forming the ion channel. This site is a highly polar and negatively charged domain, which primarily plays the role as a cation selector. In general, an NCI compound does not compete with the neurotransmitter ligand of the receptor for binding to the neurotransmitter ligand binding site of the receptor located on the external surface both α subunits in a pocket approximately 30-35 Å from the transmembrane portion of the subunit (that is, above the surface membrane when the receptor is expressed on in a cell), as described by Arias [Arias, H. R. (2000) Localization of agonist and competitive antagonist binding sites on nicotinic acetylcholine receptors Neurochem. Int 36, 595-645].
 Such drugs as mecamylamine, ketamine, bupropion or barbiturates bind in the narrowest region of the channel on the cell membrane level. Inhibitors acting there are mainly amines. It is believed that the ligands, bind into this region and sterically plug the channel, blocking the flux of ions.
 “Non-luminal” sites are the population of 10-30 binding sites located mostly at the lipid-protein interface for which an allosteric mechanism of non-competitive inhibition was proposed. Agents of different origin (steroids, fatty acids, alcohols, local anesthetics etc.) can bind to those sites and modulate nAChR activity.
 Other classes of ligand-gated ion channels include GABA (Johnston G. A. (2002) Medicinal chemistry and molecular pharmacology of GABA(C) receptors. Curr Top Med Chem 2, 903-13), 5HT3 (D. C. Reeves, S. C. Lummis, (2002) The molecular basis of the structure and function of the 5-HT3 receptor: a model ligand-gated ion channel (review). Mol. Membr. Biol. 19, 11-26), AMPA (T. B. Stensbol, U Madsen, P. Krogsgaard-Larsen, (2002) The AMPA receptor binding site: focus on agonists and competitive antagonists. Curr. Pharm. Des. 8, 857-72) and NMDA (K. A. Macritchie, A. H. Young, (2001) Emerging targets for the treatment of depressive disorder. Expert Opin. Ther. Targets 5, 601-612) receptors, etc. Although the molecular structure of these receptors differ significantly, it is believed that the luminal domains are homologous to the luminal domain of nAChRs. There are five (or occasionally four) transmembrane helices forming the wall of the channel with “rings” of polar amino-acids exposed on the pre-forming surface and the same non-competitive inhibition phenomenon can be observed.
 In summary, the luminal high affinity NCI binding domain is located on the surface of the internal lumen forming the ion channel. Drugs of different origin bind in this region and sterically plug the channel blocking the flux of ions.
 Non-competitive inhibition of the nAChR can be responsible for severe adverse drug effects. On the other hand, designing ligands that specifically interact with this site can be part of the development of new treatments of Alzheimer's and Parkinson's diseases, for example by identifying compounds likely to exhibit side effects through non-competitive inhibition of a LGIC. Furthermore, the compounds identified as NCIs by the present method are likely to find use in treating Tourette's syndrome and cognitive disorders, pain [see, Lloyd, G. K. and Williams, M. (2000) J. Pharmacol. Exper. Ther. 292, 461-467.], anxiety, depression, neurodegeneration and addictions caused by an overactive LGIC receptor, especially diseases in which nicotine agonist activity against a neuronal nAChR is part of the etiology (e.g. smoking addiction). The invention can also be used to evaluate cardiovascular toxicity of a compound mediated by non-competitive inhibition of a LGIC receptor, e.g. arrythmia and GI spasming or diarrheal side effects of a compound caused by inhibition of a muscle nAChR.
 Classical methods of NCI identification are time consuming and not effective in rapid screening of chemical libraries of drug candidates.
 Several different molecular models of the nAChR transmembrane domain have been reported (Capener C E, Kim H J, Arinaminpathy Y, Sansom M S (2002) Ion channels: structural bioinformatics and modelling. Hum Mol Genet 11:2425-33). However, none of those models were used to investigate interaction with channel blockers. A computer based model for in silico simulations of NCI interactions with the luminal domain of LGICs is needed to better understand the phenomenon of the receptor's inhibition by NCIs.
 Furthermore, in drug discovery, the potential adverse effects of drug candidates are of great importance. In depth understanding of mechanistic interaction of luminal NCIs with different subtypes of LGICs, especially of nAChRs, is required to remove potential unwanted side effects at this site. In this respect, a rapid screening technology that would identify NCIs of LGICs, and especially of nAChRs would be greatly desired.
 The functional determination and characterization of a NCI of an LGIC is very complex and time consuming. One approach is affinity chromatography based on immobilized receptor protein. This is a versatile tool for investigation of intermolecular interactions of a receptor with its ligands. The chemometric approach of affinity chromatography can be employed for determination of reliable relative affinities of ligands as well as kinetic characterization, which otherwise would be inaccessible, for large set of compounds (Kaliszan R., Wainer I. W. (1997) Combination of Biochromatography and Chemometrics: A Potential New Research Strategy in Molecular Pharmacology and Drug Design. In Chromatographic Separations Based on Molecular Recognition. K. Jinno, editors Wiley-VCH).
 Methods using nAChR and other receptors immobilized on a chromatographic support have been elaborated (U.S. Pat. Nos. 6,387,268, 6,139,735, provisional application No. 60/337,172). It was shown that the obtained stationary phases worked as selective binding materials for competitive cholinergic ligands and can be used for high throughput screening of various competitive agonists and antagonists (R. Moaddel, I. W. Wainer, (2003) Immobilized nicotinic receptor stationary phases: going with the flow in high-throughput screening and pharmacological studies J Pharm Biomed Anal. 30, 1715-24). The usefulness of such columns based on immobilized nAChR for investigations and modeling of NCI affinity has also been demonstrated. Using a novel non-linear chromatography approach off and on kinetics of ligand interaction with the receptor can be determined. (K. Jozwiak, J. Haginaka et al., (2002) Displacement and nonlinear chromatographic techniques in the investigation of interaction of noncompetitive inhibitors with an immobilized α3β4 nicotinic acetylcholine receptor liquid chromatographic stationary phase. Anal Chem 74: 4618-4624).
