| Malaria is an infectious disease caused by malaria parasite transmitted via mosquito. After being parasitic on host, their proliferation will cause periodic headaches, fever and vomiting. In severe cases, malaria may even cause anemia and death. Among species of Plasmodium that take human as host, Plasmodiumfalciparum is the most severe form with a strong drug resistance. It is responsible for severe malaria. As several commonly used antimalarial drugs such as chloroquine,artemisinine and folate are no longer as effective as before, it is urgent to develop new anti-malaria drugs.Plasmodium Falciparum glucose-6-phosphate dehydrogenase (PfG6PD) is the key enzyme for energy metabolism at red blood cell stage. It has become a potential novel target for anti-malaria drugs. In addition, Plasmodium Falciparum Dihydroorotatedehydrogenase (PfDHODH) is another important target for the crucial role in pyrimidine synthesis. It is the rate-limited enzyme in de novo pyrimidine biosynthesis pathway, the only pathway to get pyrimidine. So it is also an essencial enzyme for Plasmodium Falciparum. This thesis mainly focuses on inhibtors of these two key enzyme of Plasmodium Falciparum: glucose-6-phosphate dehydrogenase (Pf G6PD) and Dihydroorotate dehydrogenase (PfDHODH), and study the relationship between the structure of small molecular inhibitors and their bioactivities.With the development of chemoinformatics, it has been applied to many fields,including drug design by QSAR (Quantitate Structure-Activity Relationship);ligand-protein interaction and virtual screening by molecular docking; and molecular interaction mechanism study by molecular dynamic simulation. These methods will shorten the time and money to discover novo inhibitors. So in this work, several SAR and QSAR models were built to classify and predict the bioactivity of PfG6PD and Pf DHODH inhibitors. Moreover, the interaction of PfDHODH and its inhibitors were studied by molecular dynamic simulation. Last but not least, we also performed a virtual screening of blood coagulation factor FXIa inhibitors by molecular docking.This dissertation includes the following content:(1) Four classification models of high active and weakly active Pf G6PD inhibitors were established by SVM (Support Vector Machine) based on 414 PfG6PD inhibitors. They are separated into highly active group (IC50 <80μM) and weakly active group (IC50 > 80 μM). Radial distribution function (RDF) molecular descriptors were selected to represent the physical and chemical properties of molecules. Self-organizing map (SOM) and random method were used to split datasets into training set and test set. Stepwise regression and SVMAttributeEval were used to select molecular descriptors. Four classification models were built to classify the Pf G6PD inhibitors. Model 2W has the best prediction ability, which gave a prediction accuracy (Q) of 94 % and a Matthew’s correlation coefficient (MCC) of 0.88 on the test set. Some properties such as a atom charge, π atom charge, and lone pair electronegativity related descriptors are important for the interaction between the PfG6PD and the inhibitor. Also, the existence of aromatic ring, conjugated system,ketone groups and alkynylgroups may increase the bioactivity of PfG6PD.(2) Quantitative models of PfG6PD inhibitors were established by MLR(Multiple Linear Regression) and SVM, separately. A dataset of 262 Pf G6PD inhibitors were built. Radial distribution function (RDF) and global molecular descriptors were selected to represent the physical and chemical properties of molecules. Self-organizing map (SOM) and random method were used to split datasets into training set and test set. All the four models gave a correlation coefficient(r) over 0.86 and a root mean square error (RMSE) under 0.16 on training set, and a correlation coefficient (r) over 0.82 and a root mean square error (RMSE) under 0.18 on test set. It was found that models based on dataset separated by SOM gave a less RMSE than those based on dataset split randomly. By analyzing the descriptors in these models, it was observed that number of hydrogen bonding acceptors, ring complexity, a atom charges, total atom charges and effective atom polarizabilitly properties are important for the bioactivity of PfG6PD inhibitor.(3) Four QSAR models were generated to predict the bioactivity of PfDHODH inhibitors by SVM and MLR. 255 Pf DHODH inhibitors were collected from literatures. 2D property-weighted autocorrelation molecular descriptor and radial distribution function (RDF) molecular descriptors were selected to represent the physical and chemical properties of molecules. Self-organizing map (SOM) and random method were used to split datasets into training set and test set. All the models display good prediction quality with a correlation coefficient (r) over 0.85 and RMSE under 0.61 on both training sets and test sets. The study indicated that the hydrogen bonding ability, ring complexity, total charge and atom polarizabilities are predominant factors for inhibitors’ antimalarial activity. Compared with the similar published work, the dataset used here occupies a larger chemical space, so that the models built in this work will have a wider applicable domain than the others.(4) Molecular docking-based virtual screening of large library compounds for finding new inhibitors of blood coagulation factor FXIa. Blood coagulation factor FXIa is a trypsin-like serine protease which plays a major role in the coagulation pathway. In this part, we filtered 8,114 small molecules in sigma database by a multiple discipline (toxicity, inactive group, multiple OH group). We got 3,423 reactants in total. First, the reactants were docked into blood coagulation factor FXIa.Afterwards, the position and orientation of small molecules were calculated and those are closed to each other were used to generate the product molecules. Then the product docking was done, and another filter in which ADMET and drug-like properties was used to filter the products. Thus we got 314 potential FXIa inhibitors.Compared with the traditional combinatorial chemical method, our method decreases the work load significantly, and it increases the docking efficiency.All in all, several SAR and QSAR models of PfG6PD inhibitors and QSAR models of Pf DHODH inhibitors were generated in this study, and they showed good prediction ability. The models are based on larger database than published work. So they could be used to predict antimalarial activity of Pf G6PD and Pf DHODH inhibitors; they could also be used for virtual screening within a certain chemical space. Molecular descriptors selected by the models are the key factors to affect the ligand-enzyme interactions, and could be used in antimalarial drug design and development. Besides, molecular docking-based virtual screening of large library compounds for finding new inhibitors of blood coagulation factor FXIa. |
PDF Full Text Request