THIS STUDY.. raison d’etre Establishing correlation tools between in vitro and in silico studies for ligand receptor affinity using oestrogen and androgen receptors as case studies • Drug discovery, by the traditional method of screening natural and synthetic compounds is both expensive and laborious • In drug design, potential compounds may be selected for performance of required function based on essential characteristics including idealized structural and physical properties • Molecular models of these new compounds may be built, and virtual tests may be run to assess its suitability, before an expensive synthesis attempt is made • Virtual experiments are cheaper, faster, and safer than real experiments, and the data elucidated from these experiments helps in the elimination of compounds that would definitely not perform the required function. THIS STUDY.. raison d’etre • Structure based drug design is a vital tool in this process • The elucidation of the 3D structure of a binding site or active site, of a target molecule such as a receptor protein, guides researchers to subsets of compounds with desired features to complement the 3D shape of the site • Using geometric and functional features of the binding site, specific moieties of a ligand are designed so as to have a high binding affinity with the target molecule VITAL TO THIS PROCESS…. • Is the understanding of binding modalities of molecules of known affinity to the receptor sites of interest • This is because molecule optimisation often represents another important avenue that contributes significantly towards the obtaining of novel therapeutic agents THIS STUDY.. raison d’etre • Also, since subtle modifications are made to a structure that is known to bind well to its receptor, the predictive ability of ligand enhancement software is enhanced • This is because the effect of a small, as opposed to a drastic change is being quantified • What is actually being done is the determination of the rank order of a list of derivative compounds • This greatly increases the confidence that proposed structures will bind in a more or less predictable manner • Computer software facilitates this job by filtering off all weakly binding compounds allowing synthesis and testing only of the most promising ligand • Thus, utilizing computer aided drug design software to aid in the refinement of weak binding lead compounds has emerged as an efficient tool in modern drug design. THIS STUDY.. raison d’etre • • • • • Drug optimization is advantageous to de novo design in that with de novo design, one cannot know with exact certainty at the outset how the designed molecular structure will interact and bind within a target receptor In drug optimization, there is a far more reliable starting point namely, a lead compound whose bound structure within the receptor has been characterized, usually through X-ray crystallography Subtle modifications can then be performed in order to generate derivative compounds using structure based drug design to improve binding affinity The fact that only subtle modifications are made in order to generate derivative compounds with a higher affinity than the original pharmacophore is a greater guarantee of success These “new” derivatives are then evaluated in order to determine which modifications improve binding. This is an iterative process, which continues until optimally binding ligands are obtained. THIS STUDY.. raison d’etre • This study aims to create reliable, validated predictive tools such that the binding affinities of ligands for oestrogen and androgen receptors, may be estimated with confidence 1 THIS STUDY.. raison d’etre THIS STUDY.. raison d’etre • The choice of oestrogen and androgen receptors lends relevance to this study given that, they are the subject of much research. • Not only are both these nuclear receptors closely linked with the development of breast and prostate cancers respectively… • Management of these conditions is generally long term, and research into non steroidal agents that can effectively treat these conditions with a minimal side effect profile continues to be a significant research avenue • Furthermore, both these receptors are capable of binding nonsteroidal molecules- generally classified as endocrine disruptors • Through this binding process, the endocrine disruptors, by mimicking, blocking or otherwise disrupting the function of endogenous hormones negatively impact the normal function of the endocrine system • These agents, are