 The features of the invention may be better understood by reference to the drawings described below.
FIG. 1 shows a structure of a neuronal nicotinic acetylcholine receptor (nAChR).
FIG. 2 is a schematic representation of a computer system useful in the practice of the invention.
FIG. 3 is a model of luminal domain of α3β4 subtype of nAChR illustrating its electrostatic potential of the inner surface of the channel. The Figure particularly shows the electronegative potential of the cation selector region of the channel.
FIG. 4 shows a luminal domain model having five helices forming the wall of the ion channel.
FIG. 5 shows a luminal domain model (in wireframe rendering) in perpendicular view. Specific residue rings depicted in different colors: going down—red—E, first orange—T, second orange S, grey—S/A, blue—V/F, green LL, last red—E/K, respectively).
FIGS. 6-8 are of example binding complexes. FIG. 6 shows the mecamylamine luminal domain of α3β4. FIG. 7 shows the MK-801 luminal domain of α3β4.
FIG. 8 shows a two cluster interaction of the ligand PCP with α3β4. Generally NCIs bind into the small pocked formed on the apolar domain (Phenylalanine/Valine rings). Tested structures primarily entered hydrophobic pocket formed between α3 and β4 strains and subsequently interacted with protein side chains forming hydrogen bonds. Ligands most likely form two separate clusters on two symmetrical active sites. Estimated free energies of docking are in the range of experimental IC50 of tested inhibitors.
FIG. 9 shows example compounds tested on α3β4 of nAChR column. Among the tested drugs are aliphatic amines like mcm, amt, mtn and such compound like bup, ket and mk-801. Also, some examples of more complicated structures include clo, pcp mtd vera. Further, the structures of two enantiomer dmt and lmt. Finally, there is a structure for ethidium: the only compound permanently ionized and binds to its specific site.
FIG. 10 shows a correlation of log k′ (chromatographic) with log (1/ki) (docking simulation).
FIG. 11 shows the enantioselectivity of dextromethorphan/levomethorphan pair determined in chromatographic experiments. Dextromethorphan had longer retention time and the profile was more asymmetric.
FIG. 12 shows a comparison of dextromethorphan/levomethorphan complexes obtained in a docking simulation: dextromethorphan—grey wireframe structure, levomethorhan—red wireframe structure. Both systems interact initially with hydrophobic pocket on the border between α3 and β4 strains this binding determines positions of amine group different for dextromethorphan (blue) than levomethorphan. In case of dextromethorphan amine group can easily form secondary interaction hydrogen bonds with neighboring polar residues (orange balls), while levomethorphan is more likely to form such interaction. This make a difference in stabilities of two complexes by ca. 0.3 kcal/mol determined by both docking and chromatographic analysis (FIG. 11).
 The present invention lies in part in a computer system that generates molecular models of ligand-gated neuronal receptors and a method of using the same. The computer system generates a computer-based model of the inner lumen of a ligand-gated ion channel having binding pockets for non-competitive inhibitors. The computer system simulates interaction of structures from chemical libraries or of any desired compound with the generated computer-based model of the ligand-gated ion channel. The simulation can serve to predict and describe the pharmacological importance of the interaction. Thus, the invention constitutes a system for drug discovery and for screening of a drug candidate for unexpected side effects and toxicities.
 In an embodiment of the present invention, as shown in FIG. 2, the computer system comprises a memory, e.g. disk 105, storing positional data of the atomic coordinates of the transmembrane portion of at least one subunit of a ligand-gated neurotransmitter receptor protein, and a processor 101 generating a molecular model having a three dimensional shape representative of the pore portion of the ligand-gated neurotransmitter receptor based on positional data. During execution of the process for generating the molecular model, it is understood that the positional data would be stored in, for example, RAM 102, or other memory readily accessible by the processor 101.
 The memory, in particular, stores data of the atomic coordinates of at least an α chain and a β chain of a nicotinic acetylcholine receptor. The data of the atomic coordinates can include atomic coordinates of at least one polypeptide having an amino acid sequence selected from the group consisting of the polypeptides shown in Table 1 (SEQ ID NOS: 1-15). The data of the atomic coordinates should include atomic coordinates of the portion of the transmembrane portion of the subunit consisting of the amino acid sequence of residues 8 to 19 of SEQ. ID NOS: 1-15.
 The processor 101 can generate a molecular model of the pore portion of a ligand-gated neurotransmitter receptor having a subunit stoichiometry ranging from (α)5(β)0 to (α)0(β)5. For example, the subunit stoichiometry can include(α)2(β)3 useful for modeling the neuronal nAChR regulating cardiovascular and GI actions.
 Modeling Step:
 In generating a molecular model and simulating its interaction, the computer system of the present invention first generates a molecular model of the receptor channel based on a template structure determined in an NMR investigation of synthetic channel model (Opella S. J., Marassi F. M., Gesell J. J., Valente A. P., Kim Y., Oblatt-Montal M., Montal M., (1999) Structures of the M2 channel-lining segments from nicotinic acetylcholine and NMDA receptors by NMR spectroscopy. Nat. Struct. Biol. 6:374-9). Using this model, the molecular structures of all of the neuronal subtypes of nAChR can be built. All subtypes of nAChR share several common structural arrangements in the luminal domain, which makes it possible to build the model of a particular subtype using a homology modeling approach.
