Current Pharmaceutical Design - Volume 13, Issue 14, 2007

In recent years, biomedical research and drug design became one of the fastest growing branches of scientific and industrial development. The tremendous scale (number of projects and volume of information to process) promoted by trillions of dollars invested in these fields called for completely new approaches to obtain, analyze, and apply information to clinical practice, pharmacological research and the medical industry. This issue was put together as a result of our long-standing interest in the various aspects of research and drug design. It is dedicated to the use of new mathematical models in various fields of medical research. Our previous Current Pharmaceutical Design issue (2004) addressed the concept of multi-functional drug targets in diverse model systems [1]. This issue, accordingly, continues our inquiry into various types of models used in research and the subsequent creation of novel agents and improved therapies. This issue aims to give the reader an inside view into the concepts of intelligent research design and biomedical information processing. The first three reviews [2-4] are orientated on the use of mathematics in basic science research to uncover the processes underlining the most complex events that are so crucial to understand in order to successfully conduct medical research, design drugs and simply practice medicine nowadays. In the first review Yang and Hamer [2] discuss an important topic in bioinformatics and systems biology - identifying functional sites in proteins. They focus on the variants of Bio-basis Function Neural Networks (BBFNN) and their applications in mining protein sequence data. Next, Goutsias and Lee [3] discuss four gene regulatory networks models: gene networks, transcriptional regulatory systems, Boolean networks, and dynamical Bayesian networks. The authors review state-of-theart functional genomics techniques, such as gene expression profiling, cis-regulatory element identification, TF target gene identification, and gene silencing by RNA interference, which can be used to extract information about gene regulation. In the third review by Qazi, Chamberlin, and Nigam [4], the authors describe the use of the difference method to investigate the information processing capabilities of GABAA receptors and predict how pharmacological agents may modify these properties. They suggest that understanding this process of transmitter-receptor interactions may be useful in the development of more specific and highly targeted modes of action. The next three reviews [5-9] are dedicated to predicting capabilities of using mathematical models in biomedical research and in medicine, such as the prediction of drug delivery efficiency or patient treatment outcome. First, Panteleev et al. discuss the use of mathematical models of blood coagulation and platelet-mediated primary hemostasis and thrombosis in clinical practice, research and drug development [5]. Micheli, Sperduti and Starita [6] introduce the reader to new developments in neural networks and Kernel machines concerning the treatment of structured domains. Focusing more on the computational side than on the experimental one, they discuss the research on these relatively new models to introduce a novel and more general approach to QSPR/QSAR analysis. Artificial intelligence approaches for rational drug design is then reviewed by Duch, Swaminathan, and Meller [7]. A special emphasis is made on methods that “enable an intuitive interpretation of the results” and facilitate gaining an insight into the nature of the problem. This discussion is continued in the next issue We would like to thank all the authors for their contributions and hope that this two part-issue will stimulate new communication and collaborations. References [1] Current Pharmaceutical Design, Volume 10, Number 15, June 2004. [2] Yang ZR, Hamer R. Bio-basis Function Neural Networks in Protein Data Mining. Curr Pharm Des 2007; 13(14): 1403-1413. [3] Goutsias J, Leeb NH. Computational and Experimental Approaches for Modeling Gene Regulatory Networks. Curr Pharm Des 2007; 13(14): 1415-1436. [4] Qazi S, Caberlin M, Nigam N. Mechanism of psychoactive drug action in the brain: Simulation modeling of GABAA receptor interactions at non-equilibrium conditions. Curr Pharm Des 2007; 13(14): 1437-1455. [5] Panteleev MA, Ananyeva NM, Radtke K-P, Ataullakhanov FI, and Saenko EL, Mathematical models of blood coagulation and platelet adhesion: clinical application. Curr Pharm Des 2007; 13(14): 1457-1467. [6] Micheli A, Sperduti A, Starita A. An Introduction to Recursive Neural Networks and Kernel Methods for Cheminformatics. Curr Pharm Des 2007; 13(14): 1469-1495. [7] Duch W, Swaminathan K, Meller J. Artificial Intelligence Approaches for Rational Drug Design and Discovery. Curr Pharm Des 2007; 13(14): 1497-1508.

Accurately identifying functional sites in proteins is one of the most important topics in bioinformatics and systems biology. In bioinformatics, identifying protease cleavage sites in protein sequences can aid drug/inhibitor design. In systems biology, post-translational protein-protein interaction activity is one of the major components for analyzing signaling pathway activities. Determining functional sites using laboratory experiments are normally time consuming and expensive. Computer programs have therefore been widely used for this kind of task. Mining protein sequence data using computer programs covers two major issues: 1) discovering how amino acid specificity affects functional sites and 2) discovering what amino acid specificity is. Both need a proper coding mechanism prior to using a proper machine learning algorithm. The development of the bio-basis function neural network (BBFNN) has made a new way for protein sequence data mining. The bio-basis function used in BBFNN is biologically sound in well coding biological information in protein sequences, i.e. well measuring the similarity between protein sequences. BBFNN has therefore been outperforming conventional neural networks in many subjects of protein sequence data mining from protease cleavage site prediction to disordered protein identification. This review focuses on the variants of BBFNN and their applications in mining protein sequence data.