mainly pesticides & represent ideal lead molecules owing to the fact that they are non-steroidal , have a known binding affinity for these receptors, and may hence be exploited from a design point of view for the development of agents that may be used in the management of both breast and prostate cancer Methodology….. Estimating the Ligand Binding Affinities of Unminimised Ligands of the Oestrogen and Androgen Receptors Preparing Parent Ligand and Allied Protein Oestrogen Receptor Androgen Receptor • The Ligand-Protein complexes to be worked with were identified from the Protein Data Bank (PDB) • pdb ID 1A52 (Oestrogen Receptorα Ligand Binding Domain Complexed to Oestradiol) was selected to model the oestrogen bound ligand • The Ligand-Protein complexes to be worked with were identified from the Protein Data Bank (PDB) • pdb ID 1e3g (Human Androgen Receptor in Complex With the Ligand Metribolone) was selected to model the androgen bound ligand Methodology….. Estimating Ligand Binding Affinities of Unminimised Ligands of Oestrogen & Androgen Receptors- Molecular modelling software SYBYL molecular modelling package (version 6.1, Tripos Associates, Inc., St. Louis, Missouri). Oestrogen Receptor Androgen Receptor • Removal of one monomer, its ligand, allied waters, the heavy metal (Au) • For the remaining monomer the water molecules close to the active site, and which consequently could affect binding were retained. All others were deleted • The Au atom was deleted from the remaining monomer. • The edited version of the pdb file was saved for further use • Water molecules close to the active site, and which consequently could affect binding were retained • All others were deleted • The edited version of the pdb file was saved for further use. Methodology….. Estimating the Ligand Binding Affinities of Unminimised Ligands of the Oestrogen and Androgen Receptors Preparing Parent Ligand and Allied Protein OESTROGEN RECEPTOR ANDROGEN RECEPTOR Methodology….. Estimating the Ligand Binding Affinities of Unminimised Ligands of the Oestrogen and Androgen Receptors • For both androgen and oestrogen ligand receptor complexes, the same process was followed from this point onwards • Edited files read into SYBYL preserving the original coordinates of the pdb files • For each pdb file, the ligand was extracted, named and saved in SYBYL mol2 format • The natural ligand substructure was then totally deleted from the protein molecule into which it was originally docked • The protein molecule, now devoid of its allied ligand, was saved in pdb format • The result of this process therefore was, a ligand- oestradiol and metribolone respectively, saved in mol2 format, and the oestrogen and androgen receptors from which the ligands were extracted and saved in pdb format. 2 Methodology….. Estimating the Ligand Binding Affinity (LBA) between the ligand and its receptor protein • The generated pdb and mol2 files were exported to Score, a software programme that uses an empirical scoring function to describe the binding free energy, which includes terms to account for van der Waals contact, metal-ligand bonding, hydrogen bonding, desolvation effect, and deformation penalty upon the binding process, thus giving a mathematical estimation of the LBA between the ligand and its receptor • This yields a value for pKd where Kd = [ligand] [protein] [ligand:protein] • Thus, if the receptors have a high affinity for the ligand, the Kd will be low and the pKd will be high, as it will take a low concentration of ligand to bind half the receptors Methodology… Preparing the Ligand Series for Scoring • For each receptor type, oestrogen and androgen, a series of ligand molecules was prepared for scoring • The ligand series selected came from the paper of Gao et. al. for the oestrogen receptor, and from the papers of Mekenyan & Bradbury & Waller for the androgen receptor These papers contained experimental Ligand Binding Affinity (LBA) data, determined using laboratory assays. The experimental LBAs were used to compare to those calculated in silico • Each ligand series was built using Sybyl in mol2 format. The individual ligands were fitted onto the co-ordinates of the ligands in the original pdb file in a process that allowed saving of the newly constructed ligand in an orientation identical to that of the original ligand. This ensured that identical coordinates would be maintained for eventual docking into the active site when using Score. Methodology… The Androgen Receptor Ligand Series