69
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Discipline
Identifying functionally important missense mutations in cancer by dynamics-based analysis and predicting pathogenicity/disease category of missense mutations
Missense mutations have various effects on protein structures, also leading to distorted protein dynamics that plausibly affects the function. We hypothesized that missense mutations in cancer-related genes selectively target hinge-neighboring residues that orchestrate collective structural dynamics. To test our hypothesis, we selected 69 cancer-related genes from the Cancer Gene Census (CGC) database and their representative protein structures from the Protein Data Bank. We first identified the hinge residues in two global modes of motion by applying the Gaussian Network Model. We then showed that missense mutations are significantly enriched on hinge-neighboring residues in oncogenes and tumor suppressor genes. We observed that several oncogenes (e.g., MAP2K1, PTPN11, and KRAS) and tumor suppressor genes (e.g., EZH2, CDKN2C, and RHOA) strongly exhibit this phenomenon. Next, we developed a computational pipeline to detect significantly enriched three-dimensional (3D) clustering of missense mutations around hinge residues by using the the Cancer Genome Atlas (TCGA) dataset. The hinge residues were also detected by applying a Gaussian network model for the modes 1 to 5. By systematically analyzing the PanCancer compendium of somatic missense mutations in nearly 10,000 tumors from TCGA, we identified candidate genes and mutations in addition to well known ones. For instance, we found significantly enriched 3D clustering of missense mutations in known cancer genes including CDK4, CDKN2A, TCL1A, and MAPK1. Besides these known genes, we also identified significantly enriched 3D clustering of missense mutations around hinge residues in PLA2G4A, which may lead to excessive phosphorylation of the extracellular signal-regulated kinases. Our results show that the consideration of clustering around hinge residues can help us explain the functional role of the mutations in known cancer genes and identify candidate genes. Furthermore, we proposed new features, named hinge-based, for pathogenicity prediction for missense mutations and show that hinge-based features improve pathogenicity prediction. Pathogenicity prediction of human missense variants remains a challenging problem. Existing computational models are basically binary classifiers predicting whether given missense variants are deleterious or neutral. We demonstrated a multilabel classification method that predicts not only the pathogenicity but also the disease category type of given missense variants. Moreover, existing computational models are based on sequence-, structural-, or protein dynamics-based analysis. We also showed that network topological properties of proteins significantly improve determining the pathogenicity of missense variants. We trained and tested our model PathDis with 20,361 missense variants. Then, we benchmarked by the area under the ROC curve (AUROC) evaluation metric score with a well-established prediction model which uses the same dataset. We observed that our model PathDis improves AUROC by 3%. Then, we tested PathDis with a different dataset. Also benchmarking based on this different dataset against other well-established prediction models demonstrated that PathDis' AUROC score is approximately 3% higher than the second highest AUROC score. In addition to high pathogenicity prediction results, PathDis has approximately 79% accuracy for predicting the disease category types (i.e., No Disease, Cancer, and Non-Cancer). Besides introducing a prediction model, we also characterized the missense variants by our sequence-, structure-, dynamics-, and network-based features for the disease category types. We observed that sequence-based, network-based, and structure-/dynamics-based features characterize No Disease, Cancer, and Non-Cancer missense variants, respectively.
Tactile rendering of digital buttons and shapes on touchscreens using novel surface haptics technologies
Touchscreens are integrated in every aspect of our daily life as mobile phones, ATMs, tablets, vending machines, and car navigation systems. These screens provide an intuitive interface for touch interactions with no or limited haptic feedback. As a result, eye-free interaction is nearly impossible and users have to rely on visual and audio feedback to perform a task, which reduces the users' performance and experience. However, with the recent advances in surface haptics technologies, it is now becoming possible to generate complex tactile effects on touchscreens and enhance the user interactions by displaying more sophisticated haptic feedback. The overall goal of this thesis is to understand how to render realistic tactile buttons and shapes on touchscreens using novel surface haptics technologies. In this regard, the first part of the thesis focuses on creating vibrotactile feedback on a touchscreen that simulates the feeling of physical buttons using piezo actuators attached to the screen. For that purpose, we first recorded and analyzed the force, acceleration, and voltage data from twelve participants interacting with three different physical buttons: latch, toggle, and push buttons. Then, a button-specific vibrotactile stimulus was generated for each button based on the recorded data. Our results showed that participants were able to match the three digital buttons with their physical counterparts with a success rate of 83%. In addition, participants rated the degree of their subjective feelings using seven adjective pairs for all the physical and digital buttons investigated in this study. Our results showed that there exist at least three adjective pairs for which participants have rated two out of three digital buttons similar to their physical counterparts. In the second part, we investigated the recognition rate and time of five tactile shapes (i.e., triangle, square, pentagon, hexagon, and octagon) rendered by electrovibration on a touchscreen using three different methods and displayed in prototypical orientations and non-prototypical orientations (i.e., 15 degrees CW and CCW to the prototypical orientation). The results showed that the correct recognition rate of the shapes was higher when the haptically active area (area where electrovibration was on) was larger. However, as the number of edges increased, the recognition time increased and the recognition rate dropped significantly, arriving to a value slightly higher than the chance rate of 20% for non-prototypical octagon. Moreover, the recognition time for inside rendering condition was significantly shorter as compared to edge and outside rendering conditions and edge rendering condition led to the longest recognition time. Our analyses of exploration strategies revealed that participants first used global scanning to extract the coarse features of the displayed shapes, and then they applied local scanning to identify finer details, but needed another global scan for final confirmation in the case of non-prototypical shapes. We also observed that it was highly difficult to follow the edges of shapes and recognize shapes with more than five edges under electrovibration when a single finger was used for exploration. The results of these studies show that richer haptic stimuli is necessary for realistic rendering of digital buttons and shapes on touchscreens. For example, lack of kinesthetic feedback in rendering buttons and displaying haptic feedback in tangential direction only in rendering shapes adversely affected the perception of the participants in our study. These results are considered as a starting point for the development of tactile stimuli, haptically improved user interfaces, and touchscreen applications. These findings can also provide some guidance to haptic interface designers in developing techniques for efficient and effective interaction with graphical elements on touchscreens.
Extensive coarse-grained molecular dynamics simulations of soft matter: From RNA to hydrogels
Coarse-grained molecular dynamics (CG-MD) simulations are indispensable for investigating the dynamical response of macromolecular structures to external probes. The purpose of this thesis is to utilize CG-MD in order to determine (a) the response of a hydrogel to a time-varying external electric field, (b) the melting time of an RNA hairpin structure as a function of the temperature and the molecule size. Controlling the mechanical response of polyelectrolyte hydrogels to external electric fields is of great importance for hydrogel-based soft actuation systems. The first part of the thesis work involves the study of a semi-infinite polyelectrolyte hydrogel slab to a transient and spatially nonuniform half-sinusoidal electric field by means of both implicit and explicit solvent models. Results with an implicit solvent model demonstrated that an electric field confined to a small volumetric section of the hydrogel slab induces a reversible contraction of the entire slab in the direction perpendicular to the field. The hydrogel initially contracts by almost half of its field-free length and then retracts to its original size, with repeating contraction/retraction cycles exponentially decaying in magnitude akin to an underdamped oscillator. In contrast, almost no contraction occurs when the field is applied uniformly on the whole hydrogel slab. Analyses of contraction times and efficiencies for varying backbone charge fractions, dielectric constants, and salt concentrations confirm the robustness of the phenomenon. Further, by tuning the electric-field frequency and amplitude, both the contraction time and the efficiency can be controlled. Also, this phenomenon is demonstrated by using the explicit solvent model. Results showed that in realistic water concentrations (0.7 M), the contraction of the hydrogel decreases up to 35\%. Although this contraction is observed when an electric field is applied to the hydrogel slab partially like in the implicit solvent model, the field area increased to half of its initial size instead of 10\% for the implicit model. Concluded that our results of the mechanical behavior of the PE hydrogel slab are unconstrained by the solvent model, hydrogel's chemical properties, and electric field parameters. In the second part of the thesis, the melting behavior of an RNA hairpin, another macromolecular system distinguished by its palindromic sequence, investigated again by means of CG-MD simulations. Determination of the structure and dynamical behavior of nucleic acids as a function of temperature is a fundamental problem in polymer physics and is relevant for understanding the intracellular processes in thermophilic bacteria. The characteristic helical structure of these macromolecules is known to play an important role in their folding and melting dynamics. The main focus is on the melting dynamics upon an increase in the ambient temperature. Our results suggest the presence of two distinct dynamical regimes. The "critical regime" close to the equilibrium melting temperature displays an unexpected, non-monotonous dependence of the melting time on the temperature. The second, high-temperature regime is characterized by the entropic competition of the two denatured ends of the linear duplex structure. The scaling behavior of the melting time in both regimes investigated and identified the crossover boundary separating the two as a function of temperature and molecule size.
