Deutschland online bookmaker 100% Bonus.

Download Template Joomla 3.0 free theme.


Thesis Defenses 2010

Finding Dense Subgraphs in Protein Interaction Network with Soft Computing

By: Hamid Ravaee
Supervisors: Dr. Ali Moeini, Dr. Ali Masoudinejad
Advisor: Dr. Dara Moazzami

Defense room: Engineering Science 1
Committee: Dr. Kambiz Badie, Dr. Ali Movaghar
Date: Sunday, March 7, 2010

Abstract: Clustering protein-protein interaction network aims to find functional modules and protein complexes. There are many computational graph clustering methods that are used in this field, but few of them are intelligent computational methods. In this thesis, we present a novel improved immune genetic algorithm to find dense subgraphs based on efficient vaccination method, variable-length antibody schema definition and new local and global mutations.
Primary processes within a cell are carried out by collaborations among different proteins, which are shown by protein-protein interactions (PPI) networks. Metabolic pathways, for instance, consist of several proteins, that are called enzymes, conduct a chain of chemical reactions for altering a form of chemical substance into the other forms, namely products. Proteins interactions happen also in signaling pathways where a set of proteins, by an ordered sequence of reactions, try to convert a type of chemical signal to other from, and they enable a cell to obtain environmental information rapidly. Proteins interactions can be found in any sort of biological processes within cells. Indeed, existence of these interactions result in a cell to function, to flourish and more importantly to survive. In recent years, the advances in the high-throughput PPI detection technology have produced a high volume of PPI datasets freely available and it makes great opportunity for investigations .
Clustering is the process of grouping data into sets (clusters), which there is more similarity between the objects in the same clusters than they were in different clusters. Clustering analysis seeks a set of clusters based on similarity between pairs of elements. The main goal is to find clusters, in which two criteria are satisfied: first, "homogeneity" that means objects placed in a same cluster are highly similar to each other. Secondly, "separation" means elements in different clusters have low similarity to each other.

To pave a secure and effective in AD-HOC Networks

By: Mohammad Zarei
Supervisors: Dr. Dara Moazzami
Advisor: Dr. Ali Moeini

Defense room: Engineering Science 1
Committee: Dr. Shahriar Heshmati, Dr. Hassan Yousefi Azari
Date: Monday, September 20, 2010

Abstract: By appearance of wireless communication between different subjects and at last the wireless and mobile networks, it took the attention of many of scientists in computer field to the problems existed in the network the same as routing and so on. But these networks ware not sufficient of adequate for all demands in the wireless communication field.
So for this reason a special network were designed that the communication were accomplished by means of radio transmitters and receivers with limited distances. Although, There were not any central construction for routing and management. In the next step the limited power consumption and operation were added to the above models, and the economic networks were introduced.
The ad-hoc networks have a wide application the same as sensors for fire alarms in jungles and cities as well as nuclear emotion or radiation leakage in a nuclear reactors are the samples of these usages.
The properties or specifications of ad-hoc networks systems can be divided into following subjects.

  1.  imited energy of substances.
  2.  Limited band wide.
  3. The networks with no time variation construction.
  4. Low communications quality.
  5. Low computation power or ability in subjects.

Of the problems considered in ad-hoc networks are the routing in them. The different algorithms for these sort of problems has been introduced that the land algorithms has the especial importance among the others.

Transcriptional regulatory network analysis of histone post–translational modifications in computational epigenetics

By: Hadi Jorjani
Supervisors: Dr. Ali Moeini
Co-advisor: Dr. Ali Masoudinejad

Defense room: Engineering Science 1
Committee: Dr. Kambiz Badie, Dr. Abbas Nozari
Date: Monday, September 20, 2010

Abstract: As we know eukaryotic cells in an organism all have the same genetic code (genotype) but they differ in expressing this same genetic material which leads to forming various cells which have different functionalities (different phenotypes). These mechanisms that cause cells to express their genes differently without changing the genetic material are called epigenetic mechanisms. Epigenetic modifications in every eukaryotic cell are hypothesized to be specific which is also called epigenetic code. Epigenetic code consists of modifications in histone tails (also known as histone code) and other additional modifications such as DNA methylation. These epigenetic codes play an important role in gene expression and are of great importance in process of cellular differentiation. Actually the first step in gene expression is transcription and in fact transcriptional machinery must have access to DNA which necessitates the chromatin state to be more open in genomic loci. Chromatin remodeling is controlled by the pattern of histone modifications such as acetylation, methylation, phosphorilation, etc. This pattern is called histone code which regulates transcription through alterations in chromatin structure and generation of binding sites for chromatin-interacting factors. It has been shown that transcription factors’ binding activity is correlated with histone modifications specifically histone acetylation patterns. Modeling the regulatory network which describes these relations between TFs and histone codes and gene transcription is the ultimate goal in deciphering histone codes.
In this thesis we are going to model the regulatory network which controls acetylation levels of 11 histone lysines, as part of histone code, in Saccharomyces cerevisiae. We are trying to find TFs which have greater role in regulating acetylation level of specific histone lysines. We have utilized LASSO regression to find important transcription factors which regulate specific lysine acetylation. LASSO inputs are transcription factors’ binding and level of acetylation for a specific lysine in all promoters of Saccharomyces cerevisiae. After building the network for each lysine and its regulators we integrated these networks to reach a whole network which illustrates an exhaustive view of these regulatory mechanisms for all 11 lysines and we also characterized some features of this network such as degree distribution which follows the power-law degree distribution.

Text Mining Using the Ontology Containing Key Concepts of Text

By: Mina Maleki Majd
Supervisors: Dr. Kambiz Badie, Dr. Dara Moazzami

Defense room: Engineering Science 1
Committee: Dr. Ali Moeini, Dr. Mehrnoosh Shams Fard
Date: Tuesday, September 21, 2010

Abstract: Textual data are particularly important and automatic text mining can extract knowledge that is hidden within them. By the formal model of ontology, Meaning in the text can be connected to concepts which are not in the text. With the framework which offered in this thesis a formal model for text mining will be introduced. Also use of formal model ontology will be uniformed and standardized in presenting of background knowledge, preprocessing steps, knowledge discovery and other steps according to proposed framework. The proposed framework prepares the text for the main idea of this thesis in order to complete the goals of text mining with ontology. The idea proposed in the current project is the best using of ontology in activation the key concepts existing in the text which have special generalization capability. Using this generalization capability causes these concepts to have the ability of being converted to abstract enough concepts. The proposed ontology has a set of concepts and properties related to it that by existing of these properties the text will be able to be checked according to already mentioned key concepts. The proposed framework and algorithm will be examined in Biomedical domain on texts related to patients by which will demonstrate the use of results obtained from the algorithm for health policies.