In drug discovery, thousands of compounds are assayed to detect activity against a biological target. The goal of drug discovery is to identify compounds that are active against the target (e.g. inhibit a virus). Statistical learning in drug discovery seeks to build a model that uses descriptors characterizing molecular structure to predict biological activity. However, the characteristics of drug discovery data can make it difficult to model the relationship between molecular descriptors and biological activity. Among these characteristics are the rarity of active compounds, the large volume of compounds tested by high-throughput screening, and the complexity of molecular structure and its relationship to activity. This thesis focuses on the design of statistical learning algorithms/models and their applications to drug discovery. The two main parts of the thesis are: an algorithm-based statistical method and a more formal model-based approach. Both approaches can facilitate and accelerate the process of developing new drugs. A unifying theme is the use of unsupervised methods as components of supervised learning algorithms/models. In the first part of the thesis, we explore a sequential screening approach, Cluster Structure-Activity Relationship Analysis (CSARA). Sequential screening integrates High Throughput Screening with mathematical modeling to sequentially select the best compounds. CSARA is a cluster-based and algorithm driven method. To gain further insight into this method, we use three carefully designed experiments to compare predictive accuracy with Recursive Partitioning, a popular structureactivity relationship analysis method. The experiments show that CSARA outperforms Recursive Partitioning. Comparisons include problems with many descriptor sets and situations in which many descriptors are not important for activity. In the second part of the thesis, we propose and develop constrained mixture discriminant analysis (CMDA), a model-based method. The main idea of CMDA is to model the distribution of the observations given the class label (e.g. active or inactive class) as a constrained mixture distribution, and then use Bayes’ rule to predict the probability of being active for each observation in the testing set. Constraints are used to deal with the otherwise explosive growth of the number of parameters with increasing dimensionality. CMDA is designed to solve several challenges in modeling drug data sets, such as multiple mechanisms, the rare target problem (i.e. imbalanced classes), and the identification of relevant subspaces of descriptors (i.e. variable selection). We focus on the CMDA1 model, in which univariate densities form the building blocks of the mixture components. Due to the unboundedness of the CMDA1 log likelihood function, it is easy for the EM algorithm to converge to degenerate solutions. A special Multi-Step EM algorithm is therefore developed and explored via several experimental comparisons. Using the multi-step EM algorithm, the CMDA1 model is compared to model-based clustering discriminant analysis (MclustDA). The CMDA1 model is either superior to or competitive with the MclustDA model, depending on which model generates the data. The CMDA1 model has better performance than the MclustDA model when the data are high-dimensional and unbalanced, an essential feature of the drug discovery problem! An alternate approach to the problem of degeneracy is penalized estimation. By introducing a group of simple penalty functions, we consider penalized maximum likelihood estimation of the CMDA1 and CMDA2 models. This strategy improves the convergence of the conventional EM algorithm, and helps avoid degenerate solutions. Extending techniques from Chen et al. (2007), we prove that the PMLE’s of the two-dimensional CMDA1 model can be asymptotically consistent.

Docking is one of the methods in virtual screeing. Studies from around 1980 to now, many docking software have been developed, but these software have many short comings. The software currently used for docking have many disadvantage: poor efficiency, rigid structure of the proteins and the ligands, poor accuracy, without the polarization after binding, leading virtual screening is still stuck in a supporting role. Our experiment with new method improves those shortcomings of docking. With this new method, we obtain the following improvements in docking process: better efficiency, flexible structure of the proteins and the ligands, better accuracy. In the depression-related protein docked with traditional Chinese medicine test. We change the conformations of ligands with the shapes of active sites before posing, this makes the conformation of complex much more reasonable, even more complicated, large ligands. In the experiment of random sites docking, we found a possible path for compounds traveling into active sites. We illustrate a docking area by linking all possible docking sites. The lead compound may not successfully travel into active site when this area is occupied by other proteins or ligands. In the docking experiment with side-chain rotation, we rotate the torsion angle to make side chains relax. We obtained a similar result with molecular dynamics, and saved a lot of time.