 Once a molecular model is generated, the model is refined. A preferred software package for refining the molecular model is the AMBER molecular modeling package, e.g. AMBER version 7, (D. A. Pearlman, D. A. Case, J. W. Caldwell, W. S. Ross, T. E. Cheatham III, S. De Bolt, D. M. Ferguson, G. L. Seibel and P. A. Kollman, (1995) AMBER, a package of computer programs for applying molecular mechanics, normal mode analysis, molecular dynamics and free energy calculations to simulate the structural and energetic properties of molecules. Comp. Phys. Comm. 91, 1-41). The AMBER package contains a set of molecular mechanical force fields for the simulation of biomolecules and a package of molecular simulation programs. In particular, the model is preferably refined using the “SANDER” program (for Simulated Annealing with NMR-Derived Energy Restraints) was used. SANDER is the main program used for molecular dynamics simulations. SANDER allows for NMR refinement based on NOE-derived distance restraints, torsion angle restraints, and penalty functions based on chemical shifts and NOESY volumes.
 Once the model has been refined using the SANDER program of AMBER, the final model is evaluated. A preferred software package for evaluating the final model is the PROCHECK package, e.g. version 3.5.4 (Laskowski R A, MacArthur M W, Moss D S & Thornton J M, (1993) PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Cryst., 26, 283-291). PROCHECK checks the stereochemical quality of a protein structure, producing a number of PostScript plots analyzing its overall and residue-by-residue geometry.
 In order to construct subtype-specific molecular models, the primary structures of the particular subtypes are required. Different subtypes can be found in different region of the human brain and are responsible for specific functions. Subtype-specific models of the nAChR luminal domain can be utilized in designing subtype-specific NCIs.
 The procedure of building the luminal model can be easily adopted to constrain models of luminal domain of other subtypes of the nAChR and with some modification to constrain lumen models of other classes of ligand-gated ion channels. The procedure is basically explained in the modeling step of Example 1. The model of the α3β4-nAChR can serve as the template to constrain other neuronal and muscular subtypes: since those subtypes are very homologous (see Table 1). Only a few residues need to be modified in order to obtain new subtype. The new model after residue modification must be subjected to energy minimization by AMBER procedures described previously and finally should be evaluated using PROCHECK. Elaborated docking procedures can be applied to those models and the entire approach can be used in detailed molecular characterization of the luminal domain of specific subtypes of nAChR and moreover, subtype specific interaction with different classes of NCIs.
 More complicated procedures must be applied if one want to obtain model of the domain formed by other classes of ligand-gated ion channels (GABA, NMDA, 5HT3 etc). First, amino-acid sequence alignment modeling is performed. An example and detailed description of such analysis can be found in the paper by Bertaccini and Trudel [E. Bertaccini and J. R. Trudell, (2002) Predicting the transmembrane secondary structure of ligand-gated ion channels Protein Eng. 15, 443-453]. Thus, homologous parts of the ion channel can be found and a new model of transmembrane domain LGIC can be made. For some LGICs, the transmembrane domain is formed by four transmembrane helices instead of five as in the case of nAChR. In such case one of the helices must be removed and remaining four need to be properly repositioned in order to form channel structure. Then the model can be relaxed and refined in AMBER procedures and finally evaluated in PROCHECK. In case of such distinct models the docking procedures need be parameterized by initial studies as described in the simulation step of Example 1 the invention and the values of the size of the grid box, the dielectric constant, the ga_num_evals must be optimized, since the size, and environment of the channel would have been changed significantly.
 Using the modeling method of the invention, it has been discovered that there are NCI binding sites at the interface between α and β helices of LGICs, especially of nicotinic AchRs. Among modeled candidate NCIs, the compound enters into a small hydrophobic pocket formed by residues 12, 15 and 18 of the transmembrane domains of the receptor subunits (e.g. SEQ ID NOS: 1-15, Table 1). A hydrophobic group of the NCI compound will interact with this portion of the NCI binding site. A polar group (e.g. an amino group) of a putative NCI can interact by hydrogen bonding with surrounding polar residues (e.g. residues 12 and 14 of SEQ ID NOS: 1-15).
 Simulation step:
 After generating the molecular model, the final molecular model is used as a target protein for docking simulation for compounds that may be potential inhibitors. A preferred software package for docking simulation is the AutoDock package, e.g. version 3.5. AutoDock allows docking of a flexible ligand into a rigid structure of the target protein using genetic algorithms as the search method.
 A particular genetic algorithm included in the AutoDock package is the Lamarckian genetic algorithm. The Lamarckian genetic algorithm was preferably used with local search in order to improve efficiency. The Lamarckian genetic algorithm works in a reverse order compared to typical genetic algorithms. In particular, new traits in an organism develop because of a need created by the environment and these acquired characteristics are transmitted to its offspring. In AutoDock the ligand's atomic coordinates represent a genotype and fitness is represented by interaction free energy with the proteins. Genotypes are found through interations of the local search and then the atomic coordinates are translated into the ligand's state coordinates, as the phenotype. In other words, in AutoDock local search is used to update the fitness associated with an individual in the genetic algorithm selection.
 The Lamarckian genetic algorithm uses as input a grid data set produced by the AutoGrid module. The AutoGrid module is used to create 3-dimensional maps over the host protein using several atom specific and electronic probes at each grid point.
