NUROP2013 @School of Computing

National Undergraduate Research Opportunities Programme Congress 2013

Nurop2013@SoC > Parallel Abstracts

Parallel Session A - Abstracts

	Title
1	Parallel Speaker: Shailendra Khemka Title: Autonomous Search for Victims in a Disaster Situation Supervisor: Assistant Professor Low Kian Hsiang Bryan Department: Department of Computer Science, School of Computing Abstract: In disaster situations, there is a significant time gap between when the incident occurs and help actually arrives. Moreover, even when help arrives, human rescuers cannot begin the rescue mission until the situation has been bought under substantial control; for example, in case of fire, human rescuers cannot enter the structure until the fire has been suppressed to a certain level. A lot of lives are lost during this time gap, which is often the most crucial time in disaster situations. Through this UROP, we will explore search strategies that a robot stationed at a building can employ to provide wounded victims with safety supplies during the time gap described above. If the victims can get the safety supplies in this time gap, a lot of lives can be saved; for instance, in case of fire breakout, the safety supply could be a fire suit, or in case of a poisonous gas leak, it could be a gas mask. We will explore efficient search strategies that a robot can pursue to locate a victim.
2	Parallel Speaker: Sudhanshu Khemka Title: Accelerating Language Model Based Information Retrieval using the Graphics Processing Unit Supervisor: Associate Professor Kan Min-Yen Department: Department of Computer Science, School of Computing Abstract: Language models (LMs) have been a core area of study for their applicability to information retrieval (IR) problems for both their probabilistic, grounded framework as well as their retrieval efficiency. With current technologies and hardware, the graphics processing unit (GPU) has become a staple in many computing devices, from desktops to supercomputer clusters. While the GPU can be gainfully employed for high throughput computation, it has not yet been applied to the formalism of LM based IR. We illustrate how the GPU can significantly accelerate both the indexing and retrieval processes. Our adaptation is novel as we contribute algorithms to realize both the offline indexing phase and the online search phase of a LM-based IR system on the GPU. To speed up offline indexing, we contribute a GPU + CPU based variant of Huffman compression. To achieve faster query processing, we present data parallel algorithms for inverted list decompression, document score computation and top-k scoring. Our experimental results demonstrate that for medium-sized datasets, one can achieve significant speedups by implementing the two phases on the GPU.
3	Parallel Speaker: Zhang Jiangwei Title: Intelligent Assistants in Collaborative Game Supervisor: Associate Professor Lee Wee Sun Department: Department of Computer Science, School of Computing Abstract: Real time collaborative games require players to collaborate to achieve a common goal. An intelligent assistant, or AI, is a software agent that cooperates with other human players to achieve the common goal. It is always an interesting research topic that how to design the AI probably so that the AI could surpass human beings. Moreover, putting a real time game online is also challenging since it requires the synchronization between the clients. In this study, two major problems are examined. Firstly, the problem of putting a real time collaborative game online is examined. Various techniques are used to enable the online communication between two clients. Moreover, to ensure the synchronization between two clients, one unique synchronized state is maintained between two clients. Secondly, the original algorithm is updated. In the original AI, it uses shortest distance method to choose the best action for the AI. However, it has limitations when the distance between AI and the player is too large. In the updated AI, to overcome the limitation, the AI will search for a cycle whenever it is necessary. Most of the techniques used are universal and can be applied to other online real time collaborative games.
4	Parallel Speaker: Aldrian Obaja Muis Title: XML Curated Database (Recognizing Tables in Wikipedia) Supervisor: Associate Professor Stéphane Bressan Department: Department of Computer Science, School of Computing Abstract: XML and XML technologies promise to enable the intelligent integration of information for disparate information sources. In this project we design and develop a curated database. The curated database is a warehouse of maritime data (port, shipping, etc.) fed on data extracted from diverse sources including the web, in which data is integrated, cleaned and in which metadata, data quality and lineage are maintained. The interface of the database is exible as little or no information on the schema or domain model is known a priori. As an initial work on this topic, in this paper we will discuss mostly about the smaller scope of the data collection part, which is recognizing and extracting data from HTML tables in Wikipedia.