Pharmaceutical biotechnology has been emerging as a defined, increasingly important area of science dedicated to the discovery and delivery of drugs and therapies for the treatment of various human diseases. In contrast to the advancement in pharmaceutical biotechnology, current drug discovery efforts are facing unprecedented challenges. Difficulties in identifying novel drug targets and developing effective and safe drugs are closely related to the complexity of the network of interacting human proteins. Protein-protein interactions mediate virtually all cellular processes. Therefore both identification and understanding of protein-protein interactions are essential to the process of deciphering disease mechanisms and developing treatments. Unfortunately, our current knowledge and understanding of the human interactome is largely incomplete. Most of the unknown protein-protein interactions are expected to be weak and/or transient, hence are not easily identified. These unknown or uncharacterized interactions could affect the efficacy and toxicity of drug candidates, contributing to the high rate of failure. In an attempt to facilitate the ongoing efforts in drug discovery, we describe herein a series of novel methods and their applications addressing the broad topic of protein-protein interactions. We have developed a highly efficient site-specific protein cross-linking technology mediated by the genetically incorporated non-canonical amino acid L-DOPA to facilitate the identification and characterization of weak protein-protein interactions. We also established a protocol to incorporate L-DOPA into proteins in mammalian cells to enable in vivo site-specific protein cross-kinking. We then applied the DOPA-mediated cross-linking methodology to design a protein probe which can potentially serve as a diagnostic tool or a modulator of protein-protein interactions in vivo. To deliver such engineered proteins or other bioanalytical reagents into single live cells, we established a laser-assisted cellular nano-surgery protocol which would enable detailed observations of cell-to-cell variability and communication. Finally we investigated a possible experimental scheme to genetically evolve a fluorescent peptide, which has tremendous potential as a tool in cellular imaging and dynamic observation of protein-protein interactions in vivo. We aim to contribute to the discovery and development of new drugs and eventually to the overall health of our society by adding the technology above to the array of currently available bioanalytical tools. / text

eEF-2K, also known as CaM kinase-III, is an atypical protein kinase which negatively regulates the global rate of protein synthesis through the phosphorylation and inactivation of its substrate eEF-2. Recently eEF-2K has been validated as a novel target for anti-cancer therapy. However, a detailed understanding of the role of eEF-2K in cancer biology is unavailable. Mechanistic studies can often provide an understanding of enzyme function. Therefore, we determined the kinetic mechanism of eEF-2K using a peptide substrate (Acetyl-RKKYKFNEDTERRRFL-amide). We found that eEF-2K adopts a ternary-complex, steady state ordered mechanism, with ATP binding required before the peptide substrate. A good cellular inhibitor is required for elucidating the role of eEF-2K in cancer biology. To date, NH125 is the only inhibitor used to investigate the activity of eEF-2K in cells. Although it is reported as a specific inhibitor of eEF-2K, its exact mode of action has not been reported. Through in-vitro assays and cellular studies, we found that NH125 is a non-specific inhibitor of eEF-2K that blocks eEF-2 phosphorylation in cells. There is a great demand for specific inhibitors of eEF-2K. We developed a fluorescence high throughput assay system for eEF-2K. The assay utilizes the peptide substrate labeled with a Sox moiety whose phosphorylation can be monitored at 485 nm in the presence of magnesium. We also validated the assay in a screen of 30,000 compounds in 384 well plates. We found the assay to be robust and identified a relatively specific inhibitor of eEF-2K and determined its mechanism of action. We found it behaved as a slowly reversible inhibitor of eEF-2K with a two step inhibition mechanism - fast initial binding at the enzyme active site, followed by a slower inactivation step. We propose that the nitrile group on the compound binds to the active site thiol in the enzyme covalently forming a reversible thioimidate adduct to inactivate the enzyme. / text