1000 genom projesinde yer alan CLOCK geni tek nükleotid polimorfizmlerinin in silico ve in vitro analizleri
Circadian rhythm is an internal process regulating ~24-h physiological and behavioral processes in organisms. In mammals, circadian rhythm is generated by transcription and translational feedback loop (TTFL) mechanism as a result of the interaction between core clock proteins. In TTFL, CLOCK and BMAL1 interact with each other and form a heterodimer to bind E-box sequences within the promoter region to initiate the transcription of the clock-controlled genes, including Period (Per) and Cryptochrome (Cry). Within the time, CRYs and PERs accumulate in the cytosol and then translocate into the nucleus with Casein Kinase Iε and repress BMAL1: CLOCK driven transcription. There are other auxiliary TTFLs exist that control circadian rhythm. Genetics and epidemiolocal studies suggest factors that disturb circadian rhythm result in susceptibility or may directly cause several diseases such as obesity, diabetes, cardiovascular diseases, aging, cancer, mood, and sleep disorders. Several single nucleotide polymorphisms (SNPs) for core clock genes have been identified and shown to be associated with different types of diseases. However, whether these SNPs contribute to the different type of the disease are ill-defined. One of the challenges in these approaches is that genome-wide association sequences (GWAS) studies have limitations in functional prediction. To address that, I developed using in vitro studies following the in-silico techniques to identify and characterize functional CLOCK SNPs from 1000 Genomes Ensemble to show the effect of a particular missense mutation on CLOCK protein on function. Such systematic approaches would allow us to discover SNPs with pathological effects and understand how these SNPs affect proteins' function. In this thesis, I performed a functional characterization of rare CLOCK missense variations (p.Phe104Cys, p.Leu118Arg, p.Asp119Val, p.Gly120Val, and p.Phe121Cys) identified from the Ensembl database. I initially analyzed these variations using computational tools. Results revealed that variants are located on the functionally important region of CLOCK. I used in vitro experimental approach and showed p.Leu118Arg, p.Asp119Val, and p.Phe121Cys CLOCK had reduced transactivation activity while p.Gly120Val CLOCK had increased transactivation along with BMAL1. However, p.Phe104Cys CLOCK had not been comparable transcriptional activity. To attrubitue these functional difference on the affinity between CLOCK SNPs and BMAL1, I performed co-immunoprecipitation between them Results indicated that p.Leu118Arg, p. Asp119Val, p.Gly120Val CLOCKs had reduced affinity to BMAL1and interestingly p.Phe121Cys CLOCK had increased the affinity to BMAL1. To gain more insight I further showed that the CRY1 had reduced repressor activity on p.Leu118Arg and p.Phe121Cys CLOCKs. Meanwhile, the estimation binding energy analysis of MD simulations supported the biochemical results, and binding energy analysis per residues was used to examine the mechanism of such SNPs effects. Collectively, I discovered that even single nucleotide changes in CLOCK directly affect the CLOCK functions. Hence, illumination of the effects of CLOCK SNPs would also help develop novel treatment strategies for diseases related to clock disruption for further studies and provide valuable information for the structure-function of CLOCK in circadian clock mechanism.
Investigating the role of nuclear receptor enhancers on gene transcription
When activated, the type-I steroid nuclear receptors (NR) translocate into the nucleus, altering the gene expression profile through non-coding cis-regulatory elements (CRE). Yet how they induce gene transcription is largely unknown. The androgen receptor (AR) and glucocorticoid receptor (GR) are well-studied examples of these nuclear receptors. Previously the AR enhancers were divided into three separate functional classes, and only a small subset of them was induced. We observed that those inducible AR enhancers are chromatin hubs and interact more with promoters of up-regulated genes upon hormone treatment. By integrating several independent publicly available functional genomic datasets, initially, we confirmed these classes for GR enhancers. We, then, investigated the intrinsic functional features of these enhancers, such as promoter activity. We discovered one particular enhancer class, which does not require NR occupancy for activity, had elevated promoter activity besides their enhancer activity. They were also accessible for binding various transcription factors (TF) in many tissues. Later, we described each factor's CRE interaction networks (cistrome) as graph diagrams and integrated the message passing algorithm to develop a scoring method for differential transcription factor occupancy on individual elements in these networks. As a proof of concept, this method effectively captured the AR or GR enrichment on the promoter of up-regulated genes; this couldn't be observed otherwise. We anticipate that our conceptual framework will improve our understanding of CRE function and gene expression regulation.
Assessing gas separation performances of COF membranes, COF/polymer MMMs, and dual filler-incorporated polymer membranes via high-throughput computational screening
Hundreds of covalent organic frameworks (COFs) have been synthesized and thousands of them have been computationally designed. However, it is impractical to experimentally test each material as membranes for gas separations. In this work, we focused on the membrane-based gas separation performances of experimentally synthesized COFs and hypothetical COFs (hypoCOFs). Gas permeabilities of COFs were computed by combining the results of grand canonical Monte Carlo (GCMC) and molecular dynamics (MD) simulations and many COFs were found to overcome the upper bound of polymeric membranes for He/H2, N2/CH4, H2/N2, He/CH4, H2/CH4, and He/N2 separations. We then examined the structure-permeability relations of the COF membranes that are above the upper bound for each of the six gas separations and based on these relations we proposed an efficient approach for the selection of best hypoCOFs from a very large database. Molecular simulations showed that 120 hypoCOFs identified to be promising. They exceed the upper bound for He/CH4, He/N2, H2/CH4, H2/N2 separations. Both COFs and hypoCOFs were then studied as fillers in 25 polymers leading to a total of 29,020 COF/polymer and hypoCOF/polymer mixed matrix membranes (MMMs), representing the largest number of COF-based MMMs investigated to date. Permeabilities and selectivities of COF/polymer MMMs were computed for six different gas separations and results revealed that 18 of the 25 polymers can be carried above the upper bound when COFs were used as fillers. In the second part of thesis, we developed computational methods to identify the best-performing MOF-COF pairs to be used as dual fillers in polymer membranes for target gas separations. We focused on COF/polymer MMMs located below the upper bound due to their low gas selectivity for five industrially important gas separations, CO2/N2, CO2/CH4, H2/N2, H2/CH4, H2/CO2. Results showed that for polymers having a relatively high gas permeability (≥104 Barrer) but low selectivity (≤2.5) such as PTMSP, addition of the MOF as the second filler can have a dramatic effect on the final gas permeability and selectivity of the MMM. Property-performance relations of MOF and COFs were analyzed to understand effects of structural and chemical properties of the fillers on the permeability of the resulting MMMs and MOFs having Zn, Cu, and Cd metals were found to lead to the highest increase in gas permeability of MMMs. The methods in this thesis will provide insight into COFs and hypoCOFs in future computational studies and to find out promising MOF-COF pairs to be used in dual filler-incorporated polymer membranes.
Context-specific signaling pathway construction in cancer through network motif search
Abnormal alterations in intracellular signaling networks are common features in many cancer types. Yet, accurately representing signaling networks by identifying context-specific interactions within complex interactomes remains a significant challenge. In this thesis, we focused on searching significant network motifs - repeating patterns in complex networks - to reconstruct context-specific networks and identify clinically important target proteins and pathways. In our novel method, we employed a motif analysis by screening a series of three-node kinase-mediated subnetworks including feed-forward and feed-back loops in a reference directed interactome and tested the significance of their presence compared to random interactomes. As a result, we found five prominent motifs: positive cascade, positive feedback loops, and coherent type-1, coherent type-2, and incoherent type-1 feed-forward loops (FFLs). Using this method, we generated tumor-specific networks for 69 ovarian cancer patients by combining their phosphoproteomic profiles in CPTAC with three-node motifs. Merging significant motifs having at least two differentially expressed phosphoproteins in the corresponding tumor followed by filtering out non-specific edges gives the final patient-specific network. On one side patient-specific networks contains intermediate nodes; on the other side, they represent causal relations between altered proteins and eventually pathways. All pair comparison of similarities between patient-specific networks resulted in 30% node and 5% network motif overlap on average. We utilized an unsupervised method to align patient groups with ovarian cancer cell lines and performed clustering based on mRNA expression profiles of four drug targets showing significant effects on patient survival. Consequently, we identified enriched targets and pathways within the clusters, and linked them to potential drugs, providing personalized therapeutic options for each patient cluster. This approach facilitated the assessment of kinase inhibitors' efficacy in a clinically relevant context and yielded valuable insights for developing personalized treatment strategies using personalized network motif profiles.
Leveraging the molecular signatures of cancer for dynamic network
Drug resistance poses a significant challenge to the effectiveness of therapies, driven by accumulation of molecular alterations within dynamic cellular networks. In this thesis, we used a discrete dynamic model, Graph-based Cellular Automata (GCA), to reveal the network-based history of tumor progression and causal association between network modules and drug resistance by data integration. GCA can simulate dynamic systems using initial static information, set of states and simple transition rules. The reference graph is a tissue-specific interactome that consists of both protein-protein interactions and the regulatory network of the transcription factor to gene interactions composed of 8,228 nodes and 63,574 edges. By incorporating known biology and statistical rules of molecular alterations, including stimulations, repressions, and (non)-linear pairwise molecular correlations, GCA simulates molecular signalling and propagates mutation effects downstream of signalling pathways and complexes. Eventually, GCA gives a trajectory of subnetwork models for each context. In comparisons of simulations with and without mutations at the node level, we detected functional subnetworks within the dynamic network structure. We used publicly available omics data from a well-established cancer cell line repository to optimize the GCA model and construct dynamic networks for each cell-line-drug pair for interpreting drug resistance mechanisms at the pathway level. The accuracy of these drug representative networks were evaluated by cross-validation and on an independent test from Patient-Derived Xenografts (PDX). Notably, we found context-specific pathways (e.g. MAPK signalling) involving proteins from drug-resistant cell lines and PDX samples, thereby linking them to investigated drug resistance mechanisms. Overall, this approach, from molecular alterations to dynamic networks, transforms already available large datasets to gain new clinically relevant insights about drug resistance, offering potential implications for cancer therapy.
Predicting drug response through learning from network-based integration of multi-omics data in cancer cell lines
Predicting the treatment response continues to be a significant hurdle due to the inherent heterogeneity among tumours, leading to divergent responses to identical therapeutic approaches. This thesis introduces a strategy aiming to overcome this challenge by integrating a network-based hybrid machine learning model to detect drug response patterns, potentially elucidating the intricate biological mechanisms at play. Toward this goal, our approach deploys baseline omics data from more than 900 cancer cell lines. We assemble multi-omics data, including mutation, proteomic, and transcriptomic profiles from a broad spectrum of cell lines, along with their drug sensitivity data and protein targets of the drugs. The collected information is integrated onto a human protein-protein interactome, and using a personalized PageRank network propagation technique, we generate around 28,000 unique context-specific subnetworks, each representing a distinct drug-cell line pair. From this pool, we filtered out the grey zone contexts and focused on 1,194 pairs involving 426 cell lines and 45 drugs, labelled as either sensitive or resistant, facilitating more accurate predictions. Subsequently, these complex networks are converted into a format compatible with machine learning models via Graph2Vec, an unsupervised graph embedding algorithm, resulting in vectorized network representations encapsulating their topological and attribute-related features into a vector. These transformed networks serve as the input for a machine learning classifier, enabling it to predict drug responses based on the distinct properties inherent to each drug-cell line combination. Validity of this prediction model is thoroughly examined through a stringent testing, including a 10-fold cross-validation, resulting in an accuracy of 0.83 and an Area Under the Precision-Recall Curve (AUC) of 0.90. Moreover, additional 10-fold blind testing on independent cell line, drug, and tissue sets yielded an AUC of 0.72, 0.89, and 0.81, respectively. These findings underscore the model's potential in enhancing the precision of drug response predictions and illuminating the causal network modules underpinning drug resistance.