 Results of these simulations allow the classification of tested compounds in terms of free energy of binding, which leads to the identification of ligands that may be potent inhibitors. The same approach can be used to design new compounds with high binding properties to a specific subtype of the nAChR. A compound that is identified as a non-competitive inhibitor of a LGIC is one having a ΔG less than −6 kcal/mol, preferably less than −7 kcal/mol, still more preferably one having a ΔG less than −10 kcal/mol.
 The ligand structures used in docking simulations are preferably made using the HyperChem package (of HyperCube Inc., Gainsville, Fla.). In particular, it is preferred that the AM1 semiempirical method implemented in HyperChem be used to minimize the system energy and to calculate atomic charges in final structures (J. J. P. Stewart, Semiempirical molecular orbital methods, in: K. B. Lipkowitz, D. B. Boyd (Eds.), Reviews in Computational Chemistry, vol. 1, VCH, New York, 1990, pp. 45-81).
 The in silico approach described above can be supported by examining the NCI-nAChR interaction by affinity chromatography (Jozwiak K, Haginaka J, Moaddel R and Wainer I W (2002) Displacement and nonlinear chromatographic techniques in the investigation of interaction of noncompetitive inhibitors with an immobilized nicotinic acetylcholine receptor liquid chromatographic stationary phase. Anal Chem 74: 4618-4624), preferably in an iterative fashion. Chromatographic affinity screening can provide experimental data that is then employed for proper parameterization of the computer-based molecular simulation. Alternatively, the results of computer-based simulation can be related and evaluated by further chromatographic and functional experiments.
 Until recently, the screening of drug candidates for activity as NCIs was not a standard procedure in the drug development process. However, the present invention will permit pharmaceutical companies to rapidly screen their potential drugs for NCI properties. In addition, the luminal domain of nAChR can be used as a target in drug discovery programs, which represents a new therapeutic approach to the treatment of diseases such as Alzheimer's and Parkinson's diseases and for treatment of drug and tobacco dependency, which are related to LGIC functions,especially to nAChR functions.
 The nAChR, for example, was found to contain two cholinergic agonist binding sites located at the interface between the α and β subunits and on the extracellular N-terminal of the α subunits. These sites are key targets for drug discovery in a variety of diseases, including Alzheimer's disease (α4β2), Parkinson's disease (α3β2), cardiovascular and GI actions (α3β4), anxiety and depression (α4β4), short term memory (α7) and auditory function and development (α9).
 Candidate NCI compounds discovered by the computational modeling method of the invention can be confirmed by in vitro experimental methods. Two preferred methods are by binding experiments or by functional assays. Either of these methods may employ the target LGIC, a population of LGICs representing the target receptor and receptors that the compound should preferably not inhibit (to avoid side effects), or a population of LGICs representing a group of target receptors (with or without a group representing LGICs that the compound should preferably not inhibit). The LGICs for the in vitro functional assays can be present either as expression products in cells, a partially purified proteins, e.g. membrane preparations made as known in the art, or as isolated proteins. If isolated proteins are used in binding experiments, the proteins are preferably immobilized.
 A preferred binding assay is a displacement assay performed as described by Jozwiak et al. [Jozwiak K, Haginaka J, Moaddel R and Wainer I W (2002) Displacement and Nonlinear chromatographic techniques in the investigation of interaction of noncompetitive inhibitors with an immobilized α3β4 nicotinic acetylcholine receptor liquid chromatographic stationary phase. Anal Chem 74: 4618-4624.] Using this assay, a compound is identified as a non-competitive inhibitor of the ligand-gated neurotransmitter receptor is one that specifically binds to the ligand-gated neurotransmitter receptor with a k′ value greater than 8, preferably with a k′ value greater than 9 or even more preferably a k′ value greater than 10.
 Specificity of NCI binding to particular LGICs can be shown by displacement of compounds that are selective to the pore portion of the desired LGIC. Specificity of the binding to a nicotinic AChR and homologous receptors can be shown by displacement by mecamylamine. Displacement of mecamylamine at a concentration of 10 μM indicates good specific binding, ability to displace mecamylamine at a concentration of 40 μM indicates strong specific binding. Preferably it is possible to displace mecamylamine at a concentration of 100 μM. Thus, a compound that is a preferred NCI of a nicotinic AChR is one that exhibits a k′ of greater than 8 in a chromotagraphic binding experiment and can be displaced by mecamylamine at a concentration of 10 to 100 μM.
 Preferred functional ion channel activity assays are described by Hernandez et al. [Hernandez S C, Bertolino M, Xiao Y, Pringle K E, Caruso F S and Kellar K J (2000) Dextromethorphan and its metabolite dextrorphan block α3β4 neuronal nicotinic receptors. J Pharmacol Exp Ther 293: 962-967] and by Jozwiak et al. [K. Jozwiak, S C Hernandez, K J Kellar, I W Wainer (2003) The Enantioselective Interactions of Dextromethorphan and Levomethorphan with the α3β4-Nicotinic Acetylcholine Receptor: Comparison of Chromatographic and Functional Data submitted to J Pharmacol Exp Ther]. In these assays a compound is identified as a NCI that inhibits the ion channel activity of the ligand-gated neurotransmitter receptor in nicotine stimulated 86Rb+ efflux with an IC50 lower than 50 μM. A more preferred NCI compound is one that inhibits ion efflux with an IC50 lower than 5 μM. Even more preferable compounds are those that inhibit ion efflux with an IC50 lower than 500 nM. One of skill in the art will recognize that compounds that are effective at even lower concentrations are still more preferable, and IC50 of 50 nM, or even 5 nM might be observed.