To understand most cellular processes, one must understand how genetic information is processed. A formidable challenge is the dissection of gene regulatory networks to delineate how eukaryotic cells coordinate and govern patterns of gene expression that ultimately lead to a phenotype. In this paper, we review several approaches for modeling eukaryotic gene regulatory networks and for reverse engineering such networks from experimental observations. Since we are interested in elucidating the transcriptional regulatory mechanisms of colon cancer progression, we use this important biological problem to illustrate various aspects of modeling gene regulation. We discuss four important models: gene networks, transcriptional regulatory systems, Boolean networks, and dynamical Bayesian networks. We review state-of-the-art functional genomics techniques, such as gene expression profiling, cisregulatory element identification, TF target gene identification, and gene silencing by RNA interference, which can be used to extract information about gene regulation. We can employ this information, in conjunction with appropriately designed reverse engineering algorithms, to construct a computational model of gene regulation that sufficiently predicts experimental observations. In the last part of this review, we focus on the problem of reverse engineering transcriptional regulatory networks by gene perturbations. We mathematically formulate this problem and discuss the role of experimental resolution in our ability to reconstruct accurate models of gene regulation. We conclude, by discussing a promising approach for inferring a transcriptional regulatory system from microarray data obtained by gene perturbations.

Synaptic transmission requires that the binding of the transmitter to the receptor to occur under rapidly changing transmitter levels, and this binding interaction is unlikely to be at equilibrium. We have sought to numerically solve for binding kinetics using ordinary differential equations and simultaneous difference equations for use in stochastic conditions. The reaction scheme of GABA interacting with the ligand-gated ion-channel demonstrates numerical stiffness. Implicit methods (Backward Euler, ode23s) performed orders of magnitude better than explicit methods (Forward Euler, ode23, RK4, ode45) in terms of step size required for stability, number of steps and cpu time. Interestingly, upon solving the system of 8 ordinary differential equations for the GABA reaction scheme we observed the existence of low dimensional invariant manifolds that may have important consequences for information processing in synapses. We also describe a mathematical approach that models complex receptor interactions in which the timing and amplitude of transmitter release are noisy. Exact solutions for simple bimolecular interactions that include stoichiometric interactions and receptor transitions can be used to model complex reaction schemes. We used the difference method to investigate the information processing capabilities of GABAA receptors and to predict how pharmacological agents may modify these properties. Initial simulations using a model for heterosynaptic regulation shows that signal to noise ratios can be decreased in the presence of background presynaptic activity both in the presence and absence of chlorpromazine. These types of simulations provide a platform for investigating the effect of psycho-active drugs on complex responses of transmitter-receptor interactions in noisy cellular environments such as the synapse. Understanding this process of transmitter-receptor interactions may be useful in the development of more specific and highly targeted modes of action.

At present, computer-assisted molecular modeling and virtual screening have become effective and widelyused tools for drug design. However, a prerequisite for design and synthesis of a therapeutic agent is determination of a correct target in the metabolic system, which should be either inhibited or stimulated. Solution of this extremely complicated problem can also be assisted by computational methods. This review discusses the use of mathematical models of blood coagulation and platelet-mediated primary hemostasis and thrombosis as cost-effective and time-saving tools in research, clinical practice, and development of new therapeutic agents and biomaterials. We focus on four aspects of their application: 1) efficient diagnostics, i.e. theoretical interpretation of diagnostic data, including sensitivity of various clotting assays to the changes in the coagulation system; 2) elucidation of mechanisms of coagulation disorders (e.g. hemophilias and thrombophilias); 3) exploration of mechanisms of action of therapeutic agents (e.g. recombinant activated factor VII) and planning rational therapeutic strategy; 4) development of biomaterials with non-thrombogenic properties in the design of artificial organs and implantable devices. Accumulation of experimental knowledge about the blood coagulation system and about platelets, combined with impressive increase of computational power, promises rapid development of this field.

The aim of this paper is to introduce the reader to new developments in Neural Networks and Kernel Machines concerning the treatment of structured domains. Specifically, we discuss the research on these relatively new models to introduce a novel and more general approach to QSPR/QSAR analysis. The focus is on the computational side and not on the experimental one.

Pattern recognition, machine learning and artificial intelligence approaches play an increasingly important role in rational drug design, screening and identification of candidate molecules and studies on quantitative structure-activity relationships (QSAR). In this review, we present an overview of basic concepts and methodology in the fields of machine learning and artificial intelligence (AI). An emphasis is put on methods that enable an intuitive interpretation of the results and facilitate gaining an insight into the structure of the problem at hand. We also discuss representative applications of AI methods to docking, screening and QSAR studies. The growing trend to integrate computational and experimental efforts in that regard and some future developments are discussed. In addition, we comment on a broader role of machine learning and artificial intelligence approaches in biomedical research.