In efforts to search for tool compounds that are capable of probing normal and disease-associated biological processes, both quality and identity of the screening collection are very important. Towards this goal, diversity-oriented synthesis (DOS) has been explored for a decade, which aims to populate the chemical space with diverse sets of small molecules distinct from the traditional ones obtained via combinatorial chemistry. In the practice of DOS, macrocyclic ring-closing metathesis (RCM) reactions have been widely used. However, the prediction and control of stereoselectivity of the reaction is often challenging; chemical transformation of the olefin moiety within the product is in general limited. Chapter I of this thesis describes a methodology that addresses both problems simultaneously and thus extends the utility of the RCM reactions. By installing a silyl group at the internal position of one of the olefin termini, the RCM reaction could proceed with high stereoselectivity to afford the (E)-alkenylsiloxane regardless of the intrinsic selectivity of the substrate. The resulting alkenylsiloxane can be transformed to a variety of functionalities in a regiospecific fashion. The conversion of the (E)-alkenylsiloxanes to alkenyl bromides could proceed with inversion of stereochemistry for some substrates allowing the selective access of both the E- and Z-trisubstituted macrocyclic alkenes. It was also found that the silyl group could trap the desired mono-cyclized product by suppressing nonproductive pathways. Chapter II of this thesis describes the application of the concept of DOS in the area of fragment-based drug discovery. Most fragment libraries used to date have been limited to aromatic heterocycles with an underrepresentation of chiral, enantiopure, \(sp^3\)-rich compounds. In order to create a more diverse fragment collection, the build/couple/pair algorithm was adopted. Starting from proline derivatives, a series of bicyclic compounds were obtained with complete sets of stereoisomers and high \(sp^3\) ratio. Efforts are also described toward the generation of diverse fragments using methodology described in Chapter I. The glycogen synthase kinase \((GSK3\beta)\) was selected as the proof-of-concept target for screening the DOS fragments. / Chemistry and Chemical Biology

Pleckstrin homology (PH) domains are structurally conserved domains, which generally bind to phosphatidylinositol phosphate (PtdInsP) lipids. They are present in a variety of proteins, including those that are upregulated in cancer growth and metastasis, and represent a crucial component of intracellular signaling cascades and membrane translocation. Thus, they may be considered as attractive targets for cancer drug therapy. AKT (protein kinase B), a pleckstrin homology lipid binding domain and a serine/threonine kinase-containing protein, is a key component of the phophatidylinositol-3-kinase (PI3K)/AKT cell survival signaling pathway which is activated in a variety of cancers, including prostate, pancreatic, and skin cancers. In this study, I report the finding of a novel inhibitor of AKT; PH-427. I describe its effects on binding to the PH domain of AKT thus preventing its binding to PtdIns3-P at the plasma membrane and subsequent activation. In vivo testing of the drug led to reduction of tumor size and numbers in a mouse pancreatic cancer model. Additional testing of PH-427 on squamous cell carcinomas revealed that the drug is able to reduce tumor burden and multiplicity in vivo when topically applied. Thus, we demonstrate proof-of-principle in targeting PH domains as a viable cancer drug therapy option. The effects of PH-427 raised the intriguing possibility that targeting PH domains may have beneficial effects in other signaling pathways with PH domain-containing proteins. Guanine exchange factors (GEFs) contain a Dbl homology (DH) domain and a PH domain and have been shown to be involved in the process of metastasis. More specifically, RacGEFs activate Rac1 GTPase by facilitating the exchange of GDP to GTP. Over-expression of certain GEFs has been shown to contribute to increased malignancy in a variety of cancers. T-lymphoma invasion and metastasis-inducing protein-1 (Tiam1) is a highly conserved GEF and contains an N-terminal pleckstrin homology domain (nPH) and a DH/C-terminal PH domain (cPH). Tiam1 has been found to be over-expressed in several cancers, including breast, colon and prostate cancers. In this study, I describe the identification, development, experimental testing, and potential mechanism of action of novel small molecule inhibitors targeting the RacGEF Tiam1 to inhibit prostate cancer bone metastasis.