 In some instances as described above, it might be preferred to have a NCI that is selective for a particular LGIC. By “selective” is meant that the NCI inhibits the target LGIC with an IC50 that is at least 5-fold higher than the IC50 of the one or more LGICs that it is desired not to inhibit. The degree of selectivity is preferably 10-fold, more preferably 20- to 50-fold, and still more preferably 100- to 500-fold or more.
 On the other hand, the binding assays or functional assays also can be used to provide initial data that can be used to constrain the in silico modeling method desribed above. Alternatively, the in silico modeling and the in vitro assays can be run iteratively to converge upon NCI compounds that have desired properties.
 A structure-activity relation for a NCI of a LGIC has been derived using the above-described methods. Thus, a compound having a bulky hydrophobic moiety (e.g., a phenyl or napthyl ring system, cyclopentyl or cyclohexyl ring system, a fused ring system including but not limited to bicyclo [2.2.1] heptane, bicyclo [2.2.2] octane, morphinan and dibenzo [1.4] diazepine) and a primary, secondary or tertiary amino group in proximity to (i.e, approximately 5 to 10 Å from, preferably from 5 to 8 Å from, more preferably less than 7 Å from) said hydrophobic moiety. The amino group can be directly bonded to the bulky hydrophobic moiety or can be linked by a spacer moiety, such as, but not limited to, a short hydrocarbon chain. The amino group can be substituted (—NR1R2, where R1 and R2 are the same or different and are selected from the group consisting of H, C1-C3 alkyl, C1-C4 alkoxy, dialkyl keto). The substituent is preferably one that retains a hydrogen-bonding potential; a preferred substituent is a keto-group, for example a dialkyl keto group, especially CH2(C═O)CH3. Another preferred substituent is a hydroxyl or alkoxyl (—CH2OH) group, e.g. a C1-C4 normal or branched alkoxyl group. Preferred substituted amino groups are a dialkyl keto amino group (e.g., HNCH2(C═O)CH3), a hydroxyl amino group or a methoxy amino group. An example of such a compound is 3-methoxy-17-propane-2-one 9 α, 13α, 14α morphinan.
 No compound listed in Table 2 is considered a compound per se of the invention. Methods of the invention for non-competitively inhibiting a LGIC, especially a nicotinic AChR, or for treatment of a disease mediated by overactivity of a nicotinic AChR, exclude the use of bupropion, ketamine, laudanosine, mecamylamine, methadone, MK-801, phenylcylclidine, ethidium, and dextromethorphan.
 Methods for synthesis of compounds of the invention are considered within the skill of the ordinary synthetic chemist. Preferred NCI compounds have the above structural features and exhibit activity of inhibiting the ion-channel activity of a ligand-gated neurotransmitter receptor in nicotine stimulated 86Rb+ efflux with an IC50 lower than 100 μM or other activities as set forth in detail above.
 Dosage of compounds used for treatment of a subject can be easily determined by the ordinarily-skilled pharmacologist using known pharmacokinetic and pharmacodynamic assays and calculations from IC50 data obtained by the inventive method. Formulation and administration of compounds useful for treatment is also well-known in the art. For example, many of the compounds listed in Table 2 have been administered therapeutically and it is expected that compounds of the invention can be similarly formulated and administered.
 The molecular model of a δ-M2—nAChR transmembrane channel determined by frozen state NMR was used as the template for further modification (atomic coordinates were found in Protein Data Bank—PDB id: 1EQ8). This model represents a channel that mimics the transmembrane arrangement of known LGICs (Opella S. J., Marassi F. M., Gesell J. J., Valente A. P., Kim Y., Oblatt-Montal M., Montal M., (1999) Structures of the M2 channel-lining segments from nicotinic acetylcholine and NMDA receptors by NMR spectroscopy. Nat. Struct. Biol. 6:374-9). The model channel consisted of 5 uniform polypeptides oriented around a central pore. The amino-acid sequence of this polypeptide is analogous to the sequence of transmembrane M2 segment of δ subunit of nAChR found in Torpedo californica. Table 1 presents the primary structure of this δ-M2-segment.
 In the δ-M2—nAChR transmembrane channel, the spatial arrangement of polypeptide helices conserves five-fold symmetry, with certain residues exposed to the center of the pore. These residues (predominantly polar) form an explicit surface of the channel. This is consistent with the concept of the presence of amino acid rings distributed along the pore and is a common property found in all subtypes of nAChR and also other ligand-gated ion channels [Changeux J. P., Galzi J. L., Devillers-Thiery A., Bertrand D., (1992) The functional architecture of the acetylcholine nicotinic receptor explored by affinity labelling and site-directed mutagenesis. Q. Rev. Biophys. 25: 395-432].
 With respect to the spatial arrangement of five helices in the luminal domain, distribution of certain amino-acid rings along the channel is a common property of all subtypes of nAChR. Since primary sequences across different subtypes are predominantly homologous as presented in Table 1, and essential (exposed) residues are highly conserved, a subtype specific model of the luminal domain can be built using homology modeling techniques.
 Based on the sequence comparison presented in Table 1, the initial model was modified by exchange of δ helix residues into α3 and β4 using the SYBYL 6.8 molecular modeling system (Tripos Inc., 1699 South Hanley Road, St. Louis, Mo., 63144, USA). Therefore, the channel containing α3, β4, α3, β4 and β4 helices, respectively, was constrained.
