David T. Manallack

Neural networks in drug discovery : have they lived up to their promise?

Over the last decade neural networks have become an efficient method for data analysis in the field of drug discovery. The early problems encountered with neural networks such as overfitting and overtraining have been addressed resulting in a technique that surpasses traditional statistical methods. Neural networks have thus largely lived up to their promise, which was to overcome QSAR statistical problems. The next revolution in QSAR will no doubt involve research into producing better descriptors used in these studies to improve our ability to relate chemical structure to biological activity. This review focuses on the applications of neural network methods and their development over the last five years.

The pK a Distribution of Drugs: Application to Drug Discovery

The acid-base dissociation constant (pKa) of a drug is a key physicochemical parameter influencing many biopharmaceutical characteristics. While this has been well established, the overall proportion of non-ionizable and ionizable compounds for drug-like substances is not well known. Even less well known is the overall distribution of acid and base pKa values. The current study has reviewed the literature with regard to both the proportion of ionizable substances and pKa distributions. Further to this a set of 582 drugs with associated pKa data was thoroughly examined to provide a representative set of observations. This was further enhanced by delineating the compounds into CNS and non-CNS drugs to investigate where differences exist. Interestingly, the distribution of pKa values for single acids differed remarkably between CNS and non-CNS substances with only one CNS compound having an acid pKa below 6.1. The distribution of basic substances in the CNS set also showed a marked cut off with no compounds ...Fingolimod (FTY720) is the first of a novel class: sphingosine 1-phosphate (S1P) receptor modulator and is currently in phase 3 clinical trials for multiple sclerosis (MS). FTY720 was first synthesized in 1992 by chemical modification of an immunosuppressive natural product, ISP-I (myriocin). ISP-I was isolated from the culture broth of Isaria sinclairii, a type of vegetative wasp that was an ‘eternal youth’ nostrum in traditional Chinese medicine. ISP-I is an amino acid having three successive asymmetric centers and some functionalities. We simplified the structure drastically to find a nonchiral symmetric 2-substitued-2-aminopropane-1,3-diol framework for an in vivo immunosuppressive activity (inhibition of rat skin allograft rejection test or prolonging effect on rat skin allograft survival) and finally discovered FTY720. During the course of the lead optimization process, we encountered an unexpected dramatic change of the mechanism of action with an in vivo output unchanged. Since it proved that FTY720 did not inhibit serine palmitoyltransferase that is the target enzyme of ISP-I, reverse pharmacological approaches have been preformed to elucidate that FTY720 is mainly phosphorylated by sphingosine kinease 2 in vivo and the phosphorylated drug acts as a potent agonist of four of the five G protein coupled receptors for S1P: S1P1, S1P3, S1P4 and S1P5. Evidence has accumulated that immunomodulation by FTY720-P is based on agonism at the S1P1 receptor. Medicinal chemistry targeting S1P1 receptor agonists is currently in progress. The FTY720 story provides a methodology where in vivo screens rather than in vitro screens play important roles in the lead optimization. Unlike recent drug discovery methodologies, such a strategy as adopted by the FTY720 program would more likely meet serendipity.Aggressive carcinomas ferment glucose to lactate even in the presence of oxygen. This particular metabolism, termed aerobic glycolysis, the glycolytic phenotype, or the Warburg effect, was discovered by Nobel laureate Otto Warburg in the 1920s. Since these times, controversial discussions about the relevance of the fermentation of glucose by tumours took place; however, a majority of cancer researchers considered the Warburg effect as a non-causative epiphenomenon. Recent research demonstrated, that several common oncogenic events favour the expression of the glycolytic phenotype. Moreover, a suppression of the phenotypic features by either substrate limitation, pharmacological intervention, or genetic manipulation was found to mediate potent tumour-suppressive effects. The discovery of the transketolase-like 1 (TKTL1) enzyme in aggressive cancers may deliver a missing link in the interpretation of the Warburg effect. TKTL1-activity could be the basis for a rapid fermentation of glucose in aggressive carcinoma cells via the pentose phosphate pathway, which leads to matrix acidification, invasive growth, and ultimately metastasis. TKTL1 expression in certain non-cancerous tissues correlates with aerobic formation of lactate and rapid fermentation of glucose, which may be required for the prevention of advanced glycation end products and the suppression of reactive oxygen species. There is evidence, that the activity of this enzyme and the Warburg effect can be both protective or destructive for the organism. These results place glucose metabolism to the centre of pathogenesis of several civilisation related diseases and raise concerns about the high glycaemic index of various food components commonly consumed in western diets.Photodynamic therapy (PDT) is a clinical treatment that combines the effects of visible light irradiation with subsequent biochemical events that arise from the presence of a photosensitising drug (possessing no dark toxicity) to cause destruction of selected cells. Today, the most common agent used in dermatological PDT is 5-aminolevulinic acid (ALA). As a result of its hydrophilic character, ALA penetrates skin lesions poorly when applied topically. Its systemic bioavailability is limited and it is known to cause significant side effects when given orally or intravenously. Numerous chemical derivatives of ALA have been synthesised with the aims of either improving topical penetration or enhancing systemic bioavailability, while reducing side effects. In vitro cell culture experiments with ALA derivatives have yielded promising results. However, if ALA derivatives are to demonstrate meaningful clinical benefits, a rational approach to topical formulation design is