Graduate School of Informatics

Graduate School of Informatics / Bioinformatics

In partial fulfillment of the requirements for the degree of Master of Science Alperen Taciroğlu will defend his thesis.

Title: ODYSSEY: A TOOL FOR MICRORNA-MRNA EXPRESSION AND INTERACTION VISUALIZATION

Date: 03th September 2018

Time: 14:00 PM

Place: A-212

Thesis Abstract : MicroRNAs (miRNAs) are non-coding short RNA molecules that are found in all metazoa studied so far. When distinct metazoa genomes considered up to 200 genes encode for unique miRNAs that show variability between species. Regulatory functions of miRNAs have been studied for 20 years starting after their discovery. The research suggests that they are involved in a wide spectrum of biological activities including apoptosis, tumorigenesis, development, homeostasis and viral infections. miRNAs regulate these cellular processes at the posttranscriptional level by binding to the messenger RNAs (mRNAs), leading to an unstable derivative of the initial biological molecule. miRNA targets are under strict evolutionary pressure which further implicates the importance of underlying biological mechanisms. Although there are several Gene/mRNA-miRNA interaction visualization and analysis tools "Odyssey" was built for improved interactive visualization the interaction network of miRNAs with along with their target expressions for a user uploaded dataset. It is built using Shiny package of the R programming language leading to seamless online access and modularity. In the end, I aim to provide users a user-friendly web-application which consists of modules that allows: uploading of their own data; performing differential expression (DGEx) analysis; and visualization of the network of which "Odyssey" builds from either experimentally validated or predicted interactions for individual miRNAs queried by the user. Odyssey further enables the user to filter selected nodes of the networks using fold change cut-offs obtained in DGEx step or expand the network using Gene Ontology (GO) terms to act as a strong predictor of the phenotype of interest for the user-specified biological data. Furthermore, the application has been demonstrated using two different public miRNA-mRNA expression datasets.

Graduate School of Informatics /Cognitive Sciences

In partial fulfillment of the requirements for the degree of Master of Science Mani Tajaddini will defend his thesis.

Title: RECURRENCE QUANTIFICATION ANALYSIS ON GROUP EYE TRACKING DATA

Date: 28nd August 2018

Time: 10:00 AM

Place: A-108

Thesis Abstract : Eye movements can provide insight into the cognitive processes behind them. To study group cognition through investigating eye movements, we have developed a software tool (GETapp) to collect eye movement data from groups of participants performing a task on a set of computers in the scope of the Group Eye Tracking (GET) paradigm. Like many real-world systems, the data from group eye tracking experiments are non-linear. We have developed a software tool, a package for the R programming language, called the generalRQA package for analyzing these non-linear data. The methods used in this package are based on recurrences of time-series and implement a part of the theory of Recurrence Quantification Analysis (RQA). In this thesis we describe both the software tools and apply them to propose measures that differentiate between different group formation patterns in a group eye tracking search task.

Graduate School of Informatics /Cognitive Sciences

In partial fulfillment of the requirements for the degree of Master of Science Efecan Yılmaz will defend his thesis.

Title: DEICTIC GAZE IN VIRTUAL ENVIRONMENTS

Date: 28nd August 2018

Time: 11:30 AM

Place: A-108

Thesis Abstract : The research in human-robot interaction (HRI) involve topics, such as interlocutor collaboration in joint action, deixis in HRI, or the properties of shared environments. Moreover, referring expressions are particularly studied in joint action from both expression generation and resolution perspectives. Selective visual attention in gaze interaction and saliency patterns are also active topics in HRI. The present thesis investigated in a virtual reality (VR) environment an HRI and joint action situation with the assistance of eye tracking in a head-mounted display device in order to explore the augmentation of non-verbal communication in HRI. For this purpose, we employed a multimodal approach in communication with both non-verbal deictic expressions and verbal references in a multi-robot agent, single human experiment setting. We also utilized two distinct robot agent designs to explore an interaction effect in robot agents’ design as we evaluated participants’ deixis resolution time and accuracy, as well as their gaze interaction patterns.

Graduate School of Informatics /Information Systems

In partial fulfillment of the requirements for the degree of Master of Science Alper Sarıkaya will defend his thesis.

