7
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Discipline
Yapay öğrenme bazlı hesaplamalı modelleme ile geniş çaplı kanser hücre hattı ilaç yanıt tahmini
Her hasta için özel olarak en iyi tedavi seçeneğini değerlendirmek, hassas tıbbın ana hedefidir. Özellikle farklı kanser tiplerinde, aynı tanıya sahip hastalar genetik heterojenite nedeniyle uygulanan tedaviye farklı seviyelerde duyarlılık gösterebilirler. İlaç yanıtlarını (duyarlılığını) önceden tahmin ederek ilaç geliştirme süreçleri için gereken süreden tasarruf etmek ve etkisiz ilaçların uygulanmasını önlemek amacıyla, hastaların genetik özelliklerini kullanan hesaplamalı yaklaşımlar geliştirilmiştir. Bu tez çalışmasında, kanser hücrelerinin ilaç yanıtlarını tahmin eden makine öğrenmesi tabanlı bir sistem olan DeepResponse-RF önerilmiştir. DeepResponse-RF, büyük ölçekli profilleme/tarama projelerinden elde edilen ve her biri ayrı bir tümörü temsil eden farklı kanser hücre hatlarının gen ekspresyonu, mutasyon, kopya sayısı varyasyonu ve metilasyon profillerini, ilaçların moleküler özellikleriyle birlikte kullanmaktadır ve rastgele orman algoritması aracılığıyla, tümörün multi-omik özellikleri ile uygulanan ilaca duyarlılığı arasındaki ilişkiyi yapay olarak öğrenmektedir. Performans sonuçları, DeepResponse-RF'nin kanser hücrelerinin ilaç duyarlılığını başarılı bir şekilde tahmin ettiğini ve özellikle multi-omik yönün öğrenme sürecine fayda sağladığını ve tek omik tabanlı duruma kıyasla daha iyi performansa yol açtığını gösterdi. DeepResponse-RF, daha ileri seviyede geliştirme aşamalarının uygulanması sonrasında, yeni ilaç adaylarının erken aşamada keşfedilmesi ve mevcut olanların dirençli tümörlere karşı yeniden konumlandırılması için kullanılabilir.
Bayesian networks for omics data analysis in hepatocellular carcinoma single-cell sequencing
Single cell multi omics techniques have shown an advancement in unrevealing complex diseases like cancer heterogeneity by providing multi-faceted insight into their individual cellular regulations. In this study, a machine learning approach, Bayesian network (BN), has been applied to integrate genomic, epigenomic, and transcriptomic data in hepatocellular carcinoma at single cell resolution. Hepatocellular carcinoma (HCC) is the most common type of liver cancer with a high metastatic rate and reckoned for poor prognosis. Heterogeneity of tumor cells is concerned with cancer progression, metastasis, therapeutic resistance, and mortality. For this purpose, a dataset from a published study of 25 single cell sequencing of hepatocellular carcinoma were used. First, DNA methylome and transcriptome data were analyzed on their own. Copy number variations were estimated from DNA methylome data by using the Hidden Markov Model method. To reveal the causal relationship between the omics, three BN models were constructed. The models were fitted to their parameters by using maximum likelihood estimation. For model evaluation, score-based criteria, Akaike information criterion and Bayesian information criterion, were used. 207 genes with significant models have been detected. The heterogeneity of the omics and their regulation mechanisms with each other have been shown, by pointing to genes that follow different BN models that take place in major pathways in HCC.