required, along with a systematic study aimed at uncovering the true potential of ALA derivatives in photodynamic therapy. With respect to systemic ALA delivery, more study is required in the developing area of ALA-containing dendrons and dendrimers.Pseudomonas tolaasii, P. reactans and Burkholderia gladioli pv. agaricicola, are responsible of diseases on some species of cultivated mushrooms. The main bioactive metabolites produced by both Pseudomonas strains are the lipodepsipeptides (LDPs) tolaasin I and II and the so called White Line Inducing Principle (WLIP), respectively, LDPs which have been extensively studied for their role in the disease process and for their biological properties. In particular, their antimicrobial activity and the alteration of biological and model membranes (red blood cell and liposomes) was established. In the case of tolaasin I interaction with membranes was also related to the tridimensional structure in solution as determined by NMR combined with molecular dynamic calculation techniques. Recently, five news minor tolaasins, tolaasins A–E, were isolated from the culture filtrates of P. tolaasii and their chemical structure was determined by extensive use of NMR and MS spectroscopy. Furthermore, their antimicrobial activity was evaluated on target micro-organisms (fungi—including the cultivated mushrooms Agaricus bisporus, Lentinus edodes, and Pleurotus spp.—chromista, yeast and bacteria). The Gram positive bacteria resulted the most sensible and a significant structure-activity relationships was apparent. The isolation and structure determination of bioactive metabolites produced by B. gladioli pv. agaricicola are still in progress but preliminary results indicate their peptide nature. Furthermore, the exopolysaccharide (EPS) from the culture filtrates of B. gladioli pv. agaricicola, as well as the O-chain and lipid A, from the lipopolysaccharide (LPS) of the three bacteria, were isolated and the structures determined.The heterogeneity of symptoms and disease progression observed in synucleinopathies, of which Parkinson’s disease (PD) is the most common representative, poses large problems for the discovery of novel therapeutics. The molecular basis for pathology is currently unclear, both in familial and in sporadic cases. While the therapeutic effects of L-DOPA and dopamine receptor agonists constitute good options for symptomatic treatment in PD, the development of neuroprotective and/or neurorestorative treatments for PD and other synucleinopathies faces significant challenges due to the poor knowledge of the putative targets. Recent experimental evidence strongly suggests a central role for neurotoxic α-synuclein oligomeric species in neurodegeneration. The events leading to protein oligomerization, as well as the oligomeric species themselves, are likely amenable to modulation by small molecules, which are beginning to emerge in high throughput compound screens in a variety of model organisms. The therapeutic potential of small molecule modulators of oligomer formation demands further exploration and validation in cellular and animal disease models in order to accelerate human drug development.Introduction A large number of antiepileptic drugs (AEDs) are available today, but they may not be satisfactory regarding clinical efficacy, tolerance, toxicity or pharmacokinetic properties. The purpose of this review is to focus upon the rationale behind the chemical modifications of several recently marketed AEDs or drugs in development and to categorize them according to the main purposes for the improvements: better efficacy or tolerability accompanied by improved pharmacokinetic properties. Material and Method AEDs that have been chemically modified to new derivatives during the last years are reviewed based on recent publications and PubMed-searches. Results and Discussion Improvement in pharmacokinetic parameters may affect both tolerability and efficacy. Modifications to improve tolerability include various valproate analogues, divided into aliphatic amides, cyclic derivatives or amino acid conjugates. Furthermore, there are the carbamazepine analogues oxcarbazepine and eslicarbazepine, the felbamate analogues fluorofelbamate and carisbamate (RWJ 33369), and the lamotrigine analogue JZP-4. The levetiracetam analogues brivaracetam and seletracetam and the derivatives of gabapentin, pregabalin and XP13512, have improved selectivity compared to their parent compounds. Other new drugs have new mechanisms of action related to GABA and glutamate receptors; the glutamate antagonists like topiramate (talampanel and NS-1209), and GABAA receptor agonists, benzodiazepine or progesterone analogues (ELB-139 and ganaxolone). Conclusion Further challenges for development of new AEDs include investigations of target molecules affected by pathophysiological processes and detailed structure-activity relationships with focus on stereoselectivity. These potential drugs may become of importance in future drug therapy in epilepsy and other CNS disorders.Lead optimization using drug metabolism and pharmacokinetics (DMPK) parameters has become one of the primary focuses of research organizations involved in drug discovery in the last decade. Using a combination of rapid in vivo and in vitro DMPK screening procedures on a large array of compounds during the lead optimization process has resulted in development of compounds that have acceptable DMPK properties. In this review, we present a general screening paradigm that is currently being used as part of drug discovery at Schering-Plough and we describe a case study using the Hepatitis C Virus (HCV) protease inhibitor program as an example. By using the DMPK optimization tools, a potent HCV protease inhibitor, SCH 503034, was selected for development as a candidate drug.Integrins have been reported to mediate cell survival, proliferation, differentiation, and migration programs. For this reason, the past few years have seen an increased interest in the implications of integrin receptors in cancer biology and tumor cell aggression. This review considers the potential role of integrins in cancer and also addresses why integrins are present attractive targets for drug design. It discusses of the several properties of the integrin-based chemotherapeutic agents currently under consideration clinically and provides an insight into cancer drug development using integrin as a target.

