In order to elucidate the function of the nifU gene product in nitrogenase maturation in Azotobacter vinelandii. the gene product has been hyperexpressed in Escherichia coli and characterized by various biophysical techniques. Following the initial characterization, site-directed mutagenesis of conserved cysteinyl residues was performed in order to gain further insight into the structure/function relationship of NifU. Both the Fe protein and the MoFe protein of nitrogenase require processing by additional nif genes including nifM (Fe protein), and nifE, N, B, H, V, and Q (MoFe protein). Two additional genes, nifU and nifS, are required for the maturation of both nitrogenase component proteins. It has been proposed that they may somehow be involved in metallocluster biosynthesis (Jacobson et al., 1989b). Our laboratory has determined that the nifS gene product (Nifs) is a pyridoxal-phosphate containing enzyme capable of catalyzing the desulfurization of L-cysteine and can provide the inorganic sulfide necessary for in vitro metallocluster biosynthesis of the Fe protein (Zheng, et al., 1993: Zheng, et al., 1994). / Ph. D.

DNA mismatch repair is one process in the preservation of genomic integrity. It has been found in Archeae, bacteria, plants, yeast and mammals. The mismatch repair system is highly conserved among species and allows the strand-specific elimination of base-base mispairs, chemical base modifications, as well as short insertion/deletion loops following DNA replication. The repair system also has important effects on homeologous recombination, contributing to the frequency of reciprocal exchanges. In humans, defects in the repair system have been found to be associated with tumorigenesis. In Escherichia coli, this pathway was originally called long patch repair before being renamed the methyl-directed mismatch repair system. It is unique in that it utilizes a DNA methylation pattern to discriminate between the parental DNA strand and the newly synthesized daughter DNA strand. The current model for the initiation of methyl-directed mismatch repair is that the mispaired bases are recognized and bound by the MutS protein with MutL as a helper protein for binding. MutL also assists the MutH protein to bind, thereby forming the completed initiation complex of MutS, MutL and MutH. In the presence of ATP, there is evidence for translocation ofthe complex along the DNA forming alpha loops. At a d(GATC) site the MutH protein binds and nicks the unmethylated daughter DNA strand 5' to the d(G) (by recognizing the N6-d(A) methylation of the parental DNA strand which it is unable to cut). This completes the initiation of the repair system and allows the hydrolysis and resynthesis of the daughter DNA strand. MutH is a monomer of 25.5 kD in solution and contains a latent Mg2+-dependent endonuclease activity. Unmethylated DNA is nicked without any discrimination on one of the two strands and fully methylated DNA is resistant to cleavage by MutH even though the protein is able to bind the d(GATC) site. The structure of MutH was recently solved and compared to a group of restriction endonucleases that share a structural common core domain with similarly placed catalytic residues. The MutH protein is comprised of two major domains that are able to pivot and rotate with respect to one another. The cleft between the two domains is large enough for double-strand DNA to bind. This research started with the determination of the MutH structure before it was known. After crystallizing the protein and collecting several heavy atom data sets, it was found that the electron density maps were too discontinuous to trace the structure of the protein. Following that work, site-directed mutagenesis was performed on several areas of MutH based on the similarity of MutH and PvuII structural models. The aims were to identify DNA binding residues (in two flexible loop regions), to determine if MutH has the same mechanism for DNA binding and catalysis as PvuII (MutH histidines 112 and 115), and to localize the residues responsible for MutH stimulation by MutL (MutH C-terminal tail region). An in-vivoscreen based on the mutator phenotype was used to select for functionally defective MutH mutants. These bacteria accumulate mutations at a greater frequency than wild-type and this was monitored by selection on plates with rifampicin. Three MutH mutants were identified from this screen (K48A, G49A, and Δ214). They were purified and assayed for total activity and binding ability. Four other mutants with wild-type phenotypic screen results were also chosen to confirm they were not involved in any MutH function (D47A, H112A, H115A, and Δ224). No DNA binding residues (such as D47A) were identified in the two flexible loop regions of MutH, although similar loops in PvuII are involved in DNA binding. The purified D47A MutH protein showed wild-type biochemical activity. Instead, the lysine residue (K48) in the first flexible loop was found to function in catalysis together with the three presumed catalytic