Title: ANOMALY-BASED CYBER INTRUSION DETECTION SYSTEM WITH ENSEMBLE CLASSIFIER

Date: 06th September 2018

Time: 14:00 PM

Place: A-212

Thesis Abstract : Nowadays, cyber attacks are beginning to occur at an increasing rate. Along with this, diversity, size and density of the cyber attacks are increasing accordingly. When we analyze log in the security devices, we find that the massive amount of attacks sign was created. Besides, It is also difficult for human to evaluate the log accurately. Therefore, the identification of key data, which can be used to distinguish attack from this very large data set, is important for both rapid detection of attack and rapid response of security devices. In this study, we will focus on selection of appropriate data from the log via machine learning and determine the distinctive attributes specific to the attack in the selection of these data. Thus, we will research on the classification of the data to be used in the attack detection via machine learning.

Health Informatics department faculty member Assoc. Prof. Dr. Nurcan Tunçbağ  has been awarded UNESCO L'OREAL National For Women in Science Award in Life sciences category.

For related video, please click.

Title: APPLICATION OF BINARY PSO TECHNIQUE FOR THE SOLUTION OF MATHEMATICAL MODELLING OF PUBLIC CLOUD RESOURCES ALLOCATION SYSTEM FOR VIDEO ON DEMAND (VOD) APPLICATIONS

PhD Candidate: Betül Aygün

Program: Informations Systems Department 

Date: 03 September 14:00

Place: Conference Hall-01

Abstract: Video streaming services whether on demand or live has become one of the most popular services used recently. However, investments made for these type of applications cause a very serious financial problem just because video type of multimedia data needs more real time storage and high data transfer than other type of multimedia data. Furthermore, for the video streaming applications, significant amount of system resource in computing is required. To tackle this problem, cloud computing emerges as a preferred technology. Cloud services organizations are becoming more and more sophisticated as they enable the organizations to offer services without investing in hardware or software. A huge number of cloud service providers offer different pricing methods for various applications in various regions. For this reason, it is of great importance that incoming service requests are assigned to appropriate cloud services with minimum cost and maximum user satisfaction (QoS). Because of issues like multiple cloud providers, different quality of service requirements, different service level agreements (SLA) and uncertainties in demand, price and availability, optimization of resource allocation has some challenges. The objective of this study is to optimize the cost and performance of video on demand services using cloud CDNs, storage and transcoders based on QoS requirements of users. In this paper, Mixed Integer Quadratic Programming (MIQP) and different variants of Particle Swarm Optimization (PSO) algorithm are used to schedule video requests to cloud resources to achieve minimum cost of cloud services and maximum of user satisfaction. Due to the nature of the problem, it is not possible to use the classic PSO, but the new algorithms which combine Binary PSO with heuristics algorithms are proposed. These algorithms are compared with LP algorithms which gives best result. The results show that proposed algorithms yield better results than the benchmarking algorithms.

Graduate School of Informatics / Bioinformatics

In partial fulfillment of the requirements for the degree of Master of Science Tuğba Kaya will defend his thesis.

Title:STRUCTURAL MAPPING AND NETWORK ANALYSIS OF PATIENT-SPECIFIC MUTATIONS IN GLIOBLASTOMA

Date: 27th August 2018

Time: 14:00 PM

Place: A-212

Thesis Abstract :Cancer is one of the most common cause of death worldwide. It occurs as a result of a collection of somatic deviations from normal state. Therefore, many efforts has been invested to profile mutations in different types of tumors; such as, the Cancer Genome Atlas (TCGA) which deposits multiple omic data for more than 11,000 tumor samples. In this thesis, we present a pipeline which retrieves patient-specific mutation data in Glioblastoma from TCGA, maps these mutations on the protein structures in Protein Databank (PDB) and finds the location and functional effect of the mutations and reconstruct functional networks by integrating mutation data with interactome. As a result of this thesis study, we found that some mutations are specific to alternative isoform sequence of the protein instead of the canonical sequence. We also showed that functional impact of mutations in interface region is more damaging compared to the surface region and more similar to the core region of the protein. We showed that most common change in the protein core is that hydrophobic residues are mutated to another hydrophobic residue. However, in the surface or interface region a charged residue is changed either to another charged residue or a polar residue when we analyzed the chemical classes of mutations. From these mutation profiles of the patients, we reconstructed 290 GBM-specific networks with Omics Integrator which solves the prize-collecting Steiner forest (PCSF) problem and optimally connects the given set of proteins in a network context. We merged the most common nodes and edges across these patients and clustered the merged network into functional communities. The ontology and pathway enrichment analyses gave us that Wnt signaling, ERBB signaling and NfKb/Ikb signaling pathways are the most commonly enriched pathways. From mutation to protein structures and functional networks, we believe that the result of this thesis will have significant contribution in cancer research.