Data modelling with neural networks: Advantages and limitations

The origins and operation of artificial neural networks are briefly described and their early application to data modelling in drug design is reviewed. Four problems in the use of neural networks in data modelling are discussed, namely overfitting, chance effects, overtraining and interpretation, and examples are given of the means by which the first three of these may be avoided. The use of neural networks as a variable selection tool is shown and the advantage of networks as a nonlinear data modelling device is discussed. The display of multivariate data in two dimensions employing a neural network is illustrated using experimental and theoretical data for a set of charge transfer complexes.

Estimation of pKa Using Semiempirical Molecular Orbital Methods. Part 1: Application to Phenols and Carboxylic Acids.

The electronic properties of small molecules can be calculated quickly and with a reasonable degree of accuracy using semiempirical QM methods. In this study a set of QM properties derived from frontier electron theory have been used to produce a predictive model of the dissociation constants of phenols, benzoic acids and aliphatic carboxylic acids. The pK a values and structures of nearly 500 compounds were extracted from the Physprop database for this purpose. Multiple linear regression was used to search for relationships between pK a and the calculated QM properties. In most cases only a single independent variable, electrophilic superdelocalisability, was needed to produce a good model of pK a . The advantages of our approach are in the speed of calculation and the simplicity of the resultant models. The merits of using semiempirical methods to predict pK a are discussed in relation to previous studies.

The significance of acid/base properties in drug discovery

While drug discovery scientists take heed of various guidelines concerning drug-like character, the influence of acid/base properties often remains under-scrutinised. Ionisation constants (pK(a) values) are fundamental to the variability of the biopharmaceutical characteristics of drugs and to underlying parameters such as logD and solubility. pK(a) values affect physicochemical properties such as aqueous solubility, which in turn influences drug formulation approaches. More importantly, absorption, distribution, metabolism, excretion and toxicity (ADMET) are profoundly affected by the charge state of compounds under varying pH conditions. Consideration of pK(a) values in conjunction with other molecular properties is of great significance and has the potential to be used to further improve the efficiency of drug discovery. Given the recent low annual output of new drugs from pharmaceutical companies, this review will provide a timely reminder of an important molecular property that influences clinical success.

