Abstract
dc:descriptionRecent advances in technology have produced a wealth of digital machines and sensors which, along with recent advancements in biotechnology and more specifically the high-throughput sequencing methods, have led to an unprecedented explosion of data on every aspect of biology. Consequently, knowledge discovery and machine learning are today, more than ever, crucial for the intelligent analysis of biological data, knowledge discovery and ultimately the answering of fundamental questions from biology and medicine. The scope of this thesis is the knowledge discovery from biological data. This thesis belongs to the research fields of knowledge discovery from databases, machine learning and bioinformatics, which is a general term used to describe any type of computational analysis applied to biological data. The contribution of this thesis lays on the knowledge discovery and machine learning field. More specifically, the aim of this thesis is to create or extend methods for the analysis of biological data and later to apply them to extract valuable knowledge from data. Moreover, a subsequent aim is to utilize and apply the methods, tools and knowledge produced, in research programs or in clinical practice. Initially, the thesis focuses on the analysis of population genomic data and more specifically single nucleotide polymorphisms (SNPs) data, which main feature is the high dimensionality, mainly focusing on the problem of selecting the most informative markers for assigning individuals to populations of origin. Α new method for feature selection based on frequent itemsets theory is presented, which achieves much superior results compared to the existing methods is also presented. Moreover, all well-known feature selection algorithms in the field, as well as algorithms for SNP dataset manipulation are presented. In the same area of population genetics, a microsatellite pattern discovery algorithm is developed along with the software application in which it is integrated. Later, methodologies which integrate different immunogenetic and clinicobiological data sources and analyze them to study the patterns of mutations that occur through Somatic Hypermutation (SHM) are developed. All methods were applied to patient data with Chronic Lymphocytic Leukemia (CLL). Moreover, a methodology based on social choice and voting theory is presented to investigate the potential ontogenetic transformation of genes towards other genes or gene families, through SHM. Finally, a method for the polyadenylation site prediction in RNA sequences is also developed. This is a modular method consisting of two parts, the one based on interesting emerging patterns and the other on a distance based scoring method, achieving high score of adjusted accuracy.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2016
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kavakiotis, Ioannis
- Καβακιώτης, Ιωάννης
Subjects
dc:subject × 14Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/39372