京都大学 / 0048 / 新制・課程博士 / 博士(農学) / 甲第22898号 / 農博第2441号 / 新制||農||1083(附属図書館) / 学位論文||R3||N5318(農学部図書室) / 京都大学大学院農学研究科応用生物科学専攻 / (主査)教授 松浦 健二, 教授 大門 高明, 教授 寺内 良平 / 学位規則第4条第1項該当 / Doctor of Agricultural Science / Kyoto University / DFAM

Background: Scientists are measuring multiple sources of massive, complex, and diverse data in hopes to better understand the principles underpinning complex phenomena. Sophisticated statistical and computational methods that reduce data complexity, harness variability, and integrate multiple sources of information are required. The ‘sparse’ class of multivariate statistical methods is becoming a promising solution to these data-driven challenges, but lacks application, testing, and development. Methods: In this thesis, efforts are three-fold. Sparse principal component analysis (sparse PCA) and sparse canonical correlation analysis (sparse CCA) are applied to a large toxicogenomic database to uncover candidate genes associated with drug toxicity. Extensive simulations are conducted to test and compare the performance of many sparse CCA methods, determining which methods are most accurate under a variety of realistic, large-data scenarios. Finally, the performance of the non-parametric bootstrap is examined, determining its ability to generate inferential measures for sparse CCA. Results: Through applications, several groups of candidate genes are obtained to point researchers towards promising genetic profiles of drug toxicity. Simulations expose one sparse CCA method that outperforms the rest in the majority of data scenarios, while suggesting the use of a combination of complimentary sparse CCA methods for specific data conditions. Simulations for the bootstrap conclude the bootstrap to be a suitable means for inference for the canonical correlation coefficient for sparse CCA but only when sample size approaches the number of variables. As well, it is shown that aggregating sparse CCA results from many bootstrap samples can improve accuracy of detection of truly cross-correlated features. Conclusions: Sparse multivariate methods can flexibly handle challenging integrative analysis tasks. Work in this thesis has demonstrated their much-needed utility in the field of toxicogenomics and strengthened our knowledge about how they perform within a complex, massive data framework, while promoting the use of bootstrapped inferential measures. / Thesis / Doctor of Philosophy (PhD) / Due to rapid advances in technology, many areas of scientific research are measuring multiple sources of massive, complex, and diverse data in hopes to better understand the principles underpinning puzzling phenomena. Now, more than ever, advancement and discovery relies upon sophisticated and robust statistical and computational methods that reduce the data complexity, harness variability, and integrate multiple sources of information. In this thesis, I test and validate the ‘sparse’ class of multivariate statistical methods that is becoming a promising, fresh solution to these data-driven challenges. Using publicly available data from genetic toxicology as motivation, I demonstrate the utility of these methods, find where they work best, and explore the possibility of improving their scientific interpretability. The work in this thesis contributes to both biostatistics and genomic literature, by meshing together rigorous statistical methodology with real-world data applications.