 The model was further refined by energy minimization using the SanderClassic module of AMBER 6.0 software. Both termini of each helix were blocked in a standard AMBER procedure: acetyl beginning groups (ACE) and N-methylamine ending group (NME) groups were attached, respectively, to each helix. The AMBER '94 force field (Cornell, W. D., Cieplak, P., Bayly, C. I., Gould, I. R., Merz, Jr. K. M., Ferguson, D. M., Spellmeyer, D. C., Fox, T., Caldwell, J. W., Kollman, P. A., (1995) J. Am. Chem. Soc. 117, 5179-5197) parameters were used for energy minimization with the convergence criterion of the root-mean-square of the gradient to be less than 1.0E-4 kcal/mole Å. Each minimization run was started with the steepest descent followed by the conjugate gradient method. A distance-dependent dielectric function was used to evaluate the electrostatic energy. The energy minimization run was carried-out in stages by relaxing i) only hydrogen atoms, ii) hydrogen+side-chain atoms, or iii) all atoms except alpha-carbons. Finally, a restrained minimization was also performed on the alpha-carbons of all the chains/residues of the model. This was to relax the structure but keep it near the initial position of the known template structure (PDB accession no. 1EQ8). Respective scripts used to run model refining with AMBER are presented in Apendix 1.
 Using PROCHECK to evaluate the model it was found that the whole luminal domain is constrained fully by α-helix secondary structure. Along the lumen model seven rings of residues exposed to the center of the channel can be found; three polar residues (E, T and S) and then three apolar residues (L, V/F and LL) and the last polar residue (E/K).
 It is believed that apolar rings in the middle of the structure form the actual “gate” of the channel and play a role in conformational change of the receptor from a closed to an open state. Polar residues on both sides of the “gate” participate in the cation selective function of the receptor. An important structural parameter found in the obtained model is the change in position from valine in the α3 sequence to phenylalanine in the β4 sequence (see residue 15 in Table 1). This provides the formation of small pockets between α3 and β4 subunits, found during the simulation of NCI-α3β4-nAChR interactions. The developed model of α3β4-nAChR luminal domain can be used as a template to constrain homologous systems of other nicotinic receptors, especially neuronal nicotinic receptors, and other ligand-gated ion channels.
 The resulting atomic coordinates represent the final model. FIG. 3 illustrates the electrostatic potential of the inner surface of the ion channel, and especially the electronegative potential of the cation selector of the channel. FIG. 4 shows an example of the resulting luminal domain model having five helicies forming the wall of the ion channel.
FIG. 5 shows a luminal domain model in perpendicular view with residue rings.
 In order to perform docking simulations, the AutoGrid module was first used to create 3-dimensional maps over the host protein using several atom specific and electronic probes at each grid point. An example parameterization file for the AutoGrid module used in this example can be found in Appendix 2. The optimal size of constrained grid maps was a 22.5×22.5×45 Å box (i.e., a grid of 60×60×120 points, each separated by 0.375 Å). This allowed exploration of the whole internal space of the lumen domain but prevented ligands from being bound on the external side. The grid-box size can be altered in the 3rd dimension (along the lumen) in order to explore interaction with a particular segment of the lumen or to calculate the interaction profile along the model.
 An important parameter to properly explore electronic interaction in ligand receptor complexes is the dielectric constant value (d) used to calculate the electronic grid map. During the initial evaluation tests, the standard distant-dependent dielectric constant did not produce proper results: the electrostatic interaction were almost zero. The simulation did not discriminate between neutral and protonated ligands.
 A detailed test of several d values was carried out using three pairs of ligands and the results are presented in Table 2. Table 2 shows an unexpected diminished difference between neutral and protonated systems when distant-dependent d was used; differences gradually increase with decreasing d. Simultaneously the increase in electrostatic impact in the ligand receptor interaction was noticed when a low dielectric value was used. However, a very low value (d≦10) produced unrealistic ΔG values. Finally, as a mater of compromising these two effects, d=15 was chosen for final calculations as the value producing suitable electronic properties of the ligand-receptor complex in the transmembrane ion channel system. This approach is in agreement with values of the dielectric constant in transmembrane pores obtained by theoretical calculations (Cheng et al., (1998) Eur. Biophys J., 27105-112 and Gutman et al., (1992) Biochim. Biophys Acta 1109: 141-148) where it was found that the actual dielectric constant in transmembrane channels remains low and ranges from 25 to 5 depending on the structure. Thus, in the case of the NCI-nAChR docking simulations d value can vary from 10 to 20.
 The resulting ligand 3D structure was loaded into the AutoDock system and was iteratively sampled over previously created grid-maps in order to find optimal positions and the lowest energy of interaction. An example parameterization file for the AutoDock module used in this example can be found in Appendix 3.
 The Lamarckian genetic algorithm with local search was used from the AutoDock package. Atomic coordinate files of ligands were transformed into format suitable to AutoDock using the HIN2PDBQ script (Johansson M. (2002) Some computational chemistry related python conversion scripts. See Web site helsinki.fi/%7Empjohans/python/).
 The ligand structures used in the docking simulations were made using the HyperChem software package. Further, the AM1 semiempirical method implemented in HyperChem was used to minimize the system energy and to calculate atomic charges in final structures.