5	Parallel Speaker: Koh Xue Ming, Brian Title: Efficient roadmap construction and updating in dynamic environments Supervisor: Dr Goh Siew Wee, Alvina Department: Department of Mathematics, Faculty of Science Abstract: Autonomous vehicles are no longer a mere fantasy. What considerations and methods are involved in the pathfinding process? This talk briefly touches on some of the aspects involved, including visibility graph searching and construction. A couple of basic examples will be used to highlight the idea of developing efficient algorithms. Finally, we consider the introduction of a dynamic element into the landscape.
6	Parallel Speaker: Navneeth Ramakrishnan Title: Investigation into Non-cyclic phase in neutrino oscillations Supervisor: Associate Professor Chan Aik Hui, Phil Department: Department of Physics, Faculty of Science Abstract: In this project, we introduce the idea of neutrino oscillations and how it can be treated with a non-cyclic phase. Neutrino oscillations refers to the phenomenon where the flavour of a neutrino changes periodically with time. Clearly, the flavour eigenstates of neutrinos cannot be the eigenstates of the free Hamiltonian (if they were, we would never observe the flavour of a neutrino changing). Instead, we can express the flavour eigenstates as a linear combination of the eigenstates of the Hamiltonian. We first derive the form of the mixing matrix that connects the flavour eigenstates and the Hamiltonian eigenstates. We do this explicitly for the two flavour case and generalize the result to an arbitrary number of neutrino flavours. This results in a mixing matrix that depends on (n − 1)2 parameters for the n flavoured neutrino oscillation problem. The task is to now conceive of an experimental way to measure these parameters. The geometric phase is a phase acquired when a system is subjected to an adiabatic cycle. That is, some parameter in the Hamiltonian is changed in a cyclic manner and this causes the eigenstates of the Hamiltonian to pick up a geometric phase. Although the flavour eigenstates are not eigenstates of the free Hamiltonian, if one looks at the state at the end of each oscillation cycle, they do behave like eigenstates and do acquire a geometric phase. Furthermore, this geometric phase is related to the parameters of the mixing matrix. In principle, one could allow neutrino beams to interfere and calculate the geometric phase from such an experiment. We explicitly derive an extension of this result: the non-cyclic phase. Here, the experimentalist is allowed to measure the phase at an arbitrary time. Our extension relaxes the previous constraint that the phase had to be measured at precisely the end of each oscillatory cycle. In principle, this makes the experimental task much easier and would allow for the calculation of the parameters of the mixing matrix.
7	Parallel Speaker: Zhang Jiyuan Title: Robust Modelling of the Second-Order Flavour Release from Selected Food Model Systems Supervisor: Professor Zhou Weibiao Department: Department of Food Science & Technology, Faculty of Science Abstract: This project aimed to develop a new approach to describe the phenomenon of flavor release from foods and to improve the value obtained from its analytical results. A mathematical model for flavor release was developed with its root in psychological modelling and chemical reactor modeling. It described the second-order flavour release behaviour of flavour compounds and was evaluated in two different food systems, i.e. chewing gum and alcoholic drink. The headspace analyses are very important methodologies to understand the relationships between the relative volatility of flavour compounds and the aroma perceived in different food matrices. Among them, headspace solid-phase microextraction (HS-SPME) is being valued as a practical, relatively fast and solventless method that can facilitate the study on flavour release. To our knowledge, researches on characterising the release of flavour compounds from food matrices using HS-SPME are limited. The major challenges lie in its low reproducibility in quantifying the amount of analytes from complex food matrices. Therefore, a mathematical model may provide a systematic description of the release and adsorption process in the headspace and therefore the corresponding flavor release behavior. The model described the second-order flavor release as a combined effect of two parallel first-order reactions, namely the release kinetics and the adsorption kinetics of volatile compounds onto the fiber. It had four parameters (k1, k2, τ1 and τ2), where k1 and k2 are the equilibrium constants, and τ1 and τ2 are time constants of the reactions. By a two-loop optimization process, the estimated values of the parameters were found to be more reproducible than the direct measurement of the compounds themselves by instrumental analysis. This model was proved to be a robust approach of describing the release of a wide range of flavour compounds in the two different food matrices, at different environmental temperatures. Moreover, with the model’s robustness and the physical significance of the model parameters, more insights were provided to the release behaviour of the flavour compounds. The newly developed model could be used to determine an optimal amount of flavour compounds incorporated in food products and its changes during processing and storage.

