When applied to drug discovery, modern computational systems can provide insight into the highly complex systems underlying drug activity and predict compounds or targets of interest. Many tools have been developed for computer aided drug discovery (CADD), focusing on small molecule ligands, protein targets, or both. The aim of this thesis is the improvement of CADD tools for describing small molecule properties and application of CADD to several stages of drug discovery regarding two targets for the treatment of obesity and related diseases: the neuropeptide Y4 receptor (Y4R) and the melanocortin-4 receptor (MC4R). In the first chapter, the major categories of CADD are outlined, including descriptions for many of the popular tools and examples where these tools have directly contributed to the discovery of new drugs. Following the introduction, several improvements for encoding stereochemistry and signed property distribution are introduced and tested in scenarios meant to simulate applications in virtual high-throughput screening. Y4R and MC4R are both class A G-protein coupled receptors (GPCRs) with endogenous peptide ligands that play critical roles in the signaling of satiety and energy metabolism. So far, no structures from either receptor family have been experimentally elucidated. CADD was combined with high-throughput screening (HTS) to discover the first small molecule positive allosteric modulators (PAMs) of Y4R. Secondly, CADD techniques were used to model the interaction of Y4R and pancreatic polypeptide based on experimental results that elucidate specific binding contacts. Similar SB-CADD approaches were used to model the interaction of MC4R with its high affinity peptide agonist α-MSH. Due to its role in monogenic forms of obesity, these models were used to predict which residues directly participate in binding and correlate mutated residues with their potential role in the binding site.

The cyclotides are a family of naturally occurring peptides characterized by cyclic cystine knot (CCK) structural motif, which comprises a cyclic head-to-tail backbone featuring six conserved cysteine residues that form three disulfide bonds. This unique structural motif makes cyclotides exceptionally resistant to chemical, thermal and enzymatic degradation. They also exhibit a wide range of biological activities including insecticidal, cytotoxic, anti-HIV and antimicrobial effects. The cyclotides found in plants exhibit considerable sequence and structural diversity, which can be linked to their evolutionary history and that of their host plants. To clarify the evolutionary link between sequence diversity and the distribution of individual cyclotides across the genus Viola, selected known cyclotides were classified using signature sequences within their precursor proteins. By mapping the classified sequences onto the phylogenetic system of Viola, we traced the flow of cyclotide genes over evolutionary history and were able to estimate the prevalence of cyclotides in this genus. In addition, the structural diversity of the cyclotides was related to specific features of the sequences of their precursor proteins, their evolutionary selection and expression levels. A number of studies have suggested that the biological activities of the cyclotides are due to their ability to interact with and disrupt biological membranes. To better explain this behavior, quantitative structure-activity relationship (QSAR) models were developed to link the cyclotides’ biological activities to the membrane-interactive physicochemical properties of their molecular surfaces. Both scalar quantities (such as molecular surface areas) and moments (such as the distributions of specific properties over the molecular surface) were systematically taken into account in the development of these models. This approach allows the physicochemical properties of cyclotides to be geometrically interpreted, facilitating the development of guidelines for drug design using cyclotide scaffolds. Finally, an optimized microwave-assisted Fmoc-SPSS procedure for the total synthesis of cyclotides was developed. Microwave irradiation is used to accelerate and improve all the key steps in cyclotide synthesis, including the assembly of the peptide backbone by Fmoc-SPPS, the cleavage of the protected peptide, and the introduction of a thioester at the C-terminal carboxylic acid to obtain the head-to-tail cyclized cyclotide backbone by native chemical ligation.

The toxicological screening of the numerous chemicals that we are exposed to requires significant cost and the use of animals. Accordingly, more efficient methods for the evaluation of toxicity are required to reduce cost and the number of animals used. Computational strategies have the potential to reduce both the cost and the use of animal testing in toxicity screening. The ultimate goal of this thesis is to develop computational models for the prediction of toxicological endpoints that can serve as an alternative to animal testing. In Paper I, an attempt was made to construct a global quantitative structure-activity relationship (QSAR)model for the acute toxicity endpoint (LD50 values) using the Munro database that represents a broad chemical landscape. Such a model could be used for acute toxicity screening of chemicals of diverse structures. Paper II focuses on the use of acute toxicity data to support the prediction of chronic toxicity. The results of this study suggest that for related chemicals having acute toxicities within a similar range, their lowest observed effect levels (LOELs) can be used in read-across strategies to fill gaps in chronic toxicity data. In Paper III a k-nearest neighbor (k-NN) classification model was developed to predict human ether-a-go-go related gene (hERG)-derived toxicity. The results suggest that the model has potential for use in identifying compounds with hERG-liabilities, e.g. in drug development.