amino acids (Asp70, Glu77, and Lys79). This purified MutH protein (K48A) had wild-type binding ability but no endonuclease activity without MutL. In the presence of MutL, the K48A protein had only a three-fold reduction in endonuclease activity. This research has shown that MutL stimulates the wild-type MutH activity by 1000-fold. The wild-type MutH stimulation by MutL for binding was only shown to be 16-fold. The G49A MutH mutant interferes with the proper functioning of the protein but is not informative about the mechanism of action. The binding ability of this mutant was the same as wild-type and the endonuclease activity was down 30-fold with a 10-fold stimulation by MutL. The extra methyl group of the alanine may cause slight structural changes in the lysine 48 side chain that slows catalysis. The two histidines (H112 and H115) in MutH that are in a similar position as the two histidines (H84 and H85) in PvuII (that signal for DNA binding and catalysis) were changed to alanines, but had wild-type activity both in-vivo and in-vitro. These results indicate that the MutH signal for DNA binding and catalysis remains unknown. The two deletion mutations (MutHΔ224 and MutHΔ214) in the C-terminal end of the protein, localized the MutL stimulation region to five amino acids (Ala220, Leu221, Leu222, Ala223, and Arg224). Mutant MutHΔ224 had wild-type MutL stimulation activity, while MutHΔ214 showed no MutL stimulation. Another deletion mutant, MutHΔ119, from another laboratory was shown to have wild-type MutL stimulation also. This leaves one (or more) of the remaining five residues as important for MutL stimulation.

Whole-exome and genome sequencing are widely used to diagnose individual patients. However, despite its success, this approach leaves many patients undiagnosed. This could be due to the need to discover more disease genes and variants or because disease phenotypes are novel and arise from a combination of variants of multiple known genes related to the disease. Recent rapid increases in available genomic, biomedical, and phenotypic data enable computational analyses, reducing the search space for disease-causing genes or variants and facilitating the prediction of causal variants. Therefore, artificial intelligence, data mining, machine learning, and deep learning are essential tools that have been used to identify biological interactions, including protein-protein interactions, gene-disease predictions, and variant--disease associations. Predicting these biological associations is a critical step in diagnosing patients with rare or complex diseases. In recent years, computational methods have emerged to improve gene-disease prioritization by incorporating phenotype information. These methods evaluate a patient's phenotype against a database of gene-phenotype associations to identify the closest match. However, inadequate knowledge of phenotypes linked with specific genes in humans and model organisms limits the effectiveness of the prediction. Information about gene product functions and anatomical locations of gene expression is accessible for many genes and can be associated with phenotypes through ontologies and machine-learning models. Incorporating this information can enhance gene-disease prioritization methods and more accurately identify potential disease-causing genes. This dissertation aims to address key limitations in gene-disease prediction and variant prioritization by developing computational methods that systematically relate human phenotypes that arise as a consequence of the loss or change of gene function to gene functions and anatomical and cellular locations of activity. To achieve this objective, this work focuses on crucial problems in the causative variant prioritization pipeline and presents novel computational methods that significantly improve prediction performance by leveraging large background knowledge data and integrating multiple techniques. Therefore, this dissertation presents novel approaches that utilize graph-based machine-learning techniques to leverage biomedical ontologies and linked biological data as background knowledge graphs. The methods employ representation learning with knowledge graphs and introduce generic models that address computational problems in gene-disease associations and variant prioritization. I demonstrate that my approach is capable of compensating for incomplete information in public databases and efficiently integrating with other biomedical data for similar prediction tasks. Moreover, my methods outperform other relevant approaches that rely on manually crafted features and laborious pre-processing. I systematically evaluate our methods and illustrate their potential applications for data analytics in biomedicine. Finally, I demonstrate how our prediction tools can be used in the clinic to assist geneticists in decision-making. In summary, this dissertation contributes to the development of more effective methods for predicting disease-causing variants and advancing precision medicine.