Methodology… The Oestrogen Receptor Ligand Series Metribolone (parent from pdb) Mibolerone Metribolone 5α-Dihydrotestosterone Oestradiol 5α-Androstane Androstenedione Progesterone 17α-hydroxyprogesterone Corticosterone Metribolone & Androstenedione Pregnenolone Superimposed to Fix Androstenedione Testosterone to the Docked CoAndrostenedione ordinates of Metribolone Methodology… The Oestrogen Receptor Ligand Series The 11 substituted ligands CHMe2 (R isomer) CHMe2 (S isomer) CH2CH=CH2 (R isomer) CH2CH=CH2 (S isomer) CH=CH2 (R isomer) CH=CH2 (S isomer) Thienyl(2) (R isomer) Thienyl(2) (S isomer) CH2C6H5 (R isomer) CH2C6H5 (S isomer) The parent ligand Oestradiol The 16 substituted ligands Br (R isomer) Br (S isomer) F (R isomer) F (S isomer) Cl (R isomer) Cl (S isomer) CH2Br (R isomer) CH2Br (S isomer) Methodology… The Oestrogen Receptor Ligand Series The 17 substituted ligands CCH (R isomer) CCH (S isomer) Me (R isomer) Me (S isomer) CH2CCH (R isomer) CH2CCH (S isomer) CCMe (R isomer) CCMe (S isomer) CCI (R isomer) CCI (S isomer) CH2CCI (R isomer) CH2CCI (S isomer) CH2CH=CH2 (R isomer) CH2CH=CH2 (S isomer) C6H5 (R isomer) C6H5 (S isomer) CH=CHI(Z) CH=CHI(E) CH=CHBr(Z) CH=CHBr(E) CH=CHCl(Z) 3 Methodology… Calculating Ligand Binding Affinity for the 2 Ligand Series • Each new ligand was then docked into the respective receptor (oestrogen or androgen) pdb file • The LBA for each superimposed ligand in both series was calculated in Score • Values obtained were retained for plotting against the experimental LBAs obtained from the literature. Methodology… Minimising the Systems… • There are several different algorithms for minimizing the energy of the system • They all involve calculating the derivative of the potential energy, and possibly the second derivative, and using that information to adjust the coordinates in order to find a lower energy for the system Minimisation Algorithm Types • Steepest Descent • Conjugate Gradient • Conjugent Gradient Powell • Newton Raphson • Adopted basis Newton Raphson • Truncated Newton Methodology Minimising the Systems…Minimisation Protocols Adopted in this Study • Minimisation was carried out using CHARMM (Chemistry at HARvard Macromolecular Mechanics)(a force field for molecular dynamics as well as the name for the molecular dynamics simulation package associated with this force field) • Individual atom charges were generated in MOPAC (Molecular Orbital PACkage)- a computer program designed to implement semi-empirical quantum chemistry • The minimisation was carried out in a stepwise fashion in order to allow for gentle bond relaxation as well as to allow for more variation & control over the protocol and its outcome Methodology… System Minimisation • After the LBA was estimated for the ligand- (oestrogen & androgen) protein complexes, the systems were minimised in order to evaluate whether or not minimisation would affect the LBA profiles Methodology… Minimising the Systems…The Algorithms used Steepest Descent The simplest minimization algorithm is steepest descent (SD) In each step of this iterative procedure, the coordinates are adjusted in the negative direction of the gradient Steepest descents does not converge in general (i.e. reach an absolute minimum), but it will rapidly improve a very poor conformation. Adopted Basis Newton-Raphson This routine performs energy minimization using a Newton-Raphson algorithm applied to a subspace of the coordinate vector spanned by the displacement coordinates of the last positions. The second derivative matrix is constructed numerically from the change in the gradient vectors, and is inverted by an eigenvector analysis allowing the routine to recognize and avoid saddle points in the energy surface. At each step the residual gradient vector is calculated and used to add a steepest descent step onto the Newton-Raphson step, incorporating new direction into the basis set. This method is the best for most circumstances.. Methodology… Minimising the Systems…Minimisation Protocols Adopted in this Study- PROTOCOLS 1&2 Ligand-Protein Complex • STEP 1: Minimise only H atoms using 500 / 1000 steps STEEPEST DESCENT • STEP 2: Minimise protein amino acid side chains and H atoms together using 2000 / 20 000 steps STEEPEST DESCENT • STEP 3: Minimise protein amino acid side chains, H atoms & ligand together using 2000 / 10 000 steps of the adopted basis Newton-Raphson algorithm • STEP4: Minimise entire system now including protein backbone & water atoms using 25000 / 10 000 steps