Methcathinone (MCAT) is the β-ketone analog of methamphetamine, and like its amphetamine analog, MCAT functions as a monoamine releaser that selectively promotes the release of dopamine (DA) and norepinephrine (NE) over serotonin (5-HT). MCAT produces amphetamine-like psychostimulant effects and is classified as a Schedule I drug of abuse by the United States Drug Enforcement Administration (DEA). Recently, synthetic MCAT analogs have emerged as designer drugs of abuse in Europe and the United States and have been marketed under deceptively benign names like “bath salts” in an attempt to evade legal restriction. These dangerous, recently emergent and novel drugs of abuse display varying selectivity to promote release of DA/NE vs. 5-HT, and selectivity for DA neurotransmission is believed to correlate with abuse liability. The goal of this dissertation was to conduct preclinical research to examine structural determinants of abuse-related behavioral and neurochemical effects produced by a series of synthetic MCAT analogs. Specifically, this project focused on one feature of the methcathinone scaffold: the para substituent of the benzene ring. A series of six novel MCAT analogs will be examined to evaluate how physicochemical parameters (steric, Es; electronic, σp; lipophilic, πp) of the para substituent influence in vitro monoamine transporter selectivity as well as in vivo neurochemical and behavioral effects. Results from this body of work implicate steric factors as being particularly important in determining a compound’s abuse-related neurochemical and behavioral effects. Thus, these data not only offer an improved understanding of the mechanism of abuse-related drug effects produced by synthetic MCAT analogs, but also help in the generation of homology models of the human DA and 5-HT transporters (DAT and SERT, respectively).

Thèse numérisée par la Direction des bibliothèques de l'Université de Montréal.

Thyroid disruptions by xenobiotics have been associated with a broad spectrum of severe adverse human health effects, such as impaired brain development and metabolic syndrome. Ingestion of indoor dust and contact with industrial chemicals are two significant human exposure routes of thyroid hormone disrupting chemicals (THDCs), raising serious concerns for human health. However, it is a laborious and costly process to identify THDCs using conventional experimental methods, due to the number of chemicals in commerce and the varieties of potential disruption mechanisms. In this thesis, we are aimed at in silico identification of novel THDCs targeting transthyretin (TTR) and thyroid hormone receptor (THR) among dust contaminants and commonly used industrial chemicals. In vitro assays were used to validate the in silico prediction results. Co-crystallization and molecular dynamics (MD) simulations were applied to reveal binding modes of THDCs at the studied biological targets and to explain their intermolecular recognition. The main findings presented in this thesis are: 1. Over 144 environmental pollutants have been confirmed as TTR-binders in vitro and these cover a wide range of environmental pollutants and show distinct chemical profiles including a large group of halogenated aromatic compounds and a second group of per- and polyfluoroalkyl substances. (Paper I) 2. In total 485 organic contaminants have been reported to be detected in household dust. The developed QSAR classification model predicted 7.6% of these dust contaminants and 53.1% of their metabolites as potential TTR-binders, which emphasizes the importance of metabolic bioactivation. After in vitro validation, four novel TTR binders with IC50 ≤ 10 µM were identified, i.e. perfluoroheptanesulfonic acid, 2,4,2',4'-tetrahydroxybenzophenone (BP2), 2,4,5-trichlorophenoxyacetic acid, and 3,5,6-trichloro-2-pyridinol. (Paper II) 3. The development of a robust structure-based virtual screening (VS) protocol resulted in the prediction of 31 dust contaminants as potential binders to THRβ1 including musk compounds, PFASs, and bisphenol A derivatives. The in vitro experiments confirmed four compounds as weak binders to THRβ1, i.e. 2,4,5-trichlorophenoxyacetic acid, bisphenol A (3-chloro-2-hydroxypropyl) (2,3-dihydroxypropyl) ether, 2,4,2',4'-tetrahydroxybenzophenone, and 2,4-dichlorophenoxyacetic acid. (Paper III) 4. We revealed the binding conformations of perfluorooctanesulfonic acid, perfluorooctanoic acid, and BP2 in the thyroxine binding sites (TBSs) of TTR by co-crystallizing TTR with the three compounds. A VS protocol was developed based on the TTR complex structures that predicted 192 industrial chemicals as potential binders to TTR. Seven novel TTR binders were confirmed by in vitro experiments including clonixin, 2,6-dinitro-p-cresol (DNPC), triclopyr, fluroxypyr, bisphenol S, picloram, and mesotrione. We further co-crystallized TTR with PBS, clonixin, DNPC, and triclopyr, and their complex structures showed that the compounds bind in the TBSs as proposed by the VS protocol. In summary, 13 indoor dust contaminants and industrial chemicals were identified as THDCs using a combination of in silico and in vitro approaches. To the best of our knowledge, none of these compounds has previously been reported to bind to TTR or THR. The identifications of these THDCs improve our understanding on the structure-activity relationships of THDCs. The crystal structures of TTR-THDC complexes and the information on THDC-Target intermolecular interactions provide a better understanding on the mechanism-of-actions behind thyroid disruption. The dataset compiled and in silico methods developed serve as a basis for identification of more diverse THDCs in the future and a tool for guiding de novo design of safer replacements.