The rapid accumulation of numerous sequenced genomes has provided a golden opportunity for ancestral state reconstruction studies, especially in the whole genome reconstruction area. However, most ancestral genome reconstruction methods developed so far only focus on gene or replicon sequences instead of whole genomes. They rely largely on either detailed modeling of evolutionary events or edit distance computation, both of which can be computationally prohibitive for large data sets. Hence, most of these methods can only be applied to a small number of features and species. In this dissertation, we describe the design, implementation, and evaluation of an ancestral genome reconstruction system (REGEN) for bacteria. It is the first bacterial genome reconstruction tool that focuses on ancestral state reconstruction at the genome scale instead of the gene scale. It not only reconstructs ancestral gene content and contiguous gene runs using either a maximum parsimony or a maximum likelihood criterion but also replicon structures of each ancestor. Based on the reconstructed genomes, it can infer all major events at both the gene scale, such as insertion, deletion, and translocation, and the replicon scale, such as replicon gain, loss, and merge. REGEN finishes by producing a visual representation of the entire evolutionary history of all genomes in the study. With a model-free reconstruction method at its core, the computational requirement for ancestral genome reconstruction is reduced sufficiently for the tool to be applied to large data sets with dozens of genomes and thousands of features. To achieve as accurate a reconstruction as possible, we also develop a homologous gene family prediction tool for preprocessing. Furthermore, we build our in-house Prokaryote Genome Evolution simulator (PEGsim) for evaluation purposes. The homologous gene family prediction refinement module can refine homologous gene family predictions generated by third party de novo prediction programs by combining phylogeny and local gene synteny. We show that such refinement can be accomplished for up to 80% of homologous gene family predictions with ambiguity (mixed families). The genome evolution simulator, PEGsim, is the first random events based high level bacteria genome evolution simulator with models for all common evolutionary events at the gene, replicon, and genome scales. The concepts of conserved gene runs and horizontal gene transfer (HGT) are also built in. We show the validation of PEGsim itself and the evaluation of the last reconstruction component with simulated data produced by it. REGEN, REconstruction of GENomes, is an ancestral genome reconstruction tool based on the concept of neighboring gene pairs (NGPs). Although it does not cover the reconstruction of actual nucleotide sequences, it is capable of reconstructing gene content, contiguous genes runs, and replicon structure of each ancestor using either a maximum parsimony or a maximum likelihood criterion. Based on the reconstructed genomes, it can infer all major events at both the gene scale, such as insertion, deletion, and translocation, and the replicon scale, such as replicon gain, loss, and merge. REGEN finishes by producing a visual representation of the entire evolutionary history of all genomes in the study. / Ph. D.

The goal of this research is to construct causal gene networks for genetical genomics experiments using expression Quantitative Trait Loci (eQTL) mapping and Structural Equation Modeling (SEM). Unlike Bayesian Networks, this approach is able to construct cyclic networks, while cyclic relationships are expected to be common in gene networks. Reconstruction of gene networks provides important knowledge about the molecular basis of complex human diseases and generally about living systems. In genetical genomics, a segregating population is expression profiled and DNA marker genotyped. An Encompassing Directed Network (EDN) of causal regulatory relationships among genes can be constructed with eQTL mapping and selection of candidate causal regulators. Several eQTL mapping approaches and local structural models were evaluated in their ability to construct an EDN. The edges in an EDN correspond to either direct or indirect causal relationships, and the EDN is likely to contain cycles or feedback loops. We implemented SEM with genetics algorithms to produce sub-models of the EDN containing fewer edges and being well supported by the data. The EDN construction and sparsification methods were tested on a yeast genetical genomics data set, as well as the simulated data. For the simulated networks, the SEM approach has an average detection power of around ninety percent, and an average false discovery rate of around ten percent. / Ph. D.

In the post-genomic era, one of the most important goals for the community of plant biologists is to take full advantage of the knowledge generated by the Arabidopsis thaliana genome project, and to employ state-of-the-art functional genomics techniques to assign function to each gene. This will be achieved through a complete understanding of what all cellular components do, and how they interact with one another to produce a phenotype. Among the proteins encoded by the Arabidopsis genome are 24 related carboxyl methyltransferases that belong to the SABATH family. Several of the SABATH methyltransferases convert plant hormones, like jasmonic acid, indole-3-acetic acid, salicylic acid, gibberellins, and other plant constituents into methyl esters, thereby regulating the biological activity of these molecules and, consequently, myriad important physiological processes. Our research aims to decipher the function of proteins belonging to the SABATH family by applying a combination of genomics tools, including genome-wide expression analysis and gas-chromatography coupled with mass spectrometry-based metabolite profiling. Our results, combined with available biochemical information, provide a better understanding of the physiological role of SABATH methyltransferases, further insights into secondary plant metabolism and deeper knowledge of the consequences of modulating the expression of SABATH methyltransferases, both at the genome-wide expression and metabolite levels. / Ph. D.

