A Bayesian molecular interaction library

Tailored Approaches in Drug Development and Diagnostics: From Molecular Design to Biological Model Systems

Approaches to increase the efficiency in developing drugs and diagnostics tools, including new drug delivery and diagnostic technologies, are needed for improved diagnosis and treatment of major diseases and health problems such as cancer, inflammatory diseases, chronic wounds, and antibiotic resistance. Development within several areas of research ranging from computational sciences, material sciences, bioengineering to biomedical sciences and bioimaging is needed to realize innovative drug development and diagnostic (DDD) approaches. Here, an overview of recent progresses within key areas that can provide customizable solutions to improve processes and the approaches taken within DDD is provided. Due to the broadness of the area, unfortunately all relevant aspects such as pharmacokinetics of bioactive molecules and delivery systems cannot be covered. Tailored approaches within (i) bioinformatics and computer-aided drug design, (ii) nanotechnology, (iii) novel materials and technologies for drug delivery and diagnostic systems, and (iv) disease models to predict safety and efficacy of medicines under development are focused on. Current developments and challenges ahead are discussed. The broad scope reflects the multidisciplinary nature of the field of DDD and aims to highlight the convergence of biological, pharmaceutical, and medical disciplines needed to meet the societal challenges of the 21st century.

Enhance the performance of current scoring functions with the aid of 3D protein-ligand interaction fingerprints

BACKGROUND

In structure-based drug design, binding affinity prediction remains as a challenging goal for current scoring functions. Development of target-biased scoring functions provides a new possibility for tackling this problem, but this approach is also associated with certain technical difficulties. We previously reported the Knowledge-Guided Scoring (KGS) method as an alternative approach (BMC Bioinformatics, 2010, 11, 193-208). The key idea is to compute the binding affinity of a given protein-ligand complex based on the known binding data of an appropriate reference complex, so the error in binding affinity prediction can be reduced effectively.

RESULTS

In this study, we have developed an upgraded version, i.e. KGS2, by employing 3D protein-ligand interaction fingerprints in reference selection. KGS2 was evaluated in combination with four scoring functions (X-Score, ChemPLP, ASP, and GoldScore) on five drug targets (HIV-1 protease, carbonic anhydrase 2, beta-secretase 1, beta-trypsin, and checkpoint kinase 1). In the in situ scoring test, considerable improvements were observed in most cases after application of KGS2. Besides, the performance of KGS2 was always better than KGS in all cases. In the more challenging molecular docking test, application of KGS2 also led to improved structure-activity relationship in some cases.

CONCLUSIONS

KGS2 can be applied as a convenient "add-on" to current scoring functions without the need to re-engineer them, and its application is not limited to certain target proteins as customized scoring functions. As an interpolation method, its accuracy in principle can be improved further with the increasing knowledge of protein-ligand complex structures and binding affinity data. We expect that KGS2 will become a practical tool for enhancing the performance of current scoring functions in binding affinity prediction. The KGS2 software is available upon contacting the authors.

Landscape of protein-small ligand binding modes

Elucidating the mechanisms of specific small-molecule (ligand) recognition by proteins is a long-standing conundrum. While the structures of these molecules, proteins and ligands, have been extensively studied, protein-ligand interactions, or binding modes, have not been comprehensively analyzed. Although methods for assessing similarities of binding site structures have been extensively developed, the methods for the computational treatment of binding modes have not been well established. Here, we developed a computational method for encoding the information about binding modes as graphs, and assessing their similarities. An all-against-all comparison of 20,040 protein-ligand complexes provided the landscape of the protein-ligand binding modes and its relationships with protein- and chemical spaces. While similar proteins in the same SCOP Family tend to bind relatively similar ligands with similar binding modes, the correlation between ligand and binding similarities was not very high (R(2)  = 0.443). We found many pairs with novel relationships, in which two evolutionally distant proteins recognize dissimilar ligands by similar binding modes (757,474 pairs out of 200,790,780 pairs were categorized into this relationship, in our dataset). In addition, there were an abundance of pairs of homologous proteins binding to similar ligands with different binding modes (68,217 pairs). Our results showed that many interesting relationships between protein-ligand complexes are still hidden in the structure database, and our new method for assessing binding mode similarities is effective to find them.

