Filtering by
- All Subjects: Bioinformatics
- Creators: Kusumi, Kenro
- Creators: Runger, George C.
- Status: Published
The first topic of this dissertation focuses on the data clustering. Data clustering is often the first step for analyzing a dataset without the label information. Clustering high-dimensional data with mixed categorical and numeric attributes remains a challenging, yet important task. A clustering algorithm based on tree ensembles, CRAFTER, is proposed to tackle this task in a scalable manner.
The second part of this dissertation aims to develop data representation methods for genome sequencing data, a special type of high-dimensional data in the biomedical domain. The proposed data representation method, Bag-of-Segments, can summarize the key characteristics of the genome sequence into a small number of features with good interpretability.
The third part of this dissertation introduces an end-to-end deep neural network model, GCRNN, for time series classification with emphasis on both the accuracy and the interpretation. GCRNN contains a convolutional network component to extract high-level features, and a recurrent network component to enhance the modeling of the temporal characteristics. A feed-forward fully connected network with the sparse group lasso regularization is used to generate the final classification and provide good interpretability.
The last topic centers around the dimensionality reduction methods for time series data. A good dimensionality reduction method is important for the storage, decision making and pattern visualization for time series data. The CRNN autoencoder is proposed to not only achieve low reconstruction error, but also generate discriminative features. A variational version of this autoencoder has great potential for applications such as anomaly detection and process control.
In order to address this, multiple comparative genomics and bioinformatics analyses were conducted to elucidate patterns of evolution in the green anole and across multiple anole species. Comparative genomics analyses were used to infer additional X-linked loci in the green anole, RNAseq data from male and female samples were anayzed to quantify patterns of sex-biased gene expression across the genome, and the extent of dosage compensation on the anole X chromosome was characterized, providing evidence that the sex chromosomes in the green anole are dosage compensated.
In addition, X-linked genes have a lower ratio of nonsynonymous to synonymous substitution rates than the autosomes when compared to other Anolis species, and pairwise rates of evolution in genes across the anole genome were analyzed. To conduct this analysis a new pipeline was created for filtering alignments and performing batch calculations for whole genome coding sequences. This pipeline has been made publicly available.
I hypothesize that duplication events grant miRNA families with enhanced regulatory capabilities, specifically through distinct targeting preferences by family members. This has relevance for our understanding of vertebrate evolution, as well disease detection and personalized medicine. To test this hypothesis, I apply a conjunction of bioinformatic and experimental approaches, and design a novel high-throughput screening platform to identify human miRNA targets. Combined with conventional approaches, this tool allows systematic testing for functional targets of human miRNAs, and the identification of novel target genes on an unprecedented scale.
In this dissertation, I explore evolutionary signatures of 62 deeply conserved metazoan miRNA families, as well as the targeting preferences for several human miRNAs. I find that constraints on miRNA processing impact sequence evolution, creating evolutionary hotspots within families that guide distinct target preferences. I apply our novel screening platform to two cancer-relevant miRNAs, and identify hundreds of previously undescribed targets. I also analyze critical features of functional miRNA target sites, finding that each miRNA recognizes surprisingly distinct features of targets. To further explore the functional distinction between family members, I analyze miRNA expression patterns in multiple contexts, including mouse embryogenesis, RNA-seq data from human tissues, and cancer cell lines. Together, my results inform a model that describes the evolution of metazoan miRNAs, and suggests that highly similar miRNA family members possess distinct functions. These findings broaden our understanding of miRNA function in vertebrate evolution and development, and how their misexpression contributes to human disease.
Agassiz’s desert tortoise (Gopherus agassizii) is a long-lived species native to the Mojave Desert and is listed as threatened under the US Endangered Species Act. To aid conservation efforts for preserving the genetic diversity of this species, we generated a whole genome reference sequence with an annotation based on deep transcriptome sequences of adult skeletal muscle, lung, brain, and blood. The draft genome assembly for G. agassizii has a scaffold N50 length of 252 kbp and a total length of 2.4 Gbp. Genome annotation reveals 20,172 protein-coding genes in the G. agassizii assembly, and that gene structure is more similar to chicken than other turtles. We provide a series of comparative analyses demonstrating (1) that turtles are among the slowest-evolving genome-enabled reptiles, (2) amino acid changes in genes controlling desert tortoise traits such as shell development, longevity and osmoregulation, and (3) fixed variants across the Gopherus species complex in genes related to desert adaptations, including circadian rhythm and innate immune response. This G. agassizii genome reference and annotation is the first such resource for any tortoise, and will serve as a foundation for future analysis of the genetic basis of adaptations to the desert environment, allow for investigation into genomic factors affecting tortoise health, disease and longevity, and serve as a valuable resource for additional studies in this species complex.
Data Availability: All genomic and transcriptomic sequence files are available from the NIH-NCBI BioProject database (accession numbers PRJNA352725, PRJNA352726, and PRJNA281763). All genome assembly, transcriptome assembly, predicted protein, transcript, genome annotation, repeatmasker, phylogenetic trees, .vcf and GO enrichment files are available on Harvard Dataverse (doi:10.7910/DVN/EH2S9K).