of the adopted basis Newton-Raphson algorithm 4 Methodology… Minimising the Systems…Minimisation Protocols Adopted in this Study- PROTOCOL 3-Gentlest Approach in 8 Stages Methodology… Minimising the Systems…Minimisation Protocols Adopted in this Study- PROTOCOL 3 TRI-LAYER MODEL •Protein : Ligand Complex envisioned as trilayered system: •Layer 1: ligand •Layer 2: protein amino acids & side chains <15A from ligand. Side chains that might crash with ligand during the minimisation could be present in this layer •Layer 3: protein amino acids & side chains > 15A from ligand. No side chain ligand clashes envisaged in this area 1. Fix ligand; Layer 2 strong harmonic constraints; Layer 3 strong harmonic constraints Fix ligand; Layer 2 strong harmonic constraints; Layer 3 weak harmonic constraints Ligand strong harmonic constraints; Layer 2 strong harmonic constraints; Layer 3 weak harmonic constraints Ligand strong harmonic constraints; Layer 2 weak harmonic constraints; Layer 3 weak harmonic constraints Ligand weak harmonic constraints; Layer 2 weak harmonic constraints; Layer 3 weak harmonic constraints Ligand weak harmonic constraints; Layer 2 weak harmonic constraints; Layer 3 no constraints Ligand weak harmonic constraints; Layer 2 no constraints; Layer 3 no constraints Ligand no constraints; Layer 2 no constraints; Layer 3 no constraints 2. 3. 4. 5. 6. 7. 8. Methodology… Estimating LBA for the minimised systems • The minimised ligand : protein complexes were imported into SYBYL • Minimised ligand saved in mol2 format • Minimised protein saved in pdb format • LBA estimated in SCORE Methodology… Oestrogen Receptor Ligands • Unminimised in silico LBA vs experimental LBA (Gao et al.) • Minimised in silico LBA vs experimental LBA (Gao et al.) Androgen Receptor Ligands • Unminimised in silico LBA vs experimental LBA (Waller et al.) • Minimised in silico LBA vs experimental LBA (Waller et al.) RESULTS Androgen Receptor Ligands RESULTS… GRAPH OF pKd (EXPERIMENTAL) vs pKd (in silico) UNMINIMISED Predicted pKd (Before minimisation) Predicted pKd (After charmm minimisation) Experimental pKd (Mekenyan & Bradbury)32, 5 Metribolone (parent from pdb) 7.41 6.22 3.00 Mibolerone 7.82 8.14 3.00 5a dihydrotestosterone 7.67 8.37 2.30 4 pKd (exptal) ANDROGEN-RECEPTOR LIGANDS 2 0 6.5 -2 7 7.5 -4 7.15 7.42 0.96 7.08 7.96 0.70 Progesterone 6.65 7.82 -2.40 17a hydroxyprogesterone 6.80 7.53 -2.40 Corticosterone 6.58 7.28 -2.70 Pregnenolone 6.65 7.68 -2.70 Testosterone 7.53 7.36 1.82 Androstenedione 7.38 7.89 0.60 y = 5.0008x - 35.589 2 R = 0.8959 pKd (in silico) GRAPH OF pKd (EXPERIMENTAL) vs pKd (IN SILICO) (MINIMISED) 4 pKd (exptal) Oestradiol 5a androstane 8 3 2 1 0 -1 6 6.5 7 7.5 8 8.5 -2 -3 pKd (in silico) y = -0.2023x + 1.7373 2 R = 0.0025 5 SUMMARY RESULTS…. Oestrogen Receptor Ligands • No linear correlations were observed when the experimental binding affinity for the ligand set was plotted against its in silico counterpart for either unminimised or minimised ligands, even when these were grouped according to position on which substitution was made Androgen Receptor Ligands • A linear correlation was observed when the experimental binding affinity for the ligand set was plotted against its in silico counterpart for the unminimised ligands • Minimisation, irrespective of the protocol utilised, disrupted the linear relationship PART 2 NON STEROIDAL ANDROGEN RECEPTOR LIGANDS • Literature based evidence indicates that a substituted phenyl ring, equivalent to the A ring of the steroid structure, is an essential feature that acts as an anchor to the molecular recognition site of the androgen receptor 18 12 19 11 1 2 • Consequently superimposition of the substituted phenyl group on the A ring of a docked steroid molecule, would constitute an ideal starting point for the docking of non steroidal molecules CH3 10 3 4 5 9 CH3 17 13 8 14 16 15 7 6 PART 2 NON STEROIDAL ANDROGEN RECEPTOR LIGANDSMapping the Active Site of the Androgen Receptor • a map of the amino acid perimeter of the active site was generated in Sybyl, using the ligand- protein contacts in pdb ID 1e3g PART 2 NON STEROIDAL ANDROGEN RECEPTOR LIGANDS Main issues to address 1. What binding modality that should be adopted in the docking of the non steroidal ligands to the active site? 2. No non steroidal co-ordinates available from the pdb 3. Also there are significant structural variations between the non-steroidal ligands known to bind the androgen receptor PART 2 NON STEROIDAL ANDROGEN RECEPTOR LIGANDSMapping