One unique feature of cancer cells that can be exploited for anticancer drug discovery is their dependence on their own cellular stress responses to survive the stressful acidotic, hypoxic and nutrient-deprived conditions within the tumor. Reasoning that desert organisms surviving under stressful conditions may have evolved to produce small molecule metabolites capable of modulating heat shock protein 90 (Hsp90) function, and/or other cell stress responses, we employed the cellular heat shock response in a moderate-throughput phenotypic assay. This strategy has resulted in the isolation and characterization of a number of small molecule natural products with heat shock induction activity from these organisms. Three such natural products are the subject of this study.In a limited structure-activity relationship (SAR) study, a previously known Hsp90 inhibitor radicicol (RAD), and several structurally related molecules including the fungal metabolite monocillin 1 (MON) were found to interact with Hsp90. In addition, RAD and MON were shown to lead to the degradation of Hsp90 client proteins involved in the cancer cell survival the estrogen receptor (ER) and the insulin-like growth factor receptor 1 (IGF-1R).We further characterized MON and showed that by targeting the molecular chaperone Hsp90, this compound induces components of the heat shock response at the transcriptional and translational levels, and leads to the acquisition of a thermotolerant phenotype in seedlings of the plant Arabidopsis thaliana. These findings support our hypothesis that there is ecological significance to the elaboration of small molecules that target stress responses.A number of extracts active in our phenotypic assay contained small molecules with no apparent Hsp90 activity. One such extract afforded terrecyclic acid A (TCA) with significant anti-tumor activity against a panel of human cancer cell lines. To characterize the biological activities of TCA we examined three key stress responsesthe heat shock, oxidative, and inflammatory responsesand show that TCA destabilizes these pathways associated with cancer cell survival through induction of oxidative stress (ROS), and inhibition of NF-kappaB transactivation.The isolation of RAD, MON and TCA from Sonoran desert organisms provides proof of principle that we have developed an effective strategy for the discovery of small molecule modulators of cellular stress responses that can serve as leads for the development of new anticancer drugs with novel mechanisms of action.

There is a need for new antiviral drugs. Especially for the treatment of adenovirus infections, since no approved anti-adenoviral drugs are available. Adenovirus infections in healthy persons are most often associated with respiratory disease, diarrhea and infections of the eye. These infections can be severe, but are most often self-limiting. However, in immunocompromised patients, adenovirus infections are associated with morbidity and high mortality rates. These patients are mainly stem cell or bone marrow transplantation recipients, however solid organ transplantation recipients or AIDS patients may be at risk as well. In addition, children are at higher risk to develop disseminated disease. Due to the need for effective anti-adenoviral drugs, we have developed a cell based screening assay, using a replication-competent GFP expressing adenovirus vector based on adenovirus type 11 (RCAd11GFP). This assay facilitates the screening of chemical libraries for antiviral activity. Using this assay, we have screened 9800 small molecules for anti-adenoviral activity with low toxicity. One compound, designated Benzavir-1, was identified with activity against representative types of all adenovirus species. In addition, Benzavir-1 was more potent than cidofovir, which is the antiviral drug used for treatment of adenovirus disease. By structure-activity relationships analysis (SAR), the potency of Benzavir-1 was improved. Hence, the improved compound is designated Benzavir-2. To assess the antiviral specificity, the activity of Benzavir-1 and -2 on both types of herpes simplex virus (HSV) was evaluated. Benzavir-2 displayed better efficacy than Benzavir-1 and had an activity comparable to acyclovir, which is the original antiviral drug used for therapy of herpes virus infections. In addition, Benzavir-2 was active against acyclovir-resistant clinical isolates of both HSV types. To expand our search for compounds with antiviral activity, we turned to the natural products. An ethyl acetate extract library was established, with extracts derived from actinobacteria isolated from sediments of the Arctic Sea. Using our screening assay, several extracts with anti-adenoviral activity and low toxicity were identified. By activity-guided fractionation of the extracts, the active compounds could be isolated. However, several compounds had previously been characterized with antiviral activity. Nonetheless, one compound had uncharacterized antiviral activity and this compound was identified as a butenolide. Additional butenolide analogues were found and we proposed a biosynthetic pathway for the production of these compounds. The antiviral activity was characterized and substantial differences in their toxic potential were observed. One of the most potent butenolide analogues had minimal toxicity and is an attractive starting point for further optimization of the anti-adenoviral activity. This thesis describes the discovery of novel antiviral compounds that targets adenovirus and HSV infections, with the emphasis on adenovirus infections. The discoveries in this thesis may lead to the development of new antiviral drugs for clinical use.