Electro-mechanical contact interactions between human finger and touchscreen under electroadhesion
Electroadhesion is a promising technology with potential applications in robotics, automation, space missions, textiles, tactile displays, and some other fields where efficient and versatile adhesion is required. However, a comprehensive understanding of the physics behind it is lacking due to the limited development of theoretical models and insufficient experimental data to validate them. In this thesis, we have developed an electro-mechanical model to estimate the magnitude of electrostatic forces between human finger and touchscreen under electroadhesion. We also measured the friction forces between the finger and touchscreen to infer the magnitude of electrostatic forces experimentally. The model is in good agreement with the experimental data and showed that the change in magnitude of the electrostatic force is mainly due to the leakage of charge from the Stratum Corneum layer of the skin to the touchscreen at frequencies lower than 250 Hz and electrical properties of the Stratum Corneum at frequencies higher than 250 Hz. In addition, we proposed a new and systematic approach based on electrical impedance measurements, where skin and touchscreen impedances are measured and subtracted from the total impedance to obtain the remaining impedance in order to estimate the electrostatic forces between the finger and the touchscreen. This approach also marks the first instance of experimental estimation of the average air gap thickness between human finger and voltage-induced capacitive touchscreen. Moreover, the effect of electrode polarization impedance on electroadhesion was investigated. Precise measurements of electrical impedances confirmed that electrode polarization impedance exists in parallel with the impedance of the air gap, particularly at low frequencies, giving rise to the commonly observed charge leakage phenomenon in electroadhesion. We also investigated tactile perception by electroadhesion for DC and AC voltage signals applied to the touchscreen using ten participants with varying finger moisture levels. Our study showed that the voltage detection threshold for an AC signal was significantly lower than that of the corresponding DC signal and we explained this discrepancy by charge leakage at lower frequencies again. We have also observed that the participants with a moist finger had significantly higher threshold levels than the rest of the participants, which is supported by our electrical impedance measurements. Finally, we aimed to investigate the effect of touchscreen's top coating layer on our tactile sensing with and without electroadhesion, and within the time frame of this thesis, we have focused on the latter only. Hence, we first performed psychophysical experiments to quantify human tactile discrimination ability of touchscreen surfaces coated with different materials, followed by multiple physical measurements. The results showed that coating material has a strong influence on our tactile perception and human finger is capable of detecting differences in surface chemistry due to, possibly, molecular interactions. In conclusion, the findings of this thesis provide new insights into the physics of finger-touchscreen interactions under electroadhesion and have implications for the design of robotic systems and haptic interfaces utilizing this technology.
Advantage actor-critic deep reinforcement learning approach for paint shop planning and scheduling
Paint shops usually act as bottlenecks in production facilities requiring a painting procedure. To enhance efficiency and optimize the process by minimizing color batch changes that can decrease productivity, it is essential to develop optimization algorithms. Traditionally, these problems have been addressed using a mixed-integer linear programming (MILP) approach. However, mathematical optimization methods face challenges in adapting to dynamic production planning environments and the real-time nature of a production facility. This is due to its memoryless structure and search for exact and optimal solutions by solving the entire every time a schedule is required. To overcome the issues, this study proposed a deep reinforcement learning algorithm to solve and optimize a paint shop scheduling and planning problem that can adapt to dynamic environments. The actor-critic approach was the best method amongst the other policy-based state-of-the-art deep reinforcement learning algorithms. To train a DRL agent, a real-life simulation model of a paint shop in a household appliance factory was built to act as an environment. After the training and inference processes, the outcome was a paint shop production plan that minimizes the inventory cost and bottlenecks while maximizing the productivity of washing machine production and achieving energy efficiency through planned production stops. Besides some advantages of linear programming methods, DRL models performed well on the selected application of paint shop scheduling and planning problems. It is seen that DRL methods are superior in terms of efficiency and computational performance on inference step while obtaining at least sub optimal solution.
Robotic learning of haptic skills from expert demonstration for contact-rich manufacturing tasks
In this study, we propose a learning from demonstration (LfD) approach that utilizes an interaction (admittance) controller and two force sensors for the robot to learn the force applied by an expert from demonstrations in contact-rich tasks such as robotic polishing. Our goal is to equip the robot with the haptic expertise of an expert by using a machine learning (ML) approach while providing the flexibility for the user to intervene in the task at any point when necessary by using an interaction controller. The utilization of two force sensors, a pivotal concept in this study, allows us to gather environmental data crucial for effectively training our system to accommodate workpieces with diverse material and surface properties and maintain the contact of polisher with their surfaces. In the demonstration phase of our approach where an expert guiding the robot to perform a polishing task, we record the force applied by the human and the interaction force via two separate force sensors for the polishing trajectory followed by the expert to extract information about the environment. An admittance controller, which takes the interaction force as the input is used to output a reference velocity to be tracked by the internal motion controller (PID) of the robot to regulate the interactions between the polisher and the surface of a workpiece. A multilayer perceptron (MLP) model was trained to learn the human force profile based on the inputs of Cartesian position and velocity of the polisher, environmental force, and friction coefficient between the polisher and the surface to the model. During the deployment phase, in which the robot executes the task autonomously, the human force estimated by our system is utilized to balance the reaction forces coming from the environment and calculate the force needs to be inputted to the admittance controller to generate a reference velocity trajectory for the robot to follow. We designed three use-case scenarios to demonstrate the benefits of the proposed system: we first compare the performance of an expert polisher with a naive user to show the importance of haptic skills in polishing. In the second use-case, we show that the proposed system can successfully learn the intrinsic changes in human force profile from the expert user for autonomous robotic polishing of workpieces of different material and surface properties. The last use-case scenario involves human intervention during the robotic polishing for the regions on the workpiece requiring more polishing. The presented use-cases highlight the capability of the proposed pHRI (physical Human Robots Interaction) system to learn from human expertise and adjust its force based on material and surface variations during automated operations, while still accommodating manual interventions as needed.
Multi-scale network inference framework using single cell transcriptomic data
Cancer, characterized by uncontrolled cell proliferation, tumor invasion and aberrant signaling, involves complex intercellular and intracellular interactions within tumors. To address this complexity, we developed a framework to integrate single cell transcriptomic data with interactomes and model tumors as multi-scale networks. Our approach integrates reconstructed gene regulatory networks (GRN), signaling networks and receptor-ligand interactions to model intracellular and intercellular communication. We applied this framework to a publicly available single-cell transcriptomic data from LUAD patient tumors and constructed a multi-scale network containing GRNs and signaling networks specific to the cell subtypes and cell-cell interactions across subtypes within the primary tumor. Our analysis identified differentially expressed genes between normal and cancer cells, and between primary tumors and brain metastases, highlighting the critical role of receptor-ligand crosstalk in oncogenic signaling. Our approach revealed the role of several receptor-ligand crosstalk between different cell subtypes that can mediate oncogenic signaling. Additionally, our results suggest that the abnormal expression of transcription factors in cancer cells crucially influences oncogenesis and tumor suppression. Our framework is easily adaptable to various single cell transcriptomic data as well as other omic data types. Overall, our framework uses multi-scale network inference from single-cell transcriptomics to enhance our understanding of intracellular pathways and cell-type interactions, integrating multiple biological network layers.
Crosstalk between cardiovascular and cognitive diseases: deciphering molecular mechanisms of vascular cognitive impairment
Vascular cognitive impairment (VCI) is a growing public health concern with significant implications for human health. VCI is an understudied complex disease; therefore, this work aims to understand this disease by studying complex molecular interactions between cardiovascular (CVD) and cognitive diseases (CD). This thesis analyzes this crosstalk by building and examining protein-protein interaction (PPI) networks related to CVD and CD. By analyzing alternative protein conformations and mutations in interfaces of interactions, this thesis suggested three mutations in the kinase DYRK1A (V165I, S337P, and D401G) may be essential for VCI via its interaction with APP. Our results indicated that chemokine-related, and stress response-related pathways are likely related to Blood-Brain-Barrier dysregulation in VCI. We found mutant and wild-type VCP, XRCC4, and LIG4 conformations interacting with BRCA1. The analysis of the effect of transcription factors on CVD-CD crosstalk showed that JUN, CREB1, NFKB1, ESR1, and NR3C1 are crucial for VCI regulation, particularly the interaction between the mutant conformation of NFKB1 (structure: 2O61 chain B) and wild-type conformation of NR3C1 structure (3H52 chain B). Lastly, we clustered and predicted disease labels with machine-learning models. We found that GUILD scores and ESMs as features could be crucial when training a model to predict disease labels. Clustering resulted in four clusters enriched in pathways that support our previous results and suggest that RhoGTPase Signaling could play an essential role in the CVD-CD crosstalk and VCI progression. This thesis uncovered biomarkers, protein-protein interactions (protein structures/conformations that interact), and potential therapeutic targets to benefit the scientific and medical communities. This work also proposed pathways linking CVD and CD to suggest new approaches for VCI intervention.
Shape and geometry preserving loss functions for computed tomography image segmentation
Segmentation networks are not explicitly imposed to learn global invariants of an image, such as the shape of an object and the geometry between multiple objects, when they are trained with a standard loss function. On the other hand, especially when there exists a limited amount of training data, incorporating such invariants into network training may help regularize the training, provided that these invariants are the intrinsic characteristics of the objects to be segmented. This thesis addresses this issue by introducing the topology-aware loss function, with alternative formulations, that penalizes shape and geometry dissimilarities between the ground truth and prediction through persistent homology. We use three different topological filtration functions, leading to alternative formulations of the topology-aware loss function. After obtaining the persistence diagrams of both the ground truth and prediction maps by a topological filtration function, the topological dissimilarity is calculated by the use of the Wasserstein distance between the corresponding persistence diagrams. Our experiments on two different datasets of CT images reveal that the increase in the network's performance is significant.