Prcell: sentetik biyoloji için sinyal yolağı programlamaya yönelik bir yazılım
Sentetik biyoloji, mühendisliğin temel ilkelerinin biyolojiye, özellikle de hücreye uygulanmasına yönelik yeni ve disiplinlerarası bir araştırma alanıdır. Yapay biyolojik sistemler tasarlamayı ve üretmeyi amaçlar. Sentetik biyoloji, hücresel işlevleri değiştirmek veya düzeltmek, hücrenin çevresel koşullara yanıt vermesini sağlamak veya hücresel gelişimi yönetmek için tasarlanıp uygulanabilen yeni sinyal yollarını tasarlamak veya ayarlamak için çok önemlidir. Araştırmacıları biyolojik devre tasarımının elektronikteki entegre devrelerin tasarımına benzeyeceği ve genetik mühendisliğinin yeteneklerini artıracağı bir gelecek bekliyor. Bir hücre programlama dili olarak geliştirilen uygulamaların amacı, canlı hücreler için genetik devreler tasarlamak ve hücrede tasarlanmış genetik devreyi gerçekleştirmektir. Bunlardan biri, elektronik devre tasarım programına dayanan, verilog olan ve elektronik devreleri tasarlamak için mantık kapılarını kullanan Cello'dur. Elektronik devre tasarım mantığının biyolojik devrelere uygulanmasının yeterli olamayacağını düşündüğümüzden, bu tezde prCell adını verdiğimiz biyolojik devrelerin tasarımı için yeni bir programlama dili geliştirdik. prCell'in ana özelliği, herhangi bir elektronik devre tasarımında biyolojik devreyi tasarlamak için kullanılan mantık geçitleri yerine, kullanıcının giriş-çıkış ve kontrol nesnelerini belirleyeceği dinamik yapılara dayanan bir mantığa sahip olmasıdır. Bu, çok büyük ve karmaşık bir biyolojik devre tasarlama esnekliği ve imkanı sağlar. İnternet üzerinden yayınlanacak olan prcell'de bir üyelik sistemi olacağından, kullanıcının bir veritabanı ve kullanıcının bir biyolojik devre tasarım ortamı olacaktır. En önemli hedefimiz sentetik biyoloji ile ilgilenen herkesin sinyal yolunu kolay ve etkili bir şekilde tasarlayıp canlı hücrede uygulayabilmesidir. Hücre içi olayları daha iyi modellemek için, prcell'in kullanıcı taleplerine göre geliştirileceğine ve canlıları yeniden programlamak için farklı bir bakış açısı sağlayacağına inanıyoruz. Anahtar kelimeler: Sentetik Biyoloji, Genetik Devreler, Hücre Programlama
Statistical methods and tools for quantitative mass spectrometry-based proteomics
Mass spectrometry, which stands out with its high sensitivity, application variety and fast results, is an indispensable tool for molecular and cellular biology to find atomic composition and sequencing over the relative masses of substances such as the clarification of the structures contained in organic and biochemical molecules, the determination of the molar masses of peptides, proteins and oligonucleotides.MSstats is an open source R package for statistical relative quantification of proteins and peptides in data dependent, targeted and data-independent proteomics. It enables the identification of different abundant proteins for MS experiments by chromatography-based quantification with complex designs, characterizing MS assays in terms of limit of blank and limit of detection (LOB/LOD), system suitability testing (SST) and quality control (QC). Users have to install R to use the MSstats package and know the R programming language or use MSstats-Shiny, which is offered online. However, many users do not know the R programming language. On the other hand, many users who know the R programming language, experience time loss during installation and coding. In addition, many users prefer to use their own computers considering the privacy of MS test results. In this thesis, a desktop application has been developed that can be used MSstats package without any programming knowledge, without installation. It can be used offline, saving user's time with ease of use, and providing privacy by analyzing on a closed local computer. This tool was developed by synchronous operation of R and Java programming languages. An interface called MSstats-J-Plugin has been prepared with Java and R programming languages.Thanks to this interface, the results are processed, made sense, visualized and finally presented to the user.Analysis of mass spectrometry data in MSstats-J-Plugin is carried out in four basic steps. These steps are "data upload", "data processing", "group comparison" and "design sample size".MSstats-J-Plugin developed as free and open-source, is designed to be accessible to everyone on the internet.