Graduate School of Informatics /Information Systems

In partial fulfillment of the requirements for the degree of Master of Science Ayşe Begüm Kılıç will defend his thesis.

Title: INVESTIGATING THE TRANSFORMATION OF SUPPLY CHAINS INTO DIGITAL SUPPLY CHAINS BY USING SOFT SYSTEMS METHODOLOGY: A CASE STUDY

Date: th August 2018

Time: 13:30 AM

Place: A-212

Thesis Abstract :Currently, Industry 4.0 is trending topic among researchers. Basically, it promises firms higher profits by using technologies that are more advanced. In order to benefit from Industry 4.0, digitization of all company operations is essential for companies. That is why today, examining the transformation into the digital world has a significant role for the future. This thesis aims to investigate the transformation of supply chains into digital supply chains (DSC) in firm level by using Soft Systems Methodology (SSM). SSM enables researchers to analyze and define a problem to find out its real causes and suggest changes to solve it. SSM does not require quantitative data for analysis and is suitable for not fully defined and not well-known problems. Since the changes in the shifting from regular supply chains to DSC are not defined with numeric data and it is hard to define all the expected transformation clearly, the SSM methodology is chosen to follow. A case study applied to a firm will be conducted via interview technique within the thesis. The SSM is expected to be beneficial on analyzing the subjective data that will be provided by these interviews. At the end of the thesis, it is predicted to define the real-world transformation process and offer suggestions to firms in order to cope with the change

Graduate School of Informatics /Information Systems

In partial fulfillment of the requirements for the degree of Master of Science Güliz Bulut will defend his thesis.

Title: AUTOMATIC INFORMATION COVERAGE ASSESSMENT OF DIABETES WEBSITES

Date: 3rd September 2018

Time: 10:00 AM

Place: A-212

Thesis Abstract :People frequently access Internet to look up health information. However, as the quality of websites may vary significantly, the treatment recommendations and guidelines provided by some of these web sites may be fallacious. Consequently, patients may unfollow their current treatments suggested by their doctors or start following unfounded treatments. In this thesis, an automated approach is presented to estimate information coverage of websites. The approach is based on a domain-dependent standard knowledge base (KB) and enhanced by open source resources. Elastic net regularized regression is used to construct a model for estimation. As a case study, corpus of type 2 diabetes related web pages is selected. “Standards of Medical Care in Diabetes” published by American Diabetes Association is processed to obtain factual data about treatment of type 2 diabetes. This standard serves as a detailed knowledge base on type 2 diabetes treatment and enables to produce a trustworthy input for evaluation. In light of this KB, the corpus of type 2 diabetes related web pages is processed to retrieve their coverage of factual information. It is observed that, extracting significant terms from a domain-dependent knowledge base provide a basis to measure information coverage of a source

Title: TRANSFER AND MAINTENANCE EFFECTS OF N-BACK WORKING MEMORY TRAINING IN INTERPRETING STUDENTS: A BEHAVIOURAL AND OPTICAL BRAIN IMAGING STUDY

PhD Candidate: Asiye Öztürk

Program: Cognitive Science Department 

Date: 29 August 13:00

Place: Conference Hall-01

Abstract:Working memory training is seen as an effective tool for enhancing performance during a variety of high level cognitive tasks for different groups such as children, older people, and individuals with cognitive problems as well as practitioners of highly cognitive demanding professions. Although there has been some controversy regarding the efficiency of working memory training interventions, it has been shown that n-back working memory training results in improvements in working memory capacity as well as in reasoning skills which points out that it yields both near and far transfer effects. Considering the crucial role of working memory in interpreting profession, this thesis targeted interpreting students in order to investigate transfer and maintenance effects of n-back working memory training through a comprehensive analysis of its near, moderate and far transfer effects as well as possible transfer to consecutive interpreting. Combining behavioural data collection methods and fNIRS (functional near-infrared spectroscopy) as an optical brain imaging technique, the thesis was designed as a longitudinal study in which participants completed a series of tests before and after the training as well as three months after the completion of the training sessions. Compared to an active control group, n-back group had larger performance gains in near and far transfer tasks, and more importantly in consecutive interpreting scores as a result of enhanced working memory capacity, suggesting that common working memory processes are employed in the tasks. Significant neural activity patterns observed in prefrontal cortex reflect the shared cognitive processes of n-back training and consecutive interpreting, which is supported by performance improvements in consecutive interpreting and increased neural efficiency. Therefore, it is believed that findings of this thesis will have original contributions to the field.