Refine
Document Type
- Doctoral Thesis (1)
- Preprint (1)
Language
- English (2)
Has Fulltext
- yes (2)
Keywords
- Ontology (2) (remove)
Faculty / Organisational entity
One of the main tasks of molecular biology is understanding the mechanisms of molecular biological processes. This brings the problem of creating regulatory networks and therefore finding key regulators. In order to do it, it is important to have such representation of the data that can reveal the distinct patterns within the big groups. On one side, there are numerous experimentally determined kinetic information about the alteration of molecular presence in the observed system. On the other side, there are documented throughout the years evidences of the involvement of molecules in different biological processes. Both sources of the information have their drawbacks: experimental data reflect only a fleeting molecular state of each individual organism and therefore are often high-variant and noisy; functional groups were determined as generalization of known roles of molecules in biological processes and therefore can be not complete and only partially relevant to certain experimental conditions and individual organisms. Our goal is to get the overview of the experimentally observed molecules and extract the knowledge from both sources, avoiding constrains of noise distractions and generalization bias. The resulted optimal representation of the experimental data then would help to pinpoint potential regulators.
The proposed method is called the Signature Topology (ST) approach, as it uses the functional topology as the prior knowledge source and creates a specific signature for the given experimental data. The ST approach is based on knowledge-and-data-driven machine learning algorithm, that is implemented via a dynamic programming approach. Based on both prior knowledge and learning from the data, the proposed approach represents a combination of supervised and unsupervised machine learning. The resulting network structure deals with data abundance and avoids an over-detailed description that may lead to misinterpretation and is able to pick out elements with minor behavior patterns.
The method is tested with artificial data and applied to real-world mass-spectrometry proteome data and NGS-transcriptome data of Chlamydomonas reinhardtii. The proposed approach helps with identification of the potential regulatory genes, whose roles are not explicitly provided in the used functional ontology. Moreover, it shows a successful reduction in data complexity while preserving all individual molecular information reported in the literature and stored in the functional ontology. If the proposed approach analyzes different experimental data with the same ontology, the resulting networks are uniform and therefore can be compared. That gives an opportunity to compare between a great variety of experimental conditions, from different organisms to different
system levels.
The term enterprise modelling, synonymous with enterprise engineering, refers to methodologies developed for modelling activities, states, time, and cost within an enterprise architecture. They serve as a vehicle for evaluating and modelling activities resources etc. CIM - OSA (Computer Integrated Manufacturing Open Systems Architecture) is a methodology for modelling computer integrated environments, and its major objective is the appropriate integration of enterprise operations by means of efficient information exchange within the enterprise. PERA is another methodology for developing models of computer integrated manufacturing environments. The department of industrial engineering in Toronto proposed the development of ontologies as a vehicle for enterprise integration. The paper reviews the work carried out by various researchers and computing departments on the area of enterprise modelling and points out other modelling problems related to enterprise integration.