Classification of surgical and recovery durations in healthcare settings utilizing machine learning models: A case study at Koç University Hospital
Efficient operating room (OR) scheduling is essential for minimizing patient wait times, optimizing resource use, and reducing operational costs. However, accurately predicting surgery and postoperative recovery durations is challenging due to patient variability, procedural complexity, and limited data availability. This study addresses these challenges by converting continuous duration data into categorical intervals, enabling classification-based machine learning methods suitable for small datasets. Using real-world data from Koç University Hospital, several algorithms—including Logistic Regression, Random Forest, Gradient Boosting, Support Vector Machines, K-Nearest Neighbors, Decision Trees, and ensemble combinations—were evaluated under Random Oversampling and Synthetic Minority Oversampling Technique strategies. Results showed that ensemble models consistently outperformed individual classifiers. For surgery time prediction, the best ensemble achieved 0.71 accuracy and 0.698 F1 score, while Random Forest reached 0.79 accuracy and 0.768 F1 score for recovery time. SMOTE proved effective in mitigating class imbalance, improving recall and F1 scores across models. Targeted feature grouping further enhanced interpretability and predictive reliability. This study provides a practical, data-driven framework for hospitals with limited records to improve OR scheduling. Reliable categorical predictions of surgery and recovery durations can enhance resource allocation, reduce scheduling conflicts, and improve patient outcomes. Keywords: Surgical duration prediction, recovery time prediction, machine learning, classification algorithms, SMOTE, ensemble models, healthcare resource optimization
Probabilistic three-dimensional fracture mechanics and its applications using FCPAS
Cracks can be seen in many engineering structures. It is important to determine the mechanical strength and life of the cracked structure or design a mechanical part with a damage tolerance approach. It is hard to determine the remaining life of machine parts exactly, since they include some uncertainties and variations in governing parameters of the problem, such as geometric dimensions and the variability of material properties and loading conditions. Therefore, for such problems, crack growth lives must be estimated by means of probabilistic approaches considering the variables that affect lives. In this study, for three-dimensional fatigue crack growth problems, a probabilistic crack growth life estimation procedure, which also involves Monte Carlo Simulations, was developed and validated by controlled laboratory experiments. The uncertainty in material properties affecting fatigue crack propagation life was determined using standard Compact Tension (CT) specimens machined from 7075-T6 aluminium alloy. Fatigue crack growth models for constant or variable amplitude loading in the literature were investigated and an improved model has been proposed. The data obtained from two-dimensional crack propagation tests were used in three-dimensional crack propagation simulations. A non-standard specimen made from Aluminium 7075-T6 has been designed for three-dimensional mode-I fatigue surface crack growth tests. Surface crack growth experiments under constant and variable amplitude loads were conducted using this specimen. Fatigue crack growth simulations were also carried out by considering the geometric tolerances of the specimen, the scatter of the fatigue crack growth-related material properties and the variability in loading. Experimental results were compared with simulations for different crack growth models, allowing validation of the proposed probabilistic fatigue crack growth methodology.
DMA verilerinin viskoelastik ana eğriye dönüştürülmesi ve Prony serisi bulunması
This thesis details the process of converting Dynamic Mechanical Analysis (DMA) data of a viscoelastic material at different temperatures into a master curve. The methodology involves shifting DMA data based on the Williams-Landel-Ferry (WLF) universal constants and iteratively adjusting the shift to construct a continuous master curve. A Prony series is then fitted to the master curve, providing insights into the relaxation modulus at various temperatures. For the iteration a numerical method has developed using MATLAB. This process involves a comparison between the numerical iteration and the manual shifting using excel which is a criterion for the numerical iteration. Fitting Prony series is executed using MATLAB algorithms, for this instance Least Squares Algorithm is used. The developed numerical method is verified by comparing the results to those predicted from a Finite Element Analysis of a brain tissue subjected to compression loading.
Farklı fiber konfigürasyonları için biyouyumlu mikroakışkan yapay akciğerlerin hesaplamalı pulsatil akış ve verimlilik analizi
Average-sized microfluidic artificial lungs consisting of rows and columns of fiber bundles with the different column to row aspect ratios (AR) are numerically analyzed for flow characteristics, maximum gas transfer performance, minimum pressure drop, and proper wall shear stress (WSS) values in terms of biocompatibility. The flow is fully laminar and assumed to be incompressible. The problem is solved with both Newtonian and Non-Newtonian Carreau models. The transport analysis is performed using a combined convection-diffusion model, and the numerical simulations are carried out with the finite element method. The inlet volumetric flow is modeled as a sinusoidal wave function to simulate the cardiac cycle and its effect on the device performance. The model is first validated with experimental studies in steady-state condition and compared with existing correlations for transient conditions. Then, the validated model is used for a parametric study in both steady and pulsatile flow conditions. The results show that increasing the aspect ratio in fiber configuration leads to converging gas transfer rate, higher pressure drop, and higher WSS. While determining the optimum configuration, the acceptable shear stress levels play a decisive role to ensure biocompatibility. Also, it is observed that the steady analysis underestimates the gas transfer for higher aspect ratios. The Newtonian model finds the pressure drop and shear stress values less than the Carreau model. In contrast, the oxygen transfer performance observed in the Newtonian model is overestimated approximately by 5\% compared to the Carreau model predictions.
Elektrik çift tabaka kapasitörlerinin kinetik ve denge özelliklerinin incelenmesi
The growing demand for efficient and scalable energy storage solutions has driven significant advancements in supercapacitor technology. This study investigates the kinetic and structural equilibrium characteristics of Electrical Double-Layer Capacitors (EDLCs) using molecular dynamics (MD) simulations to examine the effects of pore size, solvent dipole moment, and applied voltage on their performance. Twenty distinct systems were designed by combining two pore sizes (7.78 Å and 14.75 Å), five solvent dipole moments (ranging from 0.91 D to 6.18 D), and two applied voltages (1 V and 2 V). The objective is to elucidate how these parameters interact to optimize the energy and power densities of EDLCs. The results indicate a complex and sensitive relationship between pore size, electrode voltage, and capacitance. For systems with a 14.75 Å pore size, a positive correlation was observed between solvent dipole moment and capacitance across all dipole moments. However, in narrower pores (7.78 Å), this correlation reversed beyond a dipole moment of 2.88 D, revealing a nuanced dependence of capacitance on solvent polarity. Additionally, electrolyte behavior within the pores exhibited distinct spatial distributions: narrower pores led to denser ion layering, whereas wider pores facilitated more uniform ion dispersal. These findings provide deeper insights into EDLC dynamics, offering valuable guidelines for optimizing the design of next-generation supercapacitors.
Zaman serisi analizini etkili motif keşfi ile geliştirme ve motif keşfinin tahmin uygulamalarına entegre edilmesi
Time series motif discovery is a powerful technique for identifying recurring patterns in sequential data, offering valuable insights into diverse applications such as finance, healthcare, and climate science. This thesis presents an innovative framework for motif discovery in time series data, addressing key challenges in identifying recurring patterns of varying lengths and instances. The proposed methodology improves sensitivity and accuracy of motif detection by using Binary Integer Programming (BIP) for optimal motif selection and employing a heuristic algorithm for efficient pattern discovery, while integrating a dynamic feedback mechanism to refine parameters. Additionally, the study demonstrates how the discovered motifs can be applied to forecasting tasks by integrating them into a Transformer model pipeline augmented with motif embeddings, assessing whether motif-augmented predictions outperform simpler baselines. Extensive experiments on real-world financial datasets reveal that while the identified motifs can effectively capture recurring structures, incorporating them into advanced forecasting models does not necessarily lead to improved predictive performance. In particular, Transformers with or without motif embeddings underperform relative to basic econometric techniques. These findings underscore the complexity of leveraging motif information within deep learning frameworks for time series forecasting, suggesting that although motif discovery is valuable for understanding temporal patterns, more targeted modeling strategies or alternative machine learning approaches may be necessary to realize its full potential in predictive tasks.
Yüksek uzamsal-zamansal çözünürlüklü i̇kli̇m veri̇si̇ i̇le hi̇droloji̇k modelleme : Bozkurt (Kastamonu) örneği̇
This thesis integrates convection-permitting climate modeling with catchment-scale hydrological simulation to analyze the 2021 flash flood in Bozkurt, Kastamonu, and assess future flood risk under climate change. High-resolution precipitation and temperature data were produced using RegCM5 with a non-hydrostatic core, and used to drive the SWAT model in the Ezine Stream basin. Flash floods, triggered by short-duration intense rainfall and steep terrain, are among the most destructive hydrometeorological hazards. Bozkurt was selected due to its topography and vulnerability, with the August 2021 flood used as the reference event. The study aims to (1) simulate the 2021 flood, (2) evaluate runoff and streamflow under current and future conditions, and (3) assess flash flood potential. Results show that RegCM5–MOLOCH can capture convective rainfall patterns when configured with high resolution and proper domain setup, though some timing and spatial shifts occurred. SWAT was useful for general runoff trends but limited in flash flood timing. Future projections indicate an increase in short, intense rainfall events. The integrated modeling approach contributes to understanding local flood risks and supports future preparedness.