As of today, there are 25 drugs approved for the treatment of HIV and AIDS. Nevertheless, HIV continues to infect and kill millions of people every year. Despite intensive research efforts, both a vaccine and a cure remain elusive and the long term efficacy of existing drugs is limited by the development of resistant HIV strains. New drugs and preventive strategies that are effective against resistant virus are therefore still needed. In this thesis an enzymological approach, primarily using SPR-based interaction kinetic analysis, has been used for identification and characterization of compounds of potential use in next generation anti-HIV drugs. By screening of a targeted non-nucleoside reverse transcriptase inhibitor (NNRTI) library, one novel and highly potent NNRTI was identified. The inhibitor was selected with respect to resilience to drug resistance and for high affinity and slow dissociation – a kinetic profile assumed to be suitable for inhibitors used in topical microbicides. In order to confirm the hypothesis that such a kinetic profile would result in an effective preventive agent with long-lasting effect, the correlation between antiviral effect and kinetic profile was investigated for a panel of NNRTIs. The kinetic profiles revealed that NNRTI efficacy is dependent on slow dissociation from the target, although the induced fit interaction mechanism prevented quantification of the rate constants. To avoid cross-resistance, the next generation anti-HIV drugs should be based on chemical entities that do not resemble drugs in clinical use, either in structure or mode-of-action. Fragment-based drug discovery was used for identification of structurally new inhibitors of HIV-enzymes. One fragment that was effective also on variants of HIV RT with resistance mutations was identified. The study revealed the possibility of identifying structurally novel NNRTIs as well as fragments interacting with other sites of the protein. The two compounds identified in this thesis represent potential starting points for a new generation of NNRTIs. The applied methodologies also show how interaction kinetic analysis can be used as an effective and versatile tool throughout the lead discovery process, especially when integrated with functional enzymological assays.

This thesis describes the synthesis and biological evaluation of a library of compounds containing the tetramic acid core in search of novel antibacterial drug candidates. Chapter 1 discusses the need for new antibiotics due to the emergence of virulent bacterial strains resistant to clinically available drugs and the hiatus in the discovery of new replacement antibitoics that has become a global threat to human health. Different platforms for antibacterial drug discovery and the re-emergence of natural products-based approach that has gained importance in the quest for novel antibiotics are discussed. In this regard, the intrinsic antibacterial activity of natural products containing a tetramate core structure and the strategies developed to synthesise the core scaffold are described. Chapter 2 discusses the use of ʟ-serine and ʟ-cysteine in tetramic acid synthesis and the application of ʟ-cysteine-derived thiazolidine templates suitable for stereoselective ring closing reactions to obtain the tetramic acid core with scope for further functionalization. Chapters 3 and 4 describe a range of synthetic routes for appropriate substitutions of the tetramate core for compound library generation. Elaboration of the tetramate core via carboxamide tetramate synthesis, Suzuki-Miyaura cross-coupling reactions, glycosylations and their aglycone analogue synthesis, etherification, tetramate-pyroglutamate systems, Buchwald aminations/amidations, cycloadditions and β-lactam hybrids as possible chemical modifications of the tetramate core structure are discussed. Chapter 5 describes the antibacetiral activity and physicochemical properties of the library of compounds synthesised. A preliminary evaluation of their antibiotic activity was conducted against S. aureus and E. coli using the hole-plate method. MICs of the tetramates synthesised were determined against several Gram-negative strains; Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa and Gram-positive strains; MRSA, Enterococcus faecalis and Streptococcus pneumoniae, in whole-cell bioassays. Physicochemical properties of the compound library were analysed to map the chemical space occupied by tetramates with potent antibacterial activity. Enzyme inhibition studies were conducted to identify possible modes of action that contribute to whole-cell antibiotic activity and in this regard, the inhibition of enzymes S. aureus topoisomerase IV, S. aureus RNA polymerase, E. coli RNA polymerase, E. coli gyrase and M. tuberculosis gyrase are discussed. Since plasma protein binding of compounds is an important factor that determines the bioavailability of antibiotics and their clinical outcome, a study of the binding affinity of these drug candidates to Human Serum Albumin (HSA) by both whole-cell bioassay and NMR spectroscopy-based protein binding experiments are discussed. Finally, a brief note on the potential of tetramic acids to function as proteasome inhibitors in anticancer chemotherapy is included at the end of this chapter.