Estimation of pKa using semiempirical molecular orbital methods. Part 2: Application to amines, anilines and various nitrogen containing heterocyclic compounds

The pK a of a compound directly influences its biopharmaceutical profile. This article describes the development of a method for estimating pK a values for a number of nitrogen containing chemical structures using semiempirical QM properties derived from frontier electron theory. Typically, the property giving the best correlation with pK d was the electrophilic superdelocalisability of the nitrogen atom resulting in regression equations with r 2 values up to 0.94. The advantages of this technique are in the simplicity of the models and the speed of calculation, suggesting that this method could be widely applied to the estimation of pK a values. The success of this approach is discussed in relation to other methods.

Theoretical hydrogen bonding parameters for drug design.

Hydrogen bonding interactions play a major role in many chemical and biological processes. This article describes the development of a method for the quantitative estimation of the hydrogen-bonding donor strengths of OH/NH moieties and of the hydrogen bonding acceptor strengths of O/N atoms in different chemical structures. The method is based on the correlation of experimentally observed hydrogen-bonding strengths with quantum-mechanical derived properties, calculated on the acceptor atom (for hydrogen-bond acceptors) and on the heavy atom attached to the donor hydrogen (for hydrogen-bond donors). The properties giving the best correlation with the experimental hydrogen bonding scales were electrophilic superdelocalizability and self-atom polarizability. The best equations found have been implemented in a Web-based tool for hydrogen-bond strength prediction.

The acid–base profile of a contemporary set of drugs: implications for drug discovery

Acid–base ionization constant (pK a) values have considerable influence on the physicochemical and pharmacokinetic properties of therapeutic substances. A set of 907 drugs was examined to determine the proportion of drugs that contain an ionizable group and the distribution of their pK a values. Using this contemporary set of compounds it was found that 64% of these compounds contained an ionizable group. Within this group of ionizable compounds, 34% contained a single basic group while only 20% contained a single acidic functional group. The single acid and single base containing substances were investigated further to examine the distributions of their pK a values. These data are discussed and analyzed with a focus on the entire set as well as central nervous system, non-central nervous system and oral drugs. The findings from this research will prompt pharmaceutical companies to assess the constitution of their screening libraries, such that focus is placed on the proportion of ionizable substances, the ratio of acids to bases and the distribution of pK a values.

Platensimycin: A Promising Antimicrobial Targeting Fatty Acid Synthesis

Platensimycin was recently discovered by Merck Research Laboratories and has created considerable interest given its potent antibacterial activity and mode of action. The use of RNA gene-silencing techniques and screening libraries of natural products allowed Merck to find this antibiotic which may have otherwise been missed using conventional methods. Interestingly, platensimycin has shown good activity against a panel of Gram positive organisms which included various resistant strains. Platensimycin works by inhibiting beta-ketoacyl synthases I/II (FabF/B) which are key enzymes in the production of fatty acids required for bacterial cell membranes. So far, a number of groups have explored synthetic strategies for platensimycin and this work has subsequently lead to the synthesis of active analogues. Given its mode of action, it is intriguing as to why Merck themselves patented only a single compound and have not apparently sought to generate further libraries. This review will discuss the origins of platensimycin, its mechanism of action, synthetic schemes and where the future may take us following this fascinating discovery.

Prediction of drug solubility from molecular structure using a drug-like training set

Using a training set of 191 drug-like compounds extracted from the AQUASOL database a quantitative structure-property relationship (QSPR) study was conducted employing a set of simple structural and physicochemical properties to predict aqueous solubility. The resultant regression model comprised five parameters (ClogP, molecular weight, indicator variable for aliphatic amine groups, number of rotatable bonds and number of aromatic rings) and demonstrated acceptable statistics (r 2 = 0.87, s = 0.51, F = 243.6, n = 191). The model was applied to two test sets consisting of a drug-like set of compounds (r 2 = 0.80, s = 0.68, n = 174) and a set of agrochemicals (r 2 = 0.88, s = 0.65, n = 200). Using the established general solubility equation (GSE) on the training and drug-like test set gave poorer results than the current study. The agrochemical test set was predicted with equal accuracy using the GSE and the QSPR equation. The results of this study suggest that increasing molecular size, rigidity and lipophilicity decrease solubility whereas increasing conformational flexibility and the presence of a non-conjugated amine group increase the solubility of drug-like compounds. Indeed, the proposed structural parameters make physical sense and provide simple guidelines for modifying solubility during lead optimisation.