Current Computer - Aided Drug Design - Volume 8, Issue 1, 2012

Chemobioinformatics: The Advancing Frontier of Computer-Aided Drug Design in the Post-Genomic Era Modern drug discovery is a highly expensive process, the cost of discovering one new drug ranging from $400 million to $2 billion. Drug design usually starts with the isolation of “lead” compounds, found by different approaches which include ethnopharmacology, natural product chemistry, screening of chemical libraries, and serendipity, to name just a few. A small fraction of such leads ends up in the pharmacist's desk. Between the chemist's laboratory and the bedside of the patient drug candidates undergo a complex series of chemical/biological/toxicological/clinical evaluations which is responsible for the staggering cost of drug discovery. The availability of high quality leads is critical in the early stages of pharmaceutical drug design. One of the crucial factors contributing to the escalating cost is that the drug developer has to produce and test a large number of derivatives of the lead structures for their beneficial and toxic effects before one useful drug candidate is identified. Computational evaluation of chemicals plays an indispensable role in modern drug design because they are cheaper and rapid alternatives to the medium throughput in vitro and low throughput in vivo bioassays which are generally reserved for later stages of discovery. The basic assumption of many computer aided drug design techniques is that beneficial or toxic effects of molecules, the result of ligand-biotarget interactions, can be expressed by the following relationship: BR = f (S, B) Eq (1) In Eq (1), BR represents the magnitude of pharmacological/toxicological effects produced by the chemical in the organism, and B represents the relevant biological target that is perturbed by the ligand resulting in the measurable response. It is believed that a major determinant of BR is the structure (S) of the ligand. The structure becomes the sole determinant of the variation of BR from molecule to molecule when the biological target, B, remains practically the same and there is alternation only in the structure of the ligand. Under such circumstances Eq 1 approximates to: BR = f (S) Eq. (2) In 1868, Crum-Brown and Fraser first stated that the constitution of permanently charged quaternary compounds determined their “physiological activity”. About two decades later, Richet (1893) observed that the toxicity of ethers, aldehydes, alcohols, ketones, and other organic compounds was inversely related to their aqueous solubility. Hammet's (1940) electronic descriptor sigma (σ) and Taft's (1952) steric parameter were important contributions in this line of research. The multiparamater linear free energy related (LFER) approach formulated by Hansch and Fujita (1964) was an outgrowth of this notion where combinations of different physicochemical properties or substituent constants were applied in models for the prediction of bioactivity of chemicals, particularly those belonging to congeneric classes. Models developed for congeneric sets of chemicals served us reasonably well initially. In practical drug design, however, one has to deal with large and structurally diverse (noncongeneric) collections of molecules for which not many experimental properties are available. The property-property relationship (PPR) approaches, exemplified by the LFER methodology, have limited applicability in such situations. A viable alternative is the use of calculated properties which can be computed directly from molecular structure without the input of any other experimental data. Topological, geometrical (3-D), and quantum chemical molecular descriptors belong to this group. For large sets of molecules, calculation of high level quantum chemical descriptors could be prohibitively costly in terms of computer resources. On the other hand, chemodescriptors derived from graph theoretic models of molecules have found applications in wide ranging computational models for new drug design. Long term administration of drugs in the management of diseases like malaria results in the development of resistance in the bugs leading to the diminished effectiveness of life saving drugs. Computed structural descriptors have been used to develop differential quantitative structure-activity relationship (DiQSAR) models to characterize the evolving differential chemical sensitivities of wild versus mutant varieties of drug targets. In the post-genomic era, the omics (genomics, proteomics, and metabolomics) technologies have generated a tremendous amount of data and brought fresh perspectives in our understanding chemical-biological interactions. New approaches have been used to formulate novel mathematical biodescriptors which have been applied to characterize complex proteomics patterns and nucleic acid sequences in an effort to understand pharmacological and toxicological effects of chemical substances as well as pathogenicity of organisms like pandemic Bird Flu (H5N1). Initial research on the combined use of chemodescriptors and biodescriptors shows that neither class of descriptors alone can explain bioactivity of molecules completely. We need both classes of descriptors to rationalize the diverse biological situations in pharmacology and toxicology. As evident from the above, we have come a long way from the initial insight that Crum-Brown and Fraser put forward in the third quarter of the nineteenth century. Right now we are in the middle of an intense intellectual fermentation which may be termed as CHEMOBIOINFORMATICS, the emerging frontier resulting from the fusion of chemoinformatics and bioinformatics. A perusal of publications in Current Computer-Aided Drug Design would testify that this journal in particular has been at the forefront of publishing high quality reviews and research articles by leading scientists working at the interface of chemobioinformatics and new drug discovery.

In the study, 18 antiepileptic hydantoin analogues were investigated by means of reversed-phase HPLC on C- 18 stationary phase and eluent acetonitrile-water. Quantitative structure-retention relationship (QSRR) study has been applied in order to understand factors that affect the retention which is closely correlated to the activity (ED50 values). To overview the compounds for similarities and dissimilarities principal component analysis (PCA) has been applied. Six multiple linear regression models based on the most relevant descriptors were developed. Descriptors for MLR were selected according to variable importance calculated by partial least squares (PLS) analysis. Besides ALOGP the most important is aromatic ratio for mobile phases with more than 45% of acetonitrile, as well as electrotopological states when the % of acetonitrile is less than 40%. High agreement between experimental and predicted retention, obtained in the validation procedure, indicated the good quality of the derived QSRR models. For individual linear models, crossvalidation squared correlation coefficients (Q2) ranging from 0.697 to 0.837 were obtained. The residual values (difference between observed and calculated) agreed well within experimental error. Additionally, models were compared in terms of the smallest residual value by recently developed method of ranking based on the sum of ranking differences (SRD).