Karadeniz kıyılarında gözlenen yağış ekstremleri ve öngörülen etkileri
Climate change has increased extreme precipitation events worldwide, leading to significant economic losses in Black Sea coastal countries (Bulgaria, Georgia, Romania, Russia, Türkiye, and Ukraine). This study investigates the relationship between the return periods of such events (2000–2024) and the resulting economic losses, while also projecting future impacts. Strong and significant relationships were found for Türkiye, Bulgaria, and Romania, while analyses for Russia and Ukraine were limited due to data gaps. For future projections, extreme precipitation trends for 2030–2060 were evaluated using the MPI-ESM1-2-LR climate model under the SSP5-8.5 scenario (CMIP6). Even small increases in rainfall are shown to cause significant economic consequences. The findings provide a foundation for disaster management, infrastructure planning, and regional climate adaptation policies
Comparison of the national-contributed emissions with the climate trace's global emissions inventory
Accurate and transparent greenhouse gas (GHG) inventories are vital for tracking global climate commitments. With advances in satellite technologies and AI-integrated platforms, alternative inventories like Climate TRACE have emerged as complements or replacements for traditional systems such as the UNFCCC. This study compares GHG emissions reported by Climate TRACE and UNFCCC country submissions for 2015–2021. Eight countries—Australia, Belgium, Brazil, Germany, Japan, Poland, Turkey, and the United States—were selected for their data availability and regional diversity. Percentage differences between the two datasets were calculated for 13 sectors, and methodological consistencies were assessed using Principal Component Analysis (PCA) and K-means clustering. A heatmap visualization illustrated percentage differences across sub-sectors, highlighting areas of agreement and divergence. Results show substantial differences across most countries and sectors, with Climate TRACE often reporting higher emissions. The Waste sector and certain LULUCF sub-sectors had the largest discrepancies, largely due to methodological, scope, and data granularity differences. Climate TRACE's satellite- and AI-based methods—such as flaring inefficiency models, wastewater plant detection, and forest degradation monitoring—contributed to these gaps. Conversely, sectors like Enteric Fermentation in some countries (e.g., Belgium) showed higher alignment where Climate TRACE applied standard IPCC methodologies. The findings underscore how methodological heterogeneity—between and within inventories—affects comparability, highlighting the need for transparent methods and consistent sectoral definitions to improve global climate accountability.
Büyük depremleri takip eden artçı şok örüntülerinin derin öğrenme ile modellenmesi
This thesis presents data-driven modeling strategies for predicting the spatial distribution, number, and magnitude of aftershocks following major earthquakes. To overcome the limitations of classical physics-based approaches, alternative model structures were developed for both classification and regression tasks using machine learning methods such as artificial neural networks, the XGBoost algorithm, and kernel-based probabilistic modeling. The study systematically evaluated various input combinations—including static components derived from stress tensors, neighborhood information, gradient-based derivative quantities, spatial coordinates, and distance to the epicenter—and additionally applied oversampling and undersampling strategies to mitigate class imbalance, while employing kernel-based continuous labeling to model the spatial continuity of aftershock probabilities. Through this comprehensive experimental analysis, the prediction performances of aftershock occurrence, count, and magnitude were compared across different data structures and modeling strategies, demonstrating the potential of stress-based machine learning models in aftershock forecasting. The findings indicate that kernel-based probabilistic models provide a more realistic and spatially coherent representation of aftershock regions, XGBoost enhances discrimination capability in classification tasks, and multilayer neural networks yield more balanced performance in numerical regression tasks. Overall, the results show that static features derived from stress tensors can successfully capture large-scale spatial patterns of aftershock behavior, yet modeling fine-scale variability will require richer temporal and geological datasets in future research.
Melanopsin ve cryptochrome proteinleri arasındaki etkileşimin hesaplamalı araştırması
Circadian rhythms are oscillations in the biochemical, physiological, and behavioralfunctions of organisms that occur with a periodicity of approximately 24 hours. Inmammals, circadian rhythm is generated by a molecular clock. The molecular clock,which is located at suprachiasmatic nuclei (SCN) part of brain, is synchronized byenvironmental light-dark cycle.The cryptochromes are the mammalian circadian photoreceptors; they absorb light andtransmit the signal to the molecular clock. The cryptochromes and melanopsin (andpossibly other opsin family pigments) have been proposed as circadian photoreceptorpigments that exist in the inner retina. Experimental studies imply that there is mostprobably an interaction between melanopsin and cryptochromes for molecular clock tofunction normally. In order to uncover this interaction; the tertiary structures ofMelanopsin and Cryptochrome; the possible interaction between those two proteins havebeen predicted by usage of different computational means.The results of this study imply that mammalian Melanopsin and Cryptochrome proteinsinteract. The N-termini of Cryptchrome protein interacts with C-termini and cytoplasmictails of Melanopsin protein. The in vivo interaction is supported by preliminaryflourescent microscopy technique.
İki boyutlu kümelemeye dayalı analizle göğüs kanseri ifade davranışındaki işlevsel ve zamansal ilişkilerin açığa çıkarılması
The emerging microarray technology has useful applications in cancer research.Reports detailing the expression profiles of various types of cancers have pointed to theutility of this approach in defining cancer classes and subclasses with distinct molecularconfigurations and clinical behaviour. Another crucial application of microarrays tocancer research is the detection and selection of diagnostic marker genes. Thesemolecular markers give valuable additional information about tumour diagnosis,prognosis and therapy development . In the case of breast cancer, estrogen receptor (ER)and progesterone receptor (PR) have been used as prognostic markers in clinicalmanagement of breast cancer patients. Patients with ER negative breast tumors have apoorer prognosis than patients with ER positive tumors. Expression profiling of themammary gland to identify tumor associated genes differentially expressed in breasttumors regarding the presence or absence of ER and PR lead to the identification ofgenes, the function of some of which is unknown. However most of these previousstudies uses clustering methods which focus on the global pattern seen in the expressionprofiles. On the other hand, in this study, we used biclustering, which captures the localexpression changes better to improve the understanding on the molecular mechanismswhich underlie the response to estrogen in breast cancer cells. In addition, functionalannotation enrichment, pathway information, transcriptional factor binding site analysiswere also utilized. The results show that estrogen responsive genes act in groups of genesor the so-called `modules? which follow an order in time. Furthermore the temporal orderis associated with a biologically very meaningful functional order, which is supported bythe transcription factor analysis. The pathway analysis indicates that estrogen isresponsible for eliciting mechanisms required for tumorigenesis. All these results showthat biclustering if very useful in the analysis of time series data and it is capable ofelucidating the underlying functional and temporal relationship between genes bettersince it catches local expression behaviours better than the global ones.Advisor: Yar. Doç. Dr. Özlem Keskin Date: 16/11/2006Co-advisor: Doç. Dr. Attila Gürsoy Date: 16/11/2006Director: Prof. Dr. Süleyman Özekici Date: 16/11/2006
Eşler arası epidemik içerik dağıtım protokol dizaynı ve performans değerlendirmesi
Peer-to-peer (P2P) cooperative systems are becoming extremely popular as they finddiverse applications. One major application area is the content distribution over large-scale networks. As the usage of the Internet grows up, the number of large contents suchas software packages and popular movie files, and also the user population retrievingthese contents increase exponentially.In this thesis, we propose and design a peer-to-peer system, SeCond, addressing thedistribution of large sized content to a large number of end systems in an efficient andeffective manner. In contrast to prior work, it employs an epidemic dissemination schemefor state propagation of available blocks and initiation of block transmissions. In order tosupport heterogeneity of peers, ease of deployment, scalability, and adaptivity to dynamicpeer arrivals/departures, and also to increase the utilization of the system resources, wepropose mechanisms such as adjusting protocol parameters according to the bandwidthusages dynamically. We describe our protocol SeCond and its discrete event simulationmodel. A comprehensive performance evaluation has been accomplished for a wide rangeof scenarios. A well known and widely used P2P content distribution system isBitTorrent which we also model and compare as a benchmark. Performance resultsinclude scalability analysis for different arrival/departure patterns, flash-crowd scenario,overhead analysis, and fairness ratio. The major metrics we study include the average filedownload time, load on the primary seed, uplink/downlink utilization, communicationoverhead, and the fairness ratio. SeCond peers download the file faster compared toBitTorrent peers for most of the scenarios and the protocol is as fair as BitTorrentalthough it has no explicit strategy addressing free-riding. We show that SeCond is ascalable and adaptive protocol which takes the heterogeneity of the peers into account.We also illustrate the applicability of an analytical fluid model to the behavior ofSeCond.
Gen düzenlemesi ağlarının yapısı ve dinamiği
The structure and dynamics of a typical biological system are complex dueto strong and inhomogeneous interactions between its constituents. The investigation of such systems with classical mathematical tools, such as differential equations for their dynamics, is not always suitable. The graph theoretical models may serve as a rough but powerful tool in such cases.In this thesis, I first consider the network modeling for the representation of the biological systems. Both the topological and dynamical investigation tools are developed and applied to the various model networks. In particular, the attractor features' scaling with system size and distributions are explored for model networks. Moreover, the theoretical robustness expressions are discussed and computational studies are done for confirmation.The main biological research in this thesis is to investigate the transcriptional regulation of gene expression with synchronously and deterministically updated Boolean network models. I explore the attractor structure and the robustness of the known interaction network of the yeast, Saccharomyces Cerevisiae and compare with the model networks. Furthermore, I discuss a recent model claiming a possible root to the topology of the yeast's gene regulation network and investigate this model dynamically.The thesis also included another study which investigates a relation between folding kinetics with a new network representation, namely, the incompatibility network of a protein's native structure. I showed that the conventional topological aspects of these networks are not statistically correlated with the phi-values, for the limited data that is available.
İklim görselleştirmesinde sebep-sonuç ilişkisinin dokunma hissi kullanılarak aktarılması
We investigate the potential role of haptics in augmenting the visualization of climate data. In existing approaches to climate visualization, different dimensions of climate data such as temperature, humidity, wind, precipitation, and cloud water are typically represented using different visual markers and dimensions such as color, size, intensity, and orientation. Since the number of dimensions in climate data is large and climate data needs to be represented in connection with the topography, purely visual representations typically overwhelm users. Rather than overloading the visual channel, we investigate an alternative approach in which some of the climate information is displayed through the haptic channel in order to alleviate the perceptual and cognitive load of the user. In this approach, haptic feedback is further used to provide guidance while exploring climate data in order to enable natural and intuitive learning of cause and effect relationships between climate variables. As the user explores the climate data interactively under the guidance of wind forces displayed by a haptic device, we believe that she/he can understand better the occurrence of events such as cloud and rain formation and the effect of climate variables on these events. We designed a set of experiments to demonstrate the effectiveness of this multimodal approach. Our experiments with 22 human subjects show that haptic feedback significantly improves the understanding of climate data and the cause and effect relations between climate variables as well as the interpretation of the variations in climate due to changes in terrain.