the Active Site of the Androgen Receptor •a map of the active site indicating polar and hydrophobic sites was generated using Ligbuilder, with testosterone bound into the androgen receptor (pdb 1E3G) PART 2 NON STEROIDAL ANDROGEN RECEPTOR LIGANDSMapping the Active Site of the Androgen Receptor • all amino acids except Thr 877 Asn 705 Gln 711 and Arg 752 were deleted • These amino acids were indicated in the literature as being of fundamental importance for binding as proved by site directed mutagenesis • the distances between the H bond sites of the amino acids and the corresponding H bond sites on testosterone were measured. This was done in order to try to maintain these distances when the non steroidal ligands were to be docked. 6 PART 2 Distance between the H bond Sites of the Critical Amino Acids and the Corresponding H bond Sites on Testosterone Thr 877: Gln 711: OG1 of Thr and H44 of testosterone: 1.927A H39 of Thr and O43 of testosterone: 3.552A H17 of Gln and O49 of testosterone: 2.660A H18 of Gln and O49 of testosterone: 4.581A Arg 752: H33 of Arg and O49 of testosterone: 2.660A H32 of Arg and O49 of testosterone: 2.406A Asn 705: ND2 of Thr and O43 of testosterone: 4.405A PART 2 NON STEROIDAL ANDROGEN RECEPTOR LIGANDSMapping the Active Site of the Androgen Receptor PART 2 Proposed Orientation for Hydroxyflutamide.. • The best representation of hydroxyflutamide in its docked form was achieved with the following superimposition methodology: • Superimposition of the nitro group of hydroxyflutamide onto the 3 keto group of testosterone to preserve the H bonding between the 3 keto moiety and Gln 711 and Arg 752. • Superimposition of the CF3 group of hydroxyflutamide onto C4 of testosterone so that it would occupy the hydrophobic region and also the same plane as C4, 5 & 6 of testosterone. • Superimposition of the hydroxyl group a to the carbonyl in hydroxyflutamide onto the 17 OH of testosterone in order to preserve H bonding with Asn 705 and Thr 877. PART 2 Proposed Orientation for Hydroxyflutamide.. • In this orientation for hydroxyflutamide the substituted ring occupies approximately the same spatial orientation as does the A ring of the steroid • The distances between the anchored regions of the hydroxyflutamide ligand and its H bond partners on amino acids Thr 877 Asn 705 Gln 711 and Arg 752 were comparable with those of testosterone • This exercise was applied to all ligands forming part of the test set TESTOSTERONE HYDROXYFLUTAMIDE • The training set ligands were then superimposed directly onto the test set ligand that was most similar to it in structure Finding the optimum docked spatial orientation for hydroxyflutamide Based on the binding modality of testosterone PART 2 The Training Set pp’-DDT op’-DDT pp’-DDE PART 2 The Training Set Hydroxyflutamide Vinclozolin Procymidone Linuron Methoxychlor Diethylstilboestrol 7 • (2-{[(3,5-dichlorophenyl)carbamoyl]oxy}-2-methyl-3butenoic acid • 3’,5’-dichloro-2-hydroxy-2-methylbut-3-en anilide • Hydroxylated analogue of PCB 153 • 2,2-bis(p-hydroxyphenyl)-1,1,1-trichloroethane • 3,5-dichlorobenzanilide-2-cyclopropanecarboxylic acid • PCB 153 • Hydroxylinuron Results Androgen Receptor Ligands- Training & Test Sets.. Graph of pKd (exptal) vs pKd (in silico) 0.5 R2 = 0.7867 0 Hydroxyflutamide -0.5 3 pKd (exptal) PART 2 The Test Set 4 5 p,p-DDE 6 p,p-DDT o,p-DDT -1.5 M1 P1 Hydroxy PCB -2 -2.5 Hpte 7 DES -1 Linuron Training Set Test Set Linear (Series1) M2 Hydroxylinuron M ethoxychlor Vinclozolin PCB Procymidone -3 -3.5 pKd in silico CONCLUSIONS… CONCLUSIONS… • The experimental ligand binding affinity data were obtained from literature. These were essentially reports of in vitro competitive assay studies • Waller et al., carried out a competitive binding assay study using [3H] R1881 radiolabelled synthetic androgen • Waller’s study used a series of steroidal and non-steroidal androgen receptor ligands. • Waller’s study is valuable for its compilation of binding data generated in one laboratory under consistent conditions, thereby minimising the contribution of biological variability to uncertainty in the modelling results • The results of the androgen receptor ligand assays were obtained from the paper by Waller, while those of the oestrogen receptor ligand binding assays were obtained from the paper of Gao et al. CONCLUSIONS… CONCLUSIONS… • Gao et al’s study was a broad based investigation of the interaction between oestrogen ligands and their allied receptor. For each substitution