 An initial simulation was performed in order to optimize the docking settings. Since previously described docking space seemed to be large in the model of α3β4-nAChR active site (22,781.25 Å3) it was important to optimize the maximum number of energy evaluations (ga_num_evals) required in each search run. It was found that too low a value of ga_num_evals could result in finishing the simulation too quickly, and the global minimum of the complex conformation may not be found. A set of test simulations on several ligands including conformationally flexible and rigid systems was performed. It was found that a ga_num_evals value of at least 5 million is required to assure obtaining a statistically significant number of lowest energy complexes. In the case of bigger ligand molecules with more than 2 rotatable bonds, the optimal value should be at least 50 million. Higher values are acceptable; however higher values may dramatically increase the time of each simulation.
 The optimal number of docking search runs was found to be 50. Again the number of docking search runs can be higher, but would take more time for simulation and have no effect on the final result.
 The AutoDock 3.5 implemented a free-energy scoring function that is based on a linear regression analysis, the AMBER force field, and a large set of diverse protein-ligand complexes with known inhibition constants (e.g. see Web site at scripps.edu/pub/olson-web/doc/autodock/). This function was employed to estimate the free energy change of the NCI-nAChR complex and eventually lead to an estimated inhibition constant of a particular ligand. Docking simulations allow quantitative classification of the stability of the NCI-nAChR complexes formed by tested ligands in terms of free energy of binding, which eventually lead to the identification of ligands exerting potent inhibitory properties. It was found that molecular systems forming the complex with ΔG value lower than −6.0 kcal/mol should be considered as potential NCIs. Lower ΔG values represent more potent NCI compounds. Preferred NCI compounds exhibit a ΔG value lower than −7.0 kcal/mol; more preferred compounds exhibit a ΔG value lower than −10.0 kcal/mol.
 Detailed exploration of the spatial arrangement of ligand-receptor conformations leads to building a pharmacophore model of α subtype specific NCI-nAChR. Simulations on the α3β4 model showed that NCIs bind predominantly into the channel on the apolar domain (F/V ring). Tested structures primarily entered a small hydrophobic pocket formed between α3 and α34 subunits and subsequently interacted with protein side chains forming hydrogen bonds. It is expected that this is a type of interaction that would not be found in those receptor subtypes that lack the bulky phenylalanine residue in this position. Since there are two quasi-symmetrical pockets between α3 and β4 helices in the model, ligands most likely form two separate clusters on these two symmetrical sites (FIG. 8) at which the energy of interaction does not significantly differ. Estimated free energies of docking are in the range of experimental IC50 of tested inhibitors and also can be related to experimental affinity chromatography results. The model can be applied to a variety of compounds and is useful for in silico designing of new drugs with particularly high non-competitive inhibitory activity.
 Chromatographic studies based on immobilized nAChRs were performed to characterize ligand binding for broad groups of compounds. In order to further understand the mechanistic action of NCIs on the molecular level, the model of the transmembrane domain of the α3β4 nAChR was built and used for computer simulations of docking inhibitors into the receptor. The entire approach allowed the classification of NCIs in terms of their functional effectiveness.
FIG. 9 presents compounds tested on an α3β4 nAChR column. The chemicals can be divided into several subgroups. The first group contains drugs from different origin, which are well known as non-competitive inhibitors of nAChRs. The second group is of the dextromethorphan family, levomethorphan, dextromethorphan and its analogues, and the final group is verapamil, its congeners, and metabolites. In order to properly assess the influence of non-specific retention, five other chemicals (acetanilide, acetaminophen, 2,4-dinitrobenzoic acid, 3,4-dimethoxybenzoic acid and phenylbutazone) were tested as negative controls. The affinity of ligands was investigated by non-linear chromatography on an α3β4, nicotinic receptor affinity column.
 106 Cells from the KXα3β4R2 cell line were suspended in Tris-HCl [50 mM, pH 7.4] (buffer A), homogenized for 30 sec, and centrifuged at 35,000×g for 10 min at 4° C. The pellet was resuspended in 2% cholate in buffer A and stirred for 2 h. The mixture was centrifuged at 35,000×g for 30 min, and the supernatant containing α3β4 nAChR-cholate solution was collected. 200 mg of the IAM stationary phase was added to the α3β4 nAChR-cholate solution. Subsequently the solution was stirred for 1 h. The suspension was dialyzed against 2×1L buffer A for 24 h at 4° C. The IAM liquid chromatographic support containing the α3β4-nAChR was packed into a HR5/2 glass column to form a chromatographic bed of 20 mm×5 mm i.d. The α3β4-nAChR column was then placed in the chromatographic system and used. The non-linear chromatography approach was used to determine kinetics of the NCI-nAChR interaction in affinity chromatography studies. The mathematical model assumes limited (and a relatively low) number of active sites on the column. Slow association and dissociation of the drug-protein complex are the main cause of band broadening and asymmetry of the peak profile. The chromatographic peak profiles were analyzed using PeakFit v4.11 for Windows Software (SPSS Inc., Chicago, Ill.). The mathematical approach used was the non-linear chromatography (NLC) model derived from Impulse Input Solution [Wade J L, Bergold A F and Carr P W (1987) Theoretical description of nonlinear chromatography, with applications to psychochemical measurements in affinity chromatography and implications for preparative-scale separations. Anal Chem 59:1286-1295.] and described by Equation 1 (PeakFit User's Manual, p. 8-25):
 y—intensity of signal,
 x—reduced retention time,
 I0( ) and I1( ) are Modified Bessel functions
 a0—area parameter,
 a1—center parameter, reveal to true thermodynamic capacity factor,
 a2—width parameter,
 a3—distortion parameter.