Protein dönme açılarını tahmin etmek için bilgi tabanlı yöntem
The three dimensional structure of a protein can be identified in terms of its torsion angles. These torsion angles can be considered as the degrees of freedom of a protein. In this study, a method grouping these torsion angles in different rotational isomeric states and estimating their probabilities is developed. Specifically, the probabilities of the various torsion angle states in Ramachandran maps is proposed and the accuracy of the method is examined using a knowledge based approach. Statistical independence and dependence of the states of different residues along the peptide chain are analyzed. The Flory isolated pair hypothesis, near neighbor correlations, context effects and long-range correlations are discussed. In the knowledge based approach, two different protein libraries i) coil library ii) full libarary are constructed and information from both these libraries is used. Results showed that amino acids have propensities for some rotational isomeric states that favor the choice of the native state torsion angles and they are context dependent, preferring different torsion states determined by the amino acid sequence of the protein. Context dependency is also related to chameleon sequences and the effect of chameleon sequences is also integrated into the method.
Protein-protein arayüzeylerinin nitelendirilmesi ve analizi
The diverse range of cellular functions is performed by the limited number of protein folds existing in nature. One may similarly expect that the number of protein-protein interface architectures would also be restricted. In this study, the recently derived dataset of protein-protein interfaces is analyzed and compared with older datasets to address questions like (i) how many different protein-protein interaction types are expected to exist in nature for necessary biological diversity; (ii) what fraction of interactions is already known toward elucidation of the organization of the cell; and (iii) whether the increase in the number of interface architectures and consequently the functional coverage and interaction map of the PDB are reaching a plateau. The results show that number of protein interfaces increases at a much faster rate compared to the number of folds and is not yet to level off. Functional coverage is also found to steadily increase. As an estimation of this study, the total number of different interfaces will be around 8000 and it will take almost 30 years to discover at the current rate of experiment. Also, despite the diversity of interface architectures, some are more favorable and frequently used, and of particular interest, those are the ones which are also preferred in single chains. Another significant result is that some species, especially eukaryotes, prefer intra chain domain-domain interactions; others, less complex organisms, prefer inter chain interaction. This adaptation may be the result of the crowded traffic in the eukaryotic cells. This thesis presents the multidirectional analysis and applications of the protein ? protein interfaces. We believe in that the dataset of protein ? protein interfaces is a rich source for researchers dealing with protein ? protein interactions, protein recognition mechanisms, drug design etc.
Doğal halde olmayan peptitlerin konformasyonlari: Kabataslak model gösterimi
The native structure of proteins is stabilized by both local and non-local interactions. The available phase space is highly reduced in size due to the local interactions, but non-local interactions determine the final, physiologically active native structure. Although it is wellknown that these two types of interactions are the key factors in determining the tertiary structure, their relative contributions is open for debate. This study will lay the groundwork for the investigation of relative contributions of these interactions. The contribution of both local and non-local neighbors to the Ramachandran map of the residue in question is examined by using statistical weight matrices (U) constructed according to the Markov assumption. An efficient matrix multiplication scheme based on rotational isomeric states model is introduced for studying realistic conformations of homotripeptides of all-alanine, tryptophan, valine, and tyrosine and AXA tripeptides, where X represents alanine, valine, tryptophan, and tyrosine in the unfolded state. This scheme is based on U?s obtained from mono and dipeptide molecular dynamics simulations. By using these matrices one can obtain the Ramachandran map of the central residue of longer sequences, such as tripeptides. Comparison of explicit tripeptide simulations with the Markov model shows that the Markov assumption fails to capture interactions specific to the tripeptide. Here, a systematic correction is proposed for efficient calculation of realistic protein conformations. Preliminary results suggest that the Markov assumption can be improved significantly by adding the contributions from hydrogen bonds, which are only present in the tripeptide sequences. Such a coarse-grained model, Modified Markov model, will help elucidate the protein folding problem and improve secondary structure prediction algorithms.
Güvenilir içerik dağıtımı için yeni bir ara bellek yönetim modelinin tasarım ve analizi
For supporting reliability in distributed content dissemination services, message loss recovery mechanism achieved via efficient buffer management is an indispensable component. The available approaches for buffer management concentrate on several aspects of the problem such as flow control, reducing the memory usage, providing message stability and the replacement of buffer items. In this thesis study, we consider buffer management problem in support of large-scale bio-inspired peer-to-peer data dissemination services. Bio-inspired epidemic protocols have considerable benefits as they are robust against network failures, scalable and provide probabilistic reliability guarantees. Coupled with an efficient buffering mechanism, system wide buffer usage can be optimized while providing reliability and scalability in such protocols. We propose a novel algorithm, Stepwise Fair-share Buffering, that is shown to provide uniform load distribution in comparison to earlier approaches and reduces the overall buffer usage where every peer has the partial view of the system. A major aim of our approach is to be able to choose bufferers uniformly throughout the system so that the load of buffering will be well balanced among participating peers and the efficiency of content dissemination will be improved as a result. This also reduces the memory usage since only a small subset of the peers is chosen as bufferers for each message. Furthermore, it is applicable to large-scale scenarios, provides reliable delivery and is adaptable to dynamic join and leaves to the system. It adjusts the buffer size to achieve message stability with a high probability. Performance evaluation of the buffering model and extensive comparisons with earlier approaches are performed. The evaluations include scalability, reliability, adaptivity to failures and uniformity analysis. We also derive analytical results for reliability of dissemination as a function of buffer levels. These results are based on a Markov chain analysis and are evaluated numerically. Comparison with simulations shows that they provide a good lower bound for reliability. For high level of reliability values, the bounds are very close to the simulation results.
Yapıya-dayalı ilaç dizayn yöntemi ile prostat kanserine tedavisine yönelik ilaç geliştirilmesi
Structure-Based Drug Design is a powerful method for designing inhibitors with high specificity. This method can be used for diseases where a single biochemical or structural function can be targeted for treatment. Structure-Based Drug Design has been successful in developing drug for several diseases involving hypertension, influenza, cancer and AIDS.In this thesis, the objective is to design novel inhibitors for the treatment of prostate cancer using structure-based drug design approach. Prostate Cancer is the most common malignancy among males. There are several treatment procedures for prostate cancer. However, many cases are left untreated due to severe side effects and resulting decrease in quality of life. Nonetheless, inhibition of a single enzyme, CYP17, offers a specific treatment possibility with very low risk of side effects. Molecular dynamics methods are used to validate the model structure of CYP17. Computational tools for molecular docking are employed for discovering and developing novel agents inhibiting CYP17. Experimental approaches are followed for testing biological activity and toxicity of these molecules.
Protein ikincil yapılarını optimal olarak katlayan amino asit çiftleri arasındaki potensiyellerin hesaplanması
We present a method to calculate the pair potentials for folding of protein secondary structures. In the first part of the method necessary training data are generated to compute the potentials. For this purpose a Go-type model and dynamic optimization is used to compute the optimal folding trajectories. A coarse-grained model for a helix and a beta sheet, each consisting of 12 residues has been constructed by representing each amino acid as a bead. The dynamic optimization gives the total optimal force acting on each residue (bead) to fold the protein from an initial configuration to its native state. Next, forces between pairs of residues are derived from this data. This is done by first projecting the optimal residue forces onto the pair-wise directions between residues and expressing these mean forces as (nonlinear) functions of pair-wise distances. We show how to compute the forces between pairs from the mean forces. We next incorporate the derived pair forces into the dynamic model. Thus, for new initial conditions folding is achieved in a predictive way by simulating this model without any need for optimization. We further show that the folding pathways obtained by such ?simple? simulation are similar to folding pathways which can be obtained by the rigorous dynamic optimization. To measure similarity between folds we use MPCA (Multi-way Principal Component Analysis). In addition, mean forces between pairs are presented and analyze
Doğrusal olmayan viskoelestik nesnelerin deneysel veriye dayalı parçacık tabanlı modellenmesi
Simulation-based training using Virtual Reality techniques is a promisingalternative to traditional training in minimally invasive surgery. Surgical simulators let thetrainee touch, feel, and manipulate virtual tissues and organs through the same surgical toolhandles used in actual minimally invasive surgery while viewing images of tool-tissueinteractions on a monitor as in real laparoscopic procedures. Developing realistic organforcemodels for simulating soft-tissue behavior is an integral part of a surgical simulator.The particle system approach provides a better solution than mesh-based methods tothe topological changes encountered in simulation of surgical cutting and tearing. Inaddition, they are computationally less expensive and easier to implement than the meshbasedmethods. However, the material coefficients of each individual mesh element shouldbe calculated and fine tuned to integrate the realistic tissue properties into particle models,which is not trivial.This thesis presents an end-to-end solution to realistic particle-based simulation ofnonlinear viscoelastic tissue behavior based on the experimental data collected by a roboticindenter. First, the strain-dependent nonlinear elastic response and time-dependentviscoelastic response of a tissue-like silicon phantom is measured via static loading andstress relaxation experiments performed by a robotic indenter. The collected experimentaldata is used to construct a lumped model of the tissue phantom represented by a nonlinearviscoelastic Maxwell Solid. Then, a 3-dimensional particle-based network is developed tomimic the behavior of the lumped Maxwell model. The material coefficients of theindividual Maxwell elements connecting the particles are estimated through a set of noveloptimization algorithms.
Yapısal bilgiyi kullanarak insan protein-protein etkileşim ağının ve kanser proteinlerinin analizi
Protein-protein interaction networks reveal that some proteins are highly connectedto others (acting as hub proteins), whereas some others have a few interactions. Thesame or overlapping binding sites should be repeatedly used in hub proteins (singleinterface hub proteins) making them promiscuous. Alternatively, multi-interface hubproteins make use of several distinct binding sites to bind to different partners.Understanding the interactions with respect to their physical and chemical propertiesrequires the atomistic details of the proteins, namely the three-dimensional structures.Then again, cancer-related proteins are more likely to act as hubs in interactionnetworks. In this thesis, we investigate ?what features of cancer-related proteininterfaces make them act as hubs? and ?how it is possible for them to bind to manydifferent proteins with varying affinity?. We provide a detailed analysis of humanprotein-protein interaction network including cancer-related interactions. First weanalyze the global behavior of cancer-related proteins, second we hold a structuralperspective to elucidate how these proteins interact and figure out which interactionscan occur simultaneously and which ones exclude each other. The results reveal thatcancer-related proteins tend to interact with their partners through distinct interfaces,thus corresponding mostly to multi-interface hubs (56% of cancer-related proteins aremulti-interface) and constituting the nodes with higher essentiality in the network (76%of them are essential). In addition, they have smaller, more planar, more charged andless hydrophobic binding sites compared to non-cancer ones which may indicate lowaffinity and high specificity of the cancer-related interactions. These findings might beimportant in obtaining new targets in cancer as well as finding the details of specificbinding regions of putative drug candidates in cancer.