A computational approach for predicting host specificity of adenoviruses
Adenoviruses are large viruses that consist of a complex protein cover enclosing its DNA sequence and core proteins. They belong to a large family called Adenoviridae with over hundred known members that can infect extensive range of vertebrates. The cross-species transmissions of adenoviruses between human, non-human primate, bat, cat, dog, pig, sheep and goat species are known. Understanding adenovirus host specificity and prediction of cross-species virus transmissions are important in the management of infections caused by adenoviruses. Nearly fifty of known adenoviruses infect human and can cause, sometimes serious, gastrointestinal, respiratory, urinary, and corneal infections. For the infection to start, adenoviruses establish a complex set of interactions with host cells, typically initiated by a binding between a host cell receptor and adenovirus fiber protein. Hence, understanding protein-protein interactions (PPIs) between receptor and fiber proteins, is an essential step towards developing a model to predict host specificity of a given adenovirus. In this work, our goal was to create a pipeline to predict host specificity of an adenovirus. For this purpose, we, first, conducted a literature review and identified all human receptors that are involved in the uptake of adenoviruses. Next, we computed human receptor orthologs from other adenovirus host species. Then, we computed a library of fiber proteins from various adenoviruses infecting different host species. Finally, we developed a meta-classifier predicting host specificity of adenoviruses. The data objects used in the classifier were the predictions computed for host receptor-fiber protein interactions by the existing PPI prediction methods, namely VHPPI, HOPITOR and DeNovo. For these data objects, four different class labels were computed based on candidate host and adenovirus's known host sharing the same taxa at taxonomic levels species, genus, family, and order. Several ensemble machine learning approaches were tested in meta-classification. The stacking ensemble classifiers gave the best accuracies. For stacking, we tested J48 decision tree, naïve bayes, multilayer perceptron, random forest, and support vector machine algorithms. The performance of classifiers were close the each other with J48 decision tree being slightly better (True positive rate:71.5%, true negative rate 71.5% and 71.0%). In this work, we demonstrated the potential and accuracy of using an ensemble PPI prediction methods in prediction of host specificity of adenoviruses. The diversity of Adenoviridae fiber and known host receptor proteins posed a challenge in the development of the pipeline. Although our results are quite promising, they also suggest the need for more accurate cross-species PPI predictions. In the future, we anticipate our work can be leveraged and improved for early detection of host shifts or in assessment of host specificity of adenoviruses. Furthermore, our approach has the potential to be extended to other virus families beyond Adenoviriade and this is particularly significant, given the rise in viral disease epidemics involving host shifts in recent years.
Experimental design and parameters optimization of bioanode for tissue based biofuel cell
In this study, a plant tissue-based biofuel cell (BFC) system was developed using plant tissues. The system utilized glucose oxidase (GOx) enzyme as reaction catalysts. The tissue-based BFC was constructed by using a carbon felt electrode (CFE) in both bioanode and biocathode. The electrodes were modified with multi-walled carbon nanotube-gold-platinum (MWCNT- Au-Pt) hybrid nanomaterial. Nanohybrids immobilize GOx on the electrode surface. The electron transfer will occur through mediators which are in this study will be Phosphate buffer saline and benzoquinone (BQ). Response surface methodology (RSM) was utilized to analyze the correlation of the experimental parameters to the output and the results are represented using a 3D model and tables. The study was divided into two stages. The first stage was experimental optimization of the variables using the chronoamperometry electrochemical method. Then response surface methodoly and experimental design were used to computationally determine the optimized values. Design-Expert software was employed to create a set of experiments using a central composite design model (CCD). CCD employs full factorial parameters limits (+1,-1), star design limits (+α, -α), and center point repeat to accurately optimize the parameters. The results of RSM were modeled using two software, Design expert and Octave. Octave software is a free software similar to Mat lab. Then the optimized values were used to determine the best system design by comparing three systems, single chamber BFC, nafion proton exchange membrane double chamber BFC and salt bridge double chamber BFC.
Hepatosellüler karsinomda hedefe yönelik moleküler ajan tedavilerinin ağa dayalı keşfi
Hepatocellular carcinoma (HCC) is one of the most-deadly cancers and the most common type of primary liver cancer. Multikinase inhibitor Sorafenib is one of FDA approved targeted agents in HCC treatment. PI3K/AKT/mTOR pathway is altered in about 51% of HCC; hence, understanding how Sorafenib and PI3K/AKT/mTOR pathway inhibitors act at signaling level is crucial for targeted therapies and to reveal the off-target effects. In this work, we use gene expression profiles (GEPs) of HCC cells (Huh7 and Mahlavu) which were treated with seven different agents and their combination. Our aim is to reveal the important targets and modulators in agent treatments by inferring the dysregulation of Interactome. In other words, we search for the mechanism of action of the agents in a network context beyond the list of genes. For this purpose, we use the DeMAND (Detecting Mechanism of Action based on Network Dysregulation) algorithm developed by Califano Lab. DeMAND compares GEPs and assesses the change in the individual interactions from weighted interactome obtained from STRING database. As a result, we reconstructed 18 agent-specific networks from each GEPs. Each gene and interaction within these networks have a value signifies how strongly these genes are affected from the chemical network perturbation. Then, we found enriched pathways in each network. We initially compared the networks of single agents and their combination; i.e. PI3Ki-α, Sorafenib and their combined treatment. Then, we compared all networks simultaneously. The simultaneous comparison of the reconstructed networks at gene and pathway levels shows that several pathways and proteins are commonly affected across agent treatments (e.g., Wnt, HIF-1, Notch pathways and MCM proteins, mTOR). On the other hand, some pathways are only affected in a specific agent treatment (e.g., SNARE interactions).