Chytridiomycosis is a fungal skin disease in amphibians that is primarily caused by Batrachochytrium dendrobatidis (Bd). We analyzed whole genome sequences of 40 bacterial isolates that had been previously cultured from the skin of four amphibian species from Virginia, USA, and tested for their ability to inhibit Bd growth via an in vitro challenge assay. These 40 isolates spanned 11 families and 13 genera. The aim of this study was to identify genomic differences among the amphibian skin bacterial isolates and generate hypotheses about possible differences that could contribute to variation in their ability to inhibit the growth of Bd. We identified sixty-five gene families that were present in all 40 isolates. We also looked for the presence of biosynthetic gene clusters. While this set of isolates contained a wide variety of biosynthetic gene clusters, the two most abundant clusters with potential anti-fungal activity were siderophores (N=17) and Type III polyketide synthases (N=20). We then analyzed the isolates belonging to the phylum Proteobacteria in more detail. We identified 197 gene families that were present in all 22 Proteobacteria. We examined various subsets of the Proteobacteria for genes for specific compounds with known activity against fungi, including chitinase and violacein. We identified a difference in the number, as well as amino acid sequences, of predicted chitinases found in two isolates belonging to the genus Agrobacterium that varied in their inhibition of Bd. After examining the annotated genomes, we identified a predicted chitinase in a Sphingomonas isolate that inhibited the growth of Bd that was absent from the five Sphingomonas isolates that did not inhibit Bd growth. The genes vioA, vioB, vioC, vioD and vioE are necessary to produce violacein, a compound which inhibits the growth of Bd. Differences in these genes were identified in three out of the four Janthinobacterium isolates. Of these three isolates, two showed strong inhibition of Bd growth, while the third inhibited Bd growth to a lesser extent. Using comparative genomics, we generated several testable hypotheses about differences among bacterial isolates that could contribute to variation in ability to inhibit Bd growth. Further work is necessary to test the various mechanisms utilized by amphibian skin bacterial isolates to inhibit Bd. / Master of Science / Many amphibian population declines around the world have been caused by chytridiomycosis, a skin disease. This disease is caused by the fungus Batrachochytrium dendrobatidis (Bd). The skin of amphibians is also home to many bacteria that can provide important functions for the amphibian host, like preventing infection by Bd. To understand how these bacteria might provide protection, we examined the entire genomes of 40 bacterial isolates that reside on the skin of four amphibian species from Virginia, USA. These bacteria were previously tested for their ability to prevent Bd growth and 40 of them were chosen for sequencing based on selecting closely related isolates that varied in their ability to inhibit Bd growth. This allowed us to compare their genomes and generate hypotheses about possible genomic differences that could contribute to the variation in Bd growth inhibition. We identified sixty-five gene families that were present in all 40 bacteria. We also looked for sets of genes (biosynthetic gene clusters) that are known to produce secondary metabolites, which are compounds that can include antifungals. The two most abundant clusters we identified that had the potential to produce compounds that inhibit fungal growth were siderophores and Type III polyketide synthases. We then looked for genes that were not part of biosynthetic gene clusters that could produce specific compounds that can inhibit Bd growth, such as chitinase and violacein. We found variation in chitinase genes in several isolates that seemed to be associated with the ability to inhibit Bd growth. In addition, there were some differences in violacein genes that should be examined more in future studies. Overall, we suggest that using comparative genomic approaches can be valuable for identifying key bacterial functions in the microbiome.

Temperate tree species exhibit seasonal growth cycling, and the timing of such transition varies with local climate. Under anthropogenic climate change, the local pattern of growth and dormancy in tree populations is expected to become uncoupled with shifting seasonal environmental signals, particularly temperature. Thus, an understanding of the genetic underpinnings of local adaptation is key to predicting the fate of tree populations in the future. In this thesis, we coupled sampling of range-wide natural accessions of P. trichocarpa with adaptive trait phenotyping and genome-wide genotyping to uncover relationships between genotype, phenotype, and environment. We detected strong correlations between adaptive phenotypes, climate, and geography, which suggested climatic selection driving adaptation of these populations to local environments. We subsequently combined genotype-phenotype association tests with sliding window analysis and identified regions strongly associated with these adaptive traits. We also compared adaptive markers identified in two independent GWAS on samples across latitude and altitude transects and found a set of associated variants shared across both transects. We further scanned the genome with three selection tests to identify regions showing evidence of recent positive and divergent selection. By comparing candidate selection regions across altitude and latitude, we detected a set of overlapping regions showing differentiation across gradients of the same climate variables. We validated the functional imortance of these selection regions by combining GWAS and showed that selection regions contain a strong signature of phenotypic associations. We also studied the distribution of deleterious allels across genome and natural populations, and found that deleterious alleles preferentially accumulate in regions of low recombination and hithihking regions. Finally, marginal populations contained more deleterious alleles compared with central populations, which is likely due to ineffective selection in small populations and recent bottlenecks associated with postglacial recolonization. These findings provide new insights into the genomic architecture underlying climatic adaptation and how selection drives adaptive evolution of tree species. / Ph. D.