type considered, a QSAR equation was derived based on the interaction between the specific ligand type and its receptor. • This paper contains experimental ligand binding affinity data that was obtained from various sources • The data presented by Gao et al. was taken from different sources, & placed on a common Relative Binding Affinity (RBA) scale with oestradiol, by definition, was assigned a value of 100 (log RBA =2), with lower and higher affinity ligands than oestradiol having lower and higher values than this benchmark level • The diverse sources could potentially detract from the consistency that it would have had it all been generated in the same laboratory • The fact that different oestrogen receptor preparations (coming from mouse, rat, lamb and calf, uterine cytosol) were used in the estimation of the ligand binding affinities of the oestrogen receptor ligands could also potentially detract from the “relative consistency” of the data 8 CONCLUSIONS… CONCLUSIONS… • The temperature at which the competitive binding affinity assays were carried out is perhaps a far more important issue. Traditionally, competitive binding affinity assays are sometimes run at 0oC and at 25oC. Purportedly, carrying out the assays at 0oC preserves receptor stability, while carrying out assays at 25oC allows for more rapid equlibration of the binding process for both competing ligand and labelled tracer • In fact, the RBA values presented by Gao et al at either temperature are different, probably due to incomplete assay equilibration at the lower temperature • This problem was more evident with higher affinity ligands, for which association and dissociation rates may be slower • This latter issue gains importance given that most of the RBA data presented was derived from biological assays carried out at 0oC. • In the case of the androgen receptor ligands, linear relationships between experimental and in silico pKd and hence reliable correlation tools were established for both unminimised steroidal and unminimised nonsteroidal ligands • Two issues must still be pointed out however: • All these reasons could account for the fact that no linear correlations between experimental and in silico pKd could be established for the oestrogen receptor ligands CONCLUSIONS… 1. 2. Minimisation did not enhance or even retain the linear relationships for both the steroidal and the non steroidal ligands. The implication is that the binding conformation of these ligands is not their lowest energy state The traces for the steroidal and the non-steroidal ligands were not super-imposable. This implies that although both categories have an affinity for the androgen receptor, the binding modality is different- a fact that may be confirmed through an investigation of the dynamics of the androgen receptor bound to steroidal and non-steroidal ligands respectively REFERENCES • Bradbury SP, Mekenyan OG. The Role of Ligand Flexibility in Predicting Biological Activity: Structure-Activity Relationships for Aryl Hydrocarbon, Estrogen and Androgen Receptor Binding Affinity. Environmental Toxicology and Chemistry 1998 17(1): 15-25. • Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathan S, Karplus M. CHARMM: A Program for Macromolecular Energy, Minimization, and Dynamics Calculations. J. Comp. Chem. 1983 4: 187-217. • Matias PM, Donner P, Coehlo R, Thomaz M, Peixoto C, Macedo S Otto N Joschko S, Scholz P, Wegg A, Basler S, Schafer M, Egner U, Carrondo MA. Structural evidence for ligand specificity in the binding domain of the human androgen receptor. Implications for pathogenic gene mutations. J.Biol.Chem. 2000 275: 26164-26171 • Mekenyan OG, Ivanov J. A Computationally- Based Hazard Identification Algorithm That Incorporates Ligand Flexibility. 1. Identification of Potential Androgen Receptor Ligands. Environ. Sci. Technol. 1997 31: 3702-3711 • Tanenbaum DM, Wang Y, Williams SP Sigler PB Crystallographic Comparison of the Estrogen and Progesterone Receptor's Ligand Binding Domains. Proc.Natl.Acad.Sci.USA 1998 95: 5998-6003 , • Waller CL, Juma,W. Three-Dimensional Structure-Activity Relationships for Androgen Receptor Ligands Toxicology and Applied Pharmacology 1996 137: 219-227. • Wang R, Lai L, Wang, S. Further Development and Validation of Empirical Scoring Functions for Structure- Based Binding Affinity Prediction. J.Comput. Aided Mol. Des. 2002 16: 11-26 • Wang R, Gao Y, Lai L. Ligbuilder: A Multipurpose Program for Structure-Based Drug Design. J. Mol. Modeling 2000 6: 498-516 9