 Experimental chromatograms obtained by single injection of ligand into the chromatographic column with immobilized receptor were processed with PeakFit v4.11 software. After standard linear baseline subtraction, each peak profile was fitted to the NLC function. The set of NLC parameters (a0, a1, a2 and a3) was collected for each profile and used for the calculation of descriptors of the kinetic interactions with the immobilized nAChR, dissociation rate constant (koff); equilibrium constant (Ka); association rate constant (kon) real thermodynamic capacity factor (k′), according to the following equations:
k′=a1 Eqn. 2
 where: t0 is the dead time of a column (time needed by non-retained substance to reach the detector); C0 is a concentration of solute injected multiplied by a width of the injection pulse (as a fraction of column dead volume).
 Thus, by analyzing the ligand in an immobilized receptor system four descriptors can be collected: retention (k′), association rate constant (kon), dissociation rate constant (koff) and equilibrium constant (logK). It was found that ligands which are non-competitive inhibitors have k′ greater than 8, kon greater than 10×10−6 M−1s−1 (preferred inhibitors have kon of greater than 15×10−6 M−1s−1 especially potent inhibitors have kon greater than 30×10−6 M−1s−1), koff smaller than 15 s−1 (preferably lower than 2 s−1) and logK greater than 5.9 (preferably greater than 6.5).
 The kon value obtained in chromatographic experiments is the one which is closely correlated with IC50 values from functional in vitro or in vivo experiments. In the docking simulation, it is preferred that ΔG be lower than −6 kcal/mol (preferably less than −7 kcal/mol, most preferably less than −10 kcal/mol). In functional nicotine stimulated Rb+ efflux experiments, the IC50 value is preferrably lower than 100 μM (preferred inhibitors exhibit an IC50<10 μM).
 Values of logK and k′ presented in Table 3 can be regarded as a measure of relative affinity of tested NCI compounds for the nicotinic AChR. Among tested compounds, ethidium, clozapine, verapamil and some of its congeners (PR-22, nor-verapamil and galapamil) have the highest affinities towards the α3β4 nicotinic receptor column as reflected by both logK and k′. Both verapamil and nor-verapamil were tested for enantioselectivity of binding towards nicotinic affinity column but chromatographic experiments as well as NLC data did not exhibit noticeable differences between enantiomers. Interestingly, dextromethorphan exhibited markedly increased affinity compared to the optical enantiomer levomethorphan.
 The NLC approach allows estimating the kinetic rates of the complex formation and dissociation, kon and koff, respectively. The well-known and potent NCIs mecamylamine, ketamine, ethidium and bupropion had high association constant rates. Ketamine, methamphetamine, amantadine and mecamylamine dissociated markedly quicker than other tested ligands. The lowest dissociation constant rates exhibit ethitium, clozapine and verapamil congeners.
 Examples of complexes resulting from simulations are provided in FIGS. 6-8. FIG. 6 shows the mecamylamine luminal domain of α3β4. FIG. 7 shows the MK-801 luminal domain of α3β4. FIG. 8 shows a two cluster interaction of the ligand PCP with α3β4.
 Quantitative results of simulated docking affinities were related to experimental results from chromatographic studies. Using AutoDock's scoring function, estimated inhibition constant were calculated. These values exhibited very good correlations with affinity data from NLC calculations (FIG. 10). This correlation can be illustrated by equation:
log k′=0.418(±10.037) log(1/K i)−0.89(0.19)
r=0.930 F=127.7 n=22
 Enantiomers have identical physiochemical properties and, therefore, all possible non-specific interactions between the enantiomers of a chiral NCI and an immobilized nAChR stationary phase should be equivalent. Any differences in the chromatographic retention between the enantiomers will be due to specific binding interactions with the active site of the protein. FIG. 11 shows chromatographic tracks of dextromethorphan (DM) and its enantiomer—levomethorphan (LM). The pair of enantiomers was further investigated by chromatographic, docking and functional studies (Table 5). It was learned from the chromatographic experiments that the drug dextromethorphan (DM) exert higher affinity on α3β4-nAChR than its enantiomer levomethorphan (LM) and the difference in ΔG of the complexes was 0.3 kcal/mol. These data were valuable in evaluating parameter selection during initial tests of the docking simulations to optimally choose the channel dielectric constant or evaluate the usefulness of the scoring function for calculating estimated ΔG implemented in AutoDock. The docking simulations give insights into chiral recognition on the molecular level (FIG. 12). Furthermore, the estimated inhibition constant obtained during the simulations is very well correlated with equilibrium measures obtained in affinity chromatographic experiments.
 The invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative rather than limiting on the invention described herein. Scope of the invention is thus indicated by the appended claims rather than by the foregoing description, and all changes that come within the meaning and range of equivalency of the claims are intended to be embraced therein.
 All patent and literature references cited herein are hereby incorporated by reference in their entirety and for all purposes, including the following references:
 1. Wainer I W, Zhang Y, Xiao Y, Kellar K J (1999) Liquid chromatographic studies with immobilized neuronal nicotinic acetylcholine receptor stationary phases: effects of receptor subtypes, pH and ionic strength on drug-receptor interactions. J Chromatogr B Biomed Sci Appl 724:65-72.
 2. Zhang Y, Xiao Y, Kellar K J, Wainer I W (1998) Immobilized nicotinic receptor stationary phase for on-line liquid chromatographic determination of drug-receptor affinities. Anal Biochem 264:22-5.
 3. Barrantes F J. (2002) Lipid matters: nicotinic acetylcholine receptor-lipid interactions (Review). Mol Membr Biol 19:277-84.
 4. Morris G M, Goodsell D S, Halliday R S, et al. (1998) Automated docking using a Lamarckian genetic algorithm and empirical binding free energy function. 19:1639-62.