Comprehensive classification and diversity assessment of atomic contacts in protein-small ligand interactions

Elucidating the molecular mechanisms of selective ligand recognition by proteins is a long-standing problem in drug discovery. Rapid increase in the availability of three-dimensional protein structural data indicates that a data-driven approach for finding the rules that govern protein-ligand interactions is increasingly attractive. However, this approach is not straightforward because of the complexity of molecular interactions and our inadequate understanding of the diversity of molecular interactions that occur during ligand recognition. Thus, we aimed to provide a comprehensive classification of the spatial arrangements of ligand atoms based on the local coordinates of each interacting "protein fragment" consisting of three atoms with covalent bonds in each amino acid. We used a pattern recognition technique based on the Gaussian mixture model and found 13,519 patterns in the spatial arrangements of interacting ligand atoms, each of which was described as a Gaussian function of the local coordinates. Some typical well-known interaction patterns such as hydrogen bonds were ubiquitous in several hundred protein families, whereas others were only observed in a few specific protein families. After removing protein sequence redundancy from the data set, we found that 63.4% of ligand atoms interacted via one or more interaction patterns and that 25.7% of ligand atoms interacted without patterns, whereas the remainder had no direct interactions. The top 3115 major patterns included 90% of the interacting pairs of residues and ligand atoms with patterns, while the top 6229 included all of them.

Probabilistic prediction of contacts in protein-ligand complexes

We introduce a statistical method for evaluating atomic level 3D interaction patterns of protein-ligand contacts. Such patterns can be used for fast separation of likely ligand and ligand binding site combinations out of all those that are geometrically possible. The practical purpose of this probabilistic method is for molecular docking and scoring, as an essential part of a scoring function. Probabilities of interaction patterns are calculated conditional on structural x-ray data and predefined chemical classification of molecular fragment types. Spatial coordinates of atoms are modeled using a Bayesian statistical framework with parametric 3D probability densities. The parameters are given distributions a priori, which provides the possibility to update the densities of model parameters with new structural data and use the parameter estimates to create a contact hierarchy. The contact preferences can be defined for any spatial area around a specified type of fragment. We compared calculated contact point hierarchies with the number of contact atoms found near the contact point in a reference set of x-ray data, and found that these were in general in a close agreement. Additionally, using substrate binding site in cathechol-O-methyltransferase and 27 small potential binder molecules, it was demonstrated that these probabilities together with auxiliary parameters separate well ligands from decoys (true positive rate 0.75, false positive rate 0). A particularly useful feature of the proposed Bayesian framework is that it also characterizes predictive uncertainty in terms of probabilities, which have an intuitive interpretation from the applied perspective.

A PRIORI CONTACT PREFERENCES IN MOLECULAR RECOGNITION

A molecular interaction library modeling favorable non-bonded interactions between atoms and molecular fragments is considered. In this paper, we represent the structure of the interaction library by a network diagram, which demonstrates that the underlying prediction model obtained for a molecular fragment is multi-layered. We clustered the molecular fragments into four groups by analyzing the pairwise distances between the molecular fragments. The distances are represented as an unrooted tree, in which the molecular fragments fall into four groups according to their function. For each fragment group, we modeled a group-specific a priori distribution with a Dirichlet distribution. The group-specific Dirichlet distributions enable us to derive a large population of similar molecular fragments that vary only in their contact preferences. Bayes' theorem then leads to a population distribution of the posterior probability vectors referred to as a "Dickey–Savage"-density. Two known methods for approximating multivariate integrals are applied to obtain marginal distributions of the Dickey–Savage density. The results of the numerical integration methods are compared with the simulated marginal distributions. By studying interactions between the protein structure of cyclohydrolase and its ligand guanosine-5′-triphosphate, we show that the marginal distributions of the posterior probabilities are more informative than the corresponding point estimates.