Afet sonrasında ihtiyaç duyulacak yardım malzemesi için stoklama kararları
Natural disasters are unexpected crisis events causing devastating human and financial losses. These events trigger a critical need for effective preparedness, mitigation, response and recovery operations to reduce the impact of the disasters.Humanitarian relief agencies participate in massive relief efforts to provide life-supporting resources, such as food, water, sanitation, emergency care, shelter and other essential non-food items; to distribute supplies and to coordinate international aid after a disaster. The success of such large scale and time-critical operations require effective pre-disaster planning.In this thesis, we consider stocking decisions for humanitarian relief agencies that provide emergency relief items to people affected by natural disasters. We present a mathematical model to determine the optimum stocking quantity for one type of relief commodity. The proposed model is an extension of the well-studied newsvendor model that incorporates the disaster risk. The probability that a disaster takes place within the lifetime of the stocked commodity and the demand distribution for this commodity under such a disaster are taken into consideration in the model. We extend our model to determine the optimum stocking quantities for two agencies that stock the same commodity at different locations prone to differing disaster risk, and that work in full cooperation, just like the case of the Turkish Red Crescent and the International Federation of Red Cross and Red Crescent agencies. An equilibrium solution to the model that can be calculated numerically is derived. We investigate the characteristics of the solution under various parameter settings and identify cases where cooperation is beneficial to one or both of the agencies. We analyze the case for Istanbul to demonstrate the use of this modeling approach. The probability of a major earthquake occurrence in Istanbul and the potential demand for relief commodities throughout the city are estimated and used as inputs of the model. A numerical analysis gives insights to the potential benefits arising from cooperation of an agency in Istanbul with an outside agency. Our analysis with respect to realistic scenarios and parameter estimations provides useful guidelines for the relief agencies in Istanbul.
JMJD2A enziminin metilasyon spesifisitesinin karşılaştırmalı moleküler dinamik çalışması
Specific patterns of post-translational modifications of histones act as a molecular ?code? recognized and used by non-histone proteins to regulate specific chromatin functions. K9 methylation on Histone 3 (H3) tail, mainly trimethylation, induces formation of constitutive heterochromatin via a well-known pathway, which employs heterochromatin formation protein (HP1) and DNA methyl transferase (DNMT). Jumonji domain containing 2A (JMJD2A) is a histone demethylase that specifically removes K9 and K36 trimethyl marks on H3 tail. This enzyme does not function on monomethyl marks and has almost 20-fold reduced activity on dimethyl forms compared to trimethyl forms.In order to gain insight into how JMJD2A discriminates between its substrates, we performed molecular dynamics simulations of mono-, di- and trimethylated histone tails in complex with JMJD2A catalytic domain and analyzed positional fluctuations, located the hydrogen bonds and calculated some critical distances. We revealed the importance of water molecules and the oxygen-enclosed environment in appropriate orientation of methylammonium head in the active site. We also calculated binding free energy and energy contribution of each residue. We found out that recognition is mostly driven by van der Waals and Coulombic interactions in enzyme-substrate interface. We also revealed the role of Arg8 on the H3 tail in binding and stabilizing the necessary conformation of substrate peptide.
Metillenmiş histon peptitlerinin JMJD2A enziminin bitişik tudor domenleri tarafından moleküler düzeyde tanınmaları
In this thesis, we report a detailed molecular dynamics simulation and MM-PBSA/GBSA approach (MM: Molecular Mechanics; PB: Poisson Boltzmann; GB: Generalized Born; SA: Surface Area) analysis, unraveling the recognition of the methylated histone tails H3K4me3, H4K20me3, H4K20me2 and H3K9me3 by JMJD2A-tudor. In this respect, 25 ns fully unrestrained molecular dynamics simulations were conducted for each of the bound and free structures. We investigated the important hydrogen bonds and coulombic interactions between the tudor domains and the peptide molecules; hence unveiled critical residues occupied in stabilizing the complexes. Normal mode and molecular mechanics calculations were performed to obtain the entropic determinants of the binding affinities. Suggested by the resulting binding free energies obtained via GB and PB approaches, we found that H4K20me3 peptide has the highest affinity to JMJD2A-tudor in GB calculations whereas H4K20me2 peptide has the highest affinity to JMJD2A-tudor in PB calculations. Furthermore, we discerned that H3K9me3 peptide has the lowest affinity to JMJD2A-tudor in both of the model calculations. We also revealed that while H4K20me2 peptide adopting the same binding mode with H4K20me3 peptide, H3K9me3 peptide adopts the same binding mode with H3K4me3 peptide. Decomposition of the enthalpic and the entropic contributions to the binding free energies indicated that the recognition of the histone peptides is mainly driven by favourable van der Waals interactions in both GB and PB models. Based on GB calculations pairwise and per residue decomposition of the binding free energies with backbone and sidechain contributions as well as their energetic constituents were also carried out to identify the hotspots of the structures. Thus, the van der Waals and the electrostatic interactions which are prominent for the recognition of the peptides were clarified.
Histon kuyruğunda bulunan lizin amino asitlerinin demetilasyonuna yönelik kuantum mekanik (KK) ve melez kuantum mekanik/moleküler mekanik (MM) yöntemleri kullanarak reaksiyon mekanizması analizi
Chromatins, the basic structural units of the genetic material, consist of DNA and histone proteins. Eukaryotic DNA is wrapped around the histone proteins and they form `bead-like? structures. Histone proteins control many crucial cell regulatory processes, e.g. gene transcription, gene silencing, DNA replication and repair and etc., via post-translational modifications. Among the post-translational modifications, methylation was recently shown to be reversible by the discovery of Lysine-Specific Demethylase (LSD1) enzyme. As many previous studies have shown the relation of some cancer types and other diseases with the abnormalities in the balance of methylation/demethylation, drug molecule design based on the information gained from reaction mechanism studies becomes very crucial for the fight against these diseases.In this thesis, a chemically-reliable reaction mechanism is proposed for the demethylation of histone tail lysine residues and the reaction path analysis of this mechanism is carried out. Specifically, demethylation of H3 tail fourth lysine residue, i.e. H3K4, is analyzed. Potential and free energy profiles as well as structural properties are calculated using available Quantum Mechanical (QM), i.e. PM3, B3LYP and MP2, Molecular Mechanical (MM), i.e. UFF, as well as the hybrid QM/MM methods, i.e. ONIOM, which are implemented in Gaussian09 software package.As the result of the calculations, it is proved that the proposed chemical mechanism is actually simulating the real-life process and suitable for the demethylation of mono- and dimethylated lysines found on the histone tails. This comment is based on the calculated reaction rates, which are in high agreement with the experimental observations. Besides, some important observations made on the chemical mechanism, which enhances the understanding of how the demethylation process occurs at the molecular level, are explained and discussed in the thesis in fine details.These results offer an understanding for the details of the reaction mechanism, which will form a fundamental knowledge basis for further studies involving inhibitor molecules. By comparing the standard (i.e. in the absence of any inhibitor molecule) energy profiles and thermodynamic properties with the ones obtained in the presence of the inhibitor candidate, or directly comparing these profiles for different inhibitor candidates, one can have an understanding of the efficiency of the drug candidate for regulating the de/methylation balance. Knowing the key points in the reaction mechanism, one can design novel inhibitor molecules that are inspired by the reaction mechanism.
Proteinlerde bulunan ligand bağlanma yerlerinin ağ yapı modeli kullanarak tespit edilmesi
Biomolecular interactions play key roles in biological activity. Investigation of those interactions, including protein-ligand interactions, is crucial for understanding the way that nature designed its biological machinery. Ligand binding particularly requires, recognition of the ligand by the protein, which in turn arranges the three dimensional structure of the protein, mostly directed by the energetic interactions involved. Based on these requirements, ligand-binding has been considered as a local process. Yet, it has been recently shown that ligand binding depends not on the local structure, but rather on an interaction pathway, that takes part in rearrangement of the protein into the most favorable conformation upon binding.The nonlocal nature of the protein-ligand binding problem is investigated via the Gaussian Network Model with which the residues lying along interaction pathways in a protein and the residues at the binding site are predicted. The predictions of the binding site residues are verified by using several benchmark systems where the topology of the unbound protein and the bound protein-ligand complex are known. Predictions are made on the unbound protein. Agreement of results with the bound complexes indicates that the information for binding resides in the unbound protein. Cliques that consist of three or more residues that are far apart along the primary structure but are in contact in the folded structure are shown to be important determinants of the binding problem.Comparison with known structures shows that the predictive capability of the method is significant.
Protein-protein etkileşimlerinin çoklu ölçekte analizi ve tahmini
Proteins act coherently in the cells and their roles span functions as diverse as being molecular machines and signaling. The mechanism behind this excellent synchronization is still uncovered. However, considerable effort has been centered on identifying of binding partners and binding regions, because the vast majority of the chores in the living cell involve protein?protein interactions. Proteins interact through their interfaces which contain hot spots, the residues contributing more to the binding energy. Hot spots are important for drug targeting and interaction specificity. In addition, structural modeling of protein interactions and incorporating them into the protein interaction networks are prerequisites for understanding cell function. Hence, the focus of this dissertation is directed to the question ?how do the proteins interact?? rather than the question ?which proteins interact?? at the top level. Towards this aim, firstly, this dissertation focuses on the prediction of hot spots in protein interfaces and their organization. Here, an efficient hot spot prediction model is developed and implemented that reaches an accuracy of 70% on the experimental data. A web server, namely HotPoint, is constructed based on this model. In another aspect, a novel graph-based method based on minimum cut trees developed to determine the organization of hot spots which reveal the cooperative relation between them. Nature presents a limited number of distinct binding site motifs and structurally different protein pairs can use the same binding architectures. Based on this origin, secondly, a multi-scale combinatorial strategy is illustrated to model protein complexes at proteome-level. This work shows how available structural information can help in modeling a pathway by using structural similarity. Here, the sample pathway is the tumor suppressor protein p53 pathway. Finally, the multi-partner proteins dataset is extracted from Protein Databank. Integration of time notion into protein interaction networks is demonstrated on two hub proteins, p53 and Mdm2 using both predictions and available structural data.