Drug development is a long and expensive process. Potential drug candidates can fail clinical trials due to numerous issues, including metabolic stability and efficacy issues, wasting years of research effort and resource. This thesis detailed the development of in silico methods to predict the metabolic stability of structures and their bioavailability. Coralie Atom-based Statistical SOM Identifier (CASSI) is a site of metabolism (SOM) predictor which provides a SOM prediction based on statistical information gathered about previously seen atoms present in similar environments. CASSI is a real-time SOM predictor accessible via graphical user interface (GUI), allowing users to view the prediction results and likelihood of each atom to undergo different types of metabolic transformation. Fast Metabolizer (FAME)1 is a ligand-based SOM predictor developed around the same time by Kirchmair et al. In the course of the evaluation of CASSI and FAME performance, the two concepts were combined to produce FamePrint. FamePrint is a tool developed within the Coralie Cheminformatics Platform developed by Lhasa Limited. which can carry out SOM predictions, as well as bioisosteric replacement identification. Same as CASSI, this is available via the Coralie application GUI. The bioavailability issues caused by the metabolic enzyme, cytochrome P450 3A4, and transporter protein P-gylcoprotein are also investigated in this work, along with the potential synergistic relationship between the two systems. In silico classifiers to distinguish substrates against non-substrates of the two systems are produced and it was envisaged that these classifiers can be integrated into FamePrint as an additional layer of information available to the user when deciding on bioisosteric replacements to use when optimising a compound.

Industrial chemicals known as polychlorinated biphenyls (PCBs) were widely used for decades until their production was banned worldwide due to their persistence and toxicities to humans and other animals. Upon oxidative metabolism by cytochrome P450, hydroxylated metabolites of PCBs (OHPCBs) are formed. OHPCBs have been shown to competitively displace thyroxine from transthyretin, block normal hormonal activity, and inhibit phenol or family 1 sulfotransferases (SULTs) which catalyze sulfation of thyroid hormones and estrogens. Recently, three OHPCBs were shown to also interact with hydroxysteroid or family 2 sulfotransferases that play a role in the homeostasis of steroid hormones such as dehydroepiandrosterone (DHEA). The objectives of the studies presented in this thesis were to further examine the effects of selected OHPCBs on the activity of human hydroxysteroid sulfotransferase (hSULT2A1), to develop a three-dimensional quantitative structure activity relationship (3D-QSAR) model for OHPCBs as inhibitors of DHEA-sulfation catalyzed by this enzyme, and to investigate the mechanism of inhibition and binding of OHPCBs to hSULT2A1. All 15 OHPCBs examined inhibited the sulfation of 1 μ M [3H] DHEA, catalyzed by hSULT2A1 with IC50 values ranging from 0.6 to 96 μ M. The OHPCBs with a 3, 5-dichloro-4-hydroxy substitution were the most potent inhibitors of DHEA sulfation, and they were also shown to be substrates for hSULT2A1. Eight OHPCBs were substrates for hSULT2A1, and seven were solely inhibitors (i.e. they inhibited the sulfation of DHEA, yet they were not themselves sulfuryl-acceptors in hSULT2A1-catalyzed reactions). A 3D-QSAR model was developed utilizing comparative molecular field analysis (CoMFA). The model fit the data well and also had good predictability. The kinetics of inhibition showed that these OHPCBs were noncompetitive inhibitors of hSULT2A1. Binding studies utilizing the displacement of a fluorescent probe, 8-anilino-1-naphthalene sulfonic acid, revealed that several of the OHPCBs interact either at more than one binding site or with more than one enzyme conformation. Further exploration of this binding by molecular modeling showed that OHPCBs bind similarly to different conformations of the enzyme. This work has helped in our understanding of the roles of sulfotransferases in the metabolism and toxicities of OHPCBs, and it opens new avenues for future work.