Bayesian methods in virtual screening and chemical biology

The Naïve Bayesian Classifier, as well as related classification and regression approaches based on Bayes' theorem, has experienced increased attention in the cheminformatics world in recent years. In this contribution, we first review the mathematical framework on which Bayes' methods are built, and then continue to discuss implications of this framework as well as practical experience under which conditions Bayes' methods give the best performance in virtual screening settings. Finally, we present an overview of applications of Bayes' methods to both virtual screening and the chemical biology arena, where applications range from bridging phenotypic and mechanistic space of drug action to the prediction of ligand-target interactions.

Molecular evolution of adrenoceptors and dopamine receptors: implications for the binding of catecholamines

We derived homology models for all human catecholamine-binding GPCRs (CABRs; the alpha-1, alpha-2, and beta-adrenoceptors and the D1-type and D2-type dopamine receptor) using the bovine rhodopsin-11-cis-retinal X-ray structure. Interactions were predicted from the endogenous ligands norepinephrine or dopamine and from the binding site and were used to optimize receptor-ligand interactions. Similar binding modes in the complexes agree with a large "binding core" conserved across the CABRs, that is, D3.32, V(I)3.33, T3.37, S5.42, S(A/C)5.43, S5.46, F6.51, F6.52, and W6.48. Model structures and docking simulations suggest that extracellular loop 2 could provide a common attachment point for the ligands' beta-hydroxyl via a hydrogen bond donated by the main-chain NH group of residue xl2.52. The modeled CABRs and docking modes are in good agreement with published experimental studies. Complementarity between the ligand and the binding site suggests that the bovine rhodopsin structure is a suitable template for modeling agonist-bound CABRs.

Model structures of α-2 adrenoceptors in complex with automatically docked antagonist ligands raise the possibility of interactions dissimilar from agonist ligands

Antagonist binding to alpha-2 adrenoceptors (alpha2-ARs) is not well understood. Structural models were constructed for the three human alpha2-AR subtypes based on the bovine rhodopsin X-ray structure. Twelve antagonist ligands (including covalently binding phenoxybenzamine) were automatically docked to the models. A hallmark of agonist binding is the electrostatic interaction between a positive charge on the agonist and the negatively charged side chain of D3.32. For antagonist binding, ion-pair formation would require deviations of the models from the rhodopsin structural template, e.g., a rotation of TM3 to relocate D3.32 more centrally within the binding cavity, and/or creation of new space near TM2/TM7 such that antagonists would be shifted away from TM5. Thus, except for the quinazolines, antagonist ligands automatically docked to the model structures did not form ion-pairs with D3.32. This binding mode represents a valid alternative, whereby the positive charge on the antagonists is stabilized by cation-pi interactions with aromatic residues (e.g., F6.51) and antagonists interact with D3.32 via carboxylate-aromatic interactions. This binding mode is in good agreement with maps derived from a molecular interaction library that predicts favorable atomic contacts; similar interaction environments are seen for unrelated proteins in complex with ligands sharing similarities with the alpha2-AR antagonists.

BODIL: a molecular modeling environment for structure-function analysis and drug design

BODIL is a molecular modeling environment geared to help the user to quickly identify key features of proteins critical to molecular recognition, especially (1) in drug discovery applications, and (2) to understand the structural basis for function. The program incorporates state-of-the-art graphics, sequence and structural alignment methods, among other capabilities needed in modern structure-function-drug target research. BODIL has a flexible design that allows on-the-fly incorporation of new modules, has intelligent memory management, and fast multi-view graphics. A beta version of BODIL and an accompanying tutorial are available at http://www.abo.fi/fak/mnf/bkf/research/johnson/bodil.html.