Proteinin içsel normal modları ve farklı protein konformasyonları arasındaki ilişki
Proteins are indispensible components of cellular functions. Although proteinstructure is determined as a static picture for most of the proteins, the dynamics or thetime-dependent behavior of proteins act as main contributors to protein function. Inaddition to ligand induced structural motions proteins also bear intrinsic motions arisingfrom thermal energy they contain. These ligand-independent motions possess functionalimportance according to experimental evidences for a large number of proteins and alink between these motions and functional motions are established both in terms ofstructure and timescale. These intrinsic fluctuations are revealed by low-frequencyindividual modes of proteins which are determined using a simplified version of normalmode analysis termed as Anisotropic Network Model (ANM). In this study, we applymodal analysis to eleven proteins including enzymes, antibodies and signal proteins.We investigate a kinetic relation between modal analysis and protein motions. For thispurpose we employ eigenvalues of Hessian matrix which carry information about thevibrational frequencies of these modes. In ANM studies these eigenvalues are used todetermine the low and high frequency modes of protein. Our findings imply acorrespondence between eigenvalues and kinetic/thermodynamic properties of proteinmotions. These intrinsic motions establish a dynamic equilibrium between distinctconformers of the protein and as we have proposed eigenvalues of Hessian matrixcorrelate well both with the timescales of these motions and thermodynamics of thesemotions.
Protein bağlanması ve proteinlerde mod bağlanması
In order to understand the the change in thermodynamic properties upon binding and determine the binding sides, two hexa-peptides and their bound complex structures were analyzed. In order to extract the thermodynamic properties and determine the binding side, a harmonic model was applied.The harmonic formulation is extended to large ? uctuations of residues in order to account for effects of anharmonicity. The ? uctuation probability function is constructed for this purpose as a tensorial Hermite series expansion with higher order moments of ? uctuations as coef ? cients.Mode coupling and anharmonicity in a native fluctuating protein is investigated in modal space. Molecular dynamics trajectories of Crambin are generated and used to evaluate the terms of the polynomials and to obtain the modal energies. Slowest modes have energies that are below that of the harmonic energy, kT/2 per mode, and a few fast modes have energies significantly larger than the harmonic which is a result of coupling. Detailed analysis of the lowest order two mode coupling terms is presented.It is was shown that mode coupling and anharmonicity are important for modeling the multidimensional energy landscape of Crambin. The effect of them on the fluctuational entropy is on the order of a few percent.The fluctuations and unbinding free energy profiles of two very similar proteins, HLA-B51 and HLA-B52, were investigated. HLA-B51 is related to the Behçet?s disease whereas HLA-B52 is not. Change in the dynamics of 1 helix were analyzed. Unbinding from HLA-B52 resulted in greater free energy differences than for HLA-B51.
Peptid dizayn stratejileri
Short peptide segments have gained importance as drug candidates. There exist three main problems for peptide design: determining the appropriate sequence with the desired function; properly docking peptide on the protein surface; and the unbound state of the peptide that is to be used as a drug. The `unbound state? means peptide chains in the denatured state at physiological conditions. The details of the potentials for peptide docking simulations and the statistical features of peptides are defined in the literature.As a solution to the peptide sequence determination problem, several experimental and in silico techniques exist to screen peptides. There is no general computational tool to determine peptide sequences. On the other hand, peptide motifs are crucial for selective and specific binding. There have been successful attempts to discover biological motifs by different research groups. To our knowledge, the efforts in the literature are based on the alignment of evolutionarily conserved motifs from proteins. The evolutionary peptide motif search algorithms/servers/software are available. However, there is no general methodology to discover a binding peptide motif for any protein target. We aim to predict peptide sequences and peptide binding motifs for any given protein using no prior information. Here, four different algorithms are developed for peptide design. The implementation of genetic algorithm, Markov model and hidden Markov model with Viterbi decoding leads to prediction of peptides for different protein targets. The algorithms are successful to determine peptide sequences with good theoretical binding affinities. The peptide motifs for two case-studies are also offered. A web-server, VitAL, is constructed based on Viterbi decoding.The statistical thermodynamics features of the unbound peptide as a small thermodynamics system in a thermal reservoir is lacking in the literature. A novel statistical thermodynamics approach is applied to the free peptide segments in order to classify them according to their conformational energies and entropies and heat capacities. The conformational partition function, Helmholtz free energy, energy, entropy and heat capacity are obtained. The model is applied to randomly produced peptides and to known peptide inhibitors. Peptides with low energy, low entropy and low heat capacity are determined to be essential for a peptide to be a good candidate inhibitor.
Düzenleyici gen topluluklarının saklı Markov modelleri kullanılarak zaman serisi mikrodizi ekspresyon profillerinden belirlenmensi
Time series microarrays capture multiple gene expression levels at discrete time points varying from minutes to days of a continuous cellular process. Analysis of high through put data requires automated and computer aided solutions. We propose a hidden Markov model (HMM) based approach to identify regulatory relations between the periodic genes from the cell cycle time-series microarrays. We train and test our models by using distinct types of biological data present literature. In our study we use Pramila time series dataset. Training gene pairs include transcriptional regulation and protein level regulation. After identification of gene to gene regulatory relationships, we form a network of gene regulation relationships: Gene Regulatory Neighborhood Networks (GRNN). We explore potential use of sub networks (communities) in GRNN by comparing gene clusters found by popular clustering algorithms such as K-means clustering. Our results indicate we manage to identify denser and more specific enrichment in community structure based clusters than the clusters acquired with K-means.
Peptit bazlı nanomalzemelerin moleküler dinamik simulasyonlar aracılığı ile yapısal ve termodinamik özelliklerinin analizi
Peptides are oligomers with aminoacids as building blocks. Discovery of naturallyoccuring functional peptides has led to a dramatic increase in research for both under-standing of natural peptides as well as design of novel synthetic ones. As individualmolecules, peptides serve a variety of purposes: such as drugs, antigens, ligands andantibiotics. They also act as building blocks for self-assembled nanofibers, nanotubes,micelles, and monolayers. Designing novel peptide-based materials with desired prop-erties and functions can only be possible by understanding the link between sequence,structure and organization of these molecules.This thesis work is composed of four main sections. In the first study, we analyzedstructure and thermodynamics of small amphiphilic peptides that spontaneously formmonolayers at the air/water interface. We accurately calculated the free energy oftransferring peptides from bulk water to the air/water interface, and analyzed itscomposition. Next, in the light of information gathered from small amphiphilic pep-tides we analyzed folding of a carefully designed 24-residue amphiphilic peptide atthe air/water interface. We calculated free energy of adsorption, and decomposedit into enthalpic and entropic contributions. We determined key elements requiredto adapt the targeted ß-hairpin conformation and for adsorption at the interface viain-silico mutations. In addition, we also determined organization of ß-hairpins withinsurface monolayers, which will help improve design strategies for manufacturing such2-D structures. In the third study, we tried to understand self-assembly of tri-blockpeptides into nanofibers in bulk water. By substituting aliphatic residues with aro-matic ones located in the central block of these peptides, we analyzed the stabilityand strength of these fibers. We also investigated possible nucleation mechanisms forfiber formation, which can be used to design stable and functional peptide nanofibers.Investigation of material properties and bulk behavior of peptides requires moreefficient computational techniques. In this regard, finally, we investigated the ?trans-ferability? of a recently developed solvent-free coarse-grained (CG) peptide model [1]that was shown to capture quantitatively structural and thermodynamic propertiesof a hydrophobic di-phenlyalanine peptide (FF). By mimicking the CG mapping usedin the FF CG model and transferring bonded and nonbonded interaction potentialsto other hydrophobic di-peptides, namely valine-phenyalanine (VF) and isoleucine-phenylalanine (IF), we tested the generality of this CG model. We devised a generalprotocol to transfer the original CG model to other hydrophobic di-peptides, whichare in the form of XF, where X represents an arbitrary amino acid. Hydrophobic di-peptides are the smallest molecules forming self-assembled hierarchical structures inaqueous solution. Therefore, development of a transferable CG model for simulationof such systems will help elucidate the driving forces important in the self-assemblyof peptide-based materials and peptide aggregation.
Yapısal bilgiyi kullanarak PER-ARNT-SIM (PAS) bölgeleri içeren biyolojik saat proteinlerinin etkileşimlerinin analizi
Per-ARNT- Sim (PAS) domains are modular protein units those are critical for regulation of clock-controlled gene expression. The mammalian BMAL1 and CLOCK are the transcription factors that contain two basic helix-loop-helix domains and bind E-box elements (CACGTG) of clock regulated genes including the Period and Cryptochrome and activate their transcription. Then the PERIOD (PER) and CRYPTOCHROME (CRY) proteins form ternary complexes with casein kinase I? (CKI?) in the cytoplasm and translocate into the nucleus, where they act as a negative regulator of BMAL1/CLOCK-driven transcription. To understand the nature of interaction between PER2-CLOCK-BMAL1 complex, we performed structure based analysis on protein-protein interactions (PPI) those formed via PAS domains of clock proteins. This complex was analyzed by using structural data and efficient structural comparison algorithms to predict potential interactions. Since there are no available atomic structures for our proteins of interest, homology models are used. In our model, BMAL1 and CLOCK interacts with each other through their PAS B domains and the PER2 interacts with this dimer through the PAS B domain of the CLOCK. On BMAL1/CLOCK interface, we found 12 hotspots, 7 residues on BMAL1 (347,349,362,405,427,429,441), 5 residues on CLOCK (317, 338, 350, 352,376). On PER2/CLOCK interface, 8 hotspots are found, 3 residues on PER2 (414, 429,431) and 5 residues on CLOCK (332,354,356,333,361). Here we show how, using structural data and efficient comparison algorithms can explain forming the clock complexes at the molecular level. This study is not only important to understand clock mechanism at structural level but also will allow us to develop drugs against clock-regulated diseases, like Jet-Lag and some forms of depression.