<p>This thesis deals with the highly potent olfactory mucosa toxicant 2,6-dichlorophenyl methylsulphone (2,6-diClPh-MeSO₂) and its non-toxic 2,5-chlorinated isomer (2,5-diClPh-MeSO₂). In mice, both substances bind firmly in the olfactory mucosa and the olfactory bulb, which are important components of the sensory system. The 2,6-isomer induces olfactory mucosal necrosis with permanent loss of olfactory neuroepithelium and olfactory nerves. A major objective was to clarify the cause of this isomer-specific toxicity, and to identify which physicochemical characteristics determine the olfactory toxicity. The neurobehavioural toxicity of these substances was also examined.</p><p>The results revealed a rapid CYP-catalysed covalent binding of 2,6-diClPh-MeSO₂ in the rat olfactory mucosa, whereas the 2,5-dichlorinated isomer was not covalently bound. </p><p>Acute and chronic olfactory mucosal pathology were investigated and compared in rats and mice. Twenty-four hours after dosing to rats, 2,6-diClPh-MeSO₂ induced Bowman’s glands necrosis and sloughing of the olfactory epithelium similar to that previously reported in mice. At 3 weeks, however, there were dramatic differences in histological lesions. In mice, large parts of olfactory epithelium were replaced by respiratory-like epithelium. Large, bilateral, fibrous, cartilage and bone containing polyps occluding the lumen were confirmed. In rats, only minor patches of olfactory epithelium were replaced by a metaplastic atypical respiratory-like epithelium. 2,5-diClPh-MeSO₂ was non-toxic in rats as well as in mice.</p><p>In mice, 2,6-diClPh-MeSO₂ induced a dose-dependent and long-lasting ( ≥12 weeks) hyperactivity as well as long-lasting maze learning deficits. At 2 weeks hyperactivity and maze learning deficits were observed also in rats. Unexpectedly, 2,5-diClPh-MeSO₂ induced hyperactivity that lasted for two weeks. No effect on maze learning was observed with this isomer. No major differences between male and female rats or mice were found.</p><p>In conclusion, the results show that a CYP-catalysed formation and covalent binding of a reactive 2,6-diClPh-MeSO₂-metabolite in the Bowman’s glands precede the high olfactory mucosal toxicity in rodents. As determined by QSAR-modelling, a 2,6-dichlorinated benzene derivative with a large, polar, and strong electron withdrawing substituent in the primary position has the potential of being an olfactory mucosal toxicant. The observed 2,6-diClPh-MeSO₂-induced increase in motor activity, and maze learning deficits, were not correlated to the olfactory mucosal lesions. I propose that 2,6-diClPh-MeSO₂ causes a direct effect in the brain leading to neurobehaviuoral deficits. </p>

<p>A new approach to analysis of protein-ligand interactions, termed proteochemometrics, has been developed. Contrary to traditional quantitative structure-activity relationship (QSAR) methods that aim to correlate a description of ligands to their interactions with one particular target protein, proteochemometrics considers many targets simultaneously.</p><p>Proteochemometrics thus analyzes the experimentally determined protein-ligand interaction activity data by correlating the data to a complex description of all interaction partners and; in a more general case even to interaction environment and assaying conditions, as well. In this way, a proteochemometric model analyzes an “interaction space,” from which only one cross-section would be contemplated by any one QSAR model.</p><p>Proteochemometric models reveal the physicochemical and structural properties that are essential for protein-ligand complementarity and determine specificity of molecular interactions. From a drug design perspective, models may find use in the design of drugs with improved selectivity and in the design of drugs for multiple targets, such as mutated proteins (e.g., drug resistant mutations of pathogens).</p><p>In this thesis, a general concept for creating of proteochemometric models and approaches for validation and interpretation of models are presented. Different types of physicochemical and structural description of ligands and macromolecules are evaluated; mathematical algorithms for proteochemometric modeling, in particular for analysis of large-scale data sets, are developed. Artificial chimeric proteins constructed according to principles of statistical design are used to derive high-resolution models for small classes of proteins.</p><p>The studies of this thesis use data sets comprising ligand interactions with several families of G protein-coupled receptors. The presented approach is, however, general and can be applied to study molecular recognition mechanisms of any class of drug targets.</p>