Human Immunodeficiency Virus (HIV)-1 protease is one of the key targets for Acquired Immunodeficiency Syndrome (AIDS). A large number of inhibitors are being designed for this target with the focus towards interactions with backbone atoms to combat drug resistance. In the present study, we have developed QSAR models for 99 inhibitors which include P1/P1' and P2/P2' substituents with diverse scaffolds. In the present work, HIV-1 protease inhibitors dataset with tanimoto similarity of 0.7 were compiled from The Binding Database (Binding DB). Multiple linear regression analysis was performed to compute the relationship between 2D and 3D structure descriptors and binding affinity. Untransformed and non-linearly transformed descriptors were used for the QSAR model development. Transformation of descriptors resulted in better QSAR model (r2=0.77) compared to the model developed using untransformed descriptors (r2=0.74). Molecular connectivity, cosmic bond angle energy and charged based descriptor were reported as a priori properties in the prediction of binding affinity. The developed models were validated using an external test set and r2 test values of 0.73 and 0.72 were obtained. Models developed in this study have potential application in the prediction of binding affinity for the newly synthesized compounds.

The transporter associated with antigen processing (TAP) is essential for peptide delivery from the cytosol into the lumen of the endoplasmic reticulum (ER), where these peptides are loaded on a major histocompatibility complex (MHC) I molecules and form peptide-MHC complex. The peptide-MHC leaves the ER and displays their antigenic cargo on the cell surface to cytotoxic T cells. In this study, 89 physicochemical properties of amino acid were collected from AAIndex database, and used to characterize the peptides which were binding to TAP. Then, the stepwise regression (STR) was used to optimize the parameters which characterized the TAP binding peptides, and the multiple linear regression (MLR) was used to construct the quantitative structural activity relationship (QSAR) model based on optimized parameters. The quantitative models had good reliability and predictive ability: the Q2 of “leave one out” validation is 0.676 and R2 of test dataset is 0.722 respectively. Additionally, the standardized coefficients of the models could demonstrate the attributions for each position of epitope and determine which special amino acid is suitable at any position of the peptide. Therefore, the QSAR model constructed by STR-MLR has many advantages, such as, easier calculation and explanation, good performance, and definite physiochemical indication, which could be used to guide the design and modification of the TAP binding peptide.

A novel computational technology based on fragmentation of the chemical compounds has been used for the fast and efficient prediction of activities of prospective protease inhibitors of the hepatitis C virus. This study spans over a discovery cycle from the theoretical prediction of new HCV NS3 protease inhibitors to the first cytotoxicity experimental tests of the best candidates. The measured cytotoxicity of the compounds indicated that at least two candidates would be suitable further development of drugs.

A major concern in the development of acetyl-CoA carboxylase-inhibiting (ACCase; EC 6.4.1.2) herbicides is the emergence of resistance as a result of the selection of distinct mutations within the CT domain. Mutations associated with resistance have been demonstrated to include both active sites and non-active sites, including Ile-1781-Leu, Trp- 2027-Cys, Ile-2041-Asn, Asp-2078-Gly, and Gly-2096-Ala (numbered according to the Alopecurus myosuroides plastid ACCase). In the present study, extensive computational simulations, including molecular dynamics (MD) simulations and molecular mechanics-Poisson-Boltzmann surface area (MM/PBSA) calculations, were carried out to compare the molecular mechanisms of active site mutation (I388N) and non-active site mutation (D425G) in Alopecurus myosuroides resistance to some commercial herbicides targeting ACCase, including haloxyfop (HF), diclofop (DF) and fenoxaprop (FR). All of the computational model and energetic results indicated that both I388N and D425G mutations have effects on the conformational change of the binding pocket. The π-π interaction between ligand and Phe377 and Tyr161' residues, which make an important contribution to the binding affinity, was decreased after mutation. As a result, the mutant-type ACCase has a lower affinity for the inhibitor than the wild-type enzyme, which accounts for the molecular basis of herbicidal resistance. The structural and mechanistic insights obtained from the present study will deepen our understanding of the interactions between ACCase and herbicides, which provides a molecular basis for the future design of a promising inhibitor with low resistance risk.

Lysine sulfonamide and its structural analogs, a class of Human Immunodeficiency Virus protease inhibitors has gained importance in recent years due to its mode of action. QSAR analysis for multiple ligand-receptor complexes can be performed using binding interaction energies derived from the molecular dynamics simulations. A Receptordependent QSAR (RD-QSAR) analysis was carried out for 65 lysine sulfonamide analogs complexed with HIV-protease using Prime Molecular Mechanics Generalized Born Surface Area (MM-GBSA) method. The lysine sulfonamide analogs were docked in the receptor active site and the obtained complexes were further rescored using Prime MM-GBSA method. The descriptors, docking score and MM-GBSA free energy of binding were used to derive a relationship with biological inhibition constant. The influence of individual free energy components on biological activity and the effect of structural flexibility (in terms of strain energies) over the prediction model has been studied using two models (with and without strain energy), built using forward entry MLR method. Inclusion of strain energies enhanced the effect of all the free energy of binding components and hence reflects their importance. The statistical significance of the derived QSAR models was described using the parameters, r2, F-test and RMSE. A test set of 11 compounds was used to ensure the predictability of the models (q2 cv, PRESS). Results from this study would be useful in identifying anti-virals with energetically favorable interactions.