The mechanism behind magnetoreception – the ability to sense magnetic fields for orientation and navigation – still remains one of the most difficult questions to answer in sensory biology, with fish being just one of many taxa known to possess this sense. Characterizing a magnetic sense in fish is crucial for understanding how they navigate their environment and can inform on how increasing anthropogenic sources of electromagnetic fields in aquatic environments may affect threatened fish species. This study examined the hypothesis put forth by Natan and Vortman (2017) that magnetotactic bacteria (MTB), bacteria that create their own chains of magnetic particles for navigational use, act in symbiosis with their animal host to convey magnetic information about their surroundings. Utilizing existing, publicly available datasets of raw genomic sequences, this study demonstrated the presence of MTB within a diverse array of fishes and identified differences in species diversity of MTB between freshwater and marine species of fish. Future research aimed at identifying MTB in specific fish tissues, such as the eye and other neural tissues, will be necessary to provide support for this hypothesis and to further examine the relationships that MTB may have with magnetically sensitive animals.

<p dir="ltr">Open-access bioinformatics data enables accessible public health research for a variety of stakeholders, including teachers and low-resourced researchers. This project outlines two case studies utilizing open-access bioinformatics data sets and analysis software as proofs of concept for the types of research projects that can be adapted for workforce development purposes. The first case study is a spatial temporal analysis of Lyme disease rates in the United States from 2008 to 2020 using freely available data from the United States Department of Agriculture and Centers for Disease Control and Prevention to determine how urbanization and other changes in land use have impacted Lyme disease rates over time. The second case study conducts a pangenome analysis using bacteriophage data from the Actinobacteriophage Database to determine conserved gene regions related to host specificity.</p>

This work describes the development and application of genomic DNA tiling arrays: microarrays designed to represent all of the DNA comprising a chromosome or other genomic locus, regardless of the genes that may be annotated in the region of interest. Because tiling arrays are intended for the unbiased interrogation of genomic sequence, they enable the discovery of novel functional elements beyond those described by existing gene annotation. This is of particular importance in mapping the gene structures of higher eukaryotes, where combinatorial exon usage produces rare splice variants or isoforms expressed in low abundance that may otherwise elude detection. Issues related to the design of both oligonucleotide- and amplicon-based tiling arrays are discussed; the latter technology presents distinct challenges related to the selection of suitable amplification targets from genomic DNA. Given the widespread fragmentation of mammalian genomes by repetitive elements, obtaining maximal coverage of the non-repetitive sequence with a set of fragments amenable to high-throughput polymerase chain reaction (PCR) amplification represents a non-trivial optimization problem. To address this issue, several algorithms are described for the efficient computation of optimal tile paths for the design of amplicon tiling arrays. Using these methods, it is possible to recover an optimal tile path that maximizes the coverage of non-repetitive DNA while minimizing the number of repetitive elements included in the resulting sequence fragments. Tiling arrays were constructed and used for the chromosome- and genome-wide assessment of human transcriptional activity, via hybridization to complementary DNA derived from polyadenylated RNA expressed in normal complex tissues. The approach is first demonstrated with amplicon arrays representing all of the non-repetitive DNA of human chromosome 22, then extended to the entire genome using maskless photolithographic DNA synthesis technology. A large-scale tiling array survey revealed the presence of over 10,000 novel transcribed regions and verified the expression of nearly 13,000 predicted genes, providing the first global transcription map of the human genome. In addition to those likely to encode protein sequences on the basis of evolutionary sequence conservation, many of the novel transcripts constitute a previously uncharacterized population of non-coding RNAs implicated in myriad structural, catalytic and regulatory functions.