Parallel Session B - Abstracts

	Title
1	Parallel Speaker: Wilson Teo Way Seng Title: Synthesis and Characterization of conducting hydrogel adhesive Supervisor:: Associate Professor Hong Liang, Dr Tay Siok Wei (IMRE) Department: Department of Chemical and Biomolecular Engineering, Faculty of Engineering Abstract: Polyelectrolyte hydrogel constitutes hydrophilic polymer network possessing polymeric backbones with pendant ionic groups. Two types of ion conducting hydrogels, poly(acrylic acid-co-N,N’-methylene bisacrylamide) P(AA-MBAM) and poly(2-hydroxyethyl methacrylate-co-ethylenedimethacrylate) (P(HEMA-EGDMA) were synthesized by free radical polymerization in an aqueous medium containing lithium hydroxide as electrolyte. An investigation into the effect of crosslinking degree of the hydrogels and concentration of electrolyte on the ionic conductivity was carried out by using Direct Current Offset (DCO) voltage and Alternating Current Impedance (ACZ) analysis. The ionic conductivity of the hydrogel was largely affected by the mobility of polymer chains, number of ion-hopping sites of the polar pendant groups, and the occurrence of free volume between the polymer networks. The final stage of the study involved the incorporation of a quaternary ammonium monomer made of diethanolamino-2-hydroxypropyl methacrylate, and 2-bromoethanol into P(HEMA-EGDMA) through copolymerization. This structural modification resulted in higher ionic conductivity according to the measurement of the offset voltage and the frequency-dependent impedance. This finally achieved hydrogel demonstrates the key structural factors required for fast ion transport: water content, sweeping role of the pendant hydrophilic groups, a proper crosslinking degree, and interconnectivity of ion migration channels.
2	Parallel Speaker: Grace Pohan Title: Biochemical Modification of Polyvinyl Alcohol Scaffold Mediates Cell Adhesion Supervisor: Dr Evelyn Yim Department: Department of Bioengineering, Faculty of Engineering Abstract: The mechanical property mismatch between blood vessel and synthetic vascular graft is known to cause thrombus formation and intimal hyperplasia. As an aid to this, drug eluting grafts have been utilized. However, when the drug elution lifespan elapses, the risk remains there.[2] In some cases, the use of drugs may prevent endothelialization on the graft.[2] Studies have shown that endothelialization of synthetic graft is necessary to maintain long term graft patency.[3-6] Amongst currently studied materials, poly(vinyl alcohol) (PVA) has an advantage in that it has similar mechanical property to that of human artery.[1] In addition, PVA has excellent film-forming, emulsifying and adhesive properties. PVA also has been widely used in food industry as a moisture barrier coating (Joint FAO/WHO Expert Committee on Food Additives) [7] and known to be non-irritant when it is in contact with soft tissue. Considering all these, PVA has become our choice of material. However, despite the superior mechanical property of PVA, graft patency still needs to be improved because occlusion still occurs occasionally in in vivo testing of such graft. As mentioned, one way to improve the patency is by endothelializing PVA. However due to its high hydrophilicity, endothelial cells lining can hardly be established on PVA substrate. Establishing PVA material which well-mediates cell attachment is the objective in this project. Extracellular matrix proteins are proteins secreted by cells and they mediate cell attachment to a substrate. Therefore, in this study, PVA was modified with extracellular matrix (ECM) proteins including RGDS peptide, fibronectin and heparin. A preliminary result on nanoparticle retainment ability in PVA was also shown. Nanoparticles serve as the protein adsorption point on PVA. The latter is preferable as compared to the bulk modification since less amount of ECM protein will be used, hence less costly, and more consistent results should be produced. In our case, adsorption point is significant because PVA is a hydrogel and therefore simple protein incubation (passive immobilization) will not work. To sum up, PVA is not only mechano-compatible but also bioinert. Since patency of a PVA graft still needs to be improved, efforts must be taken to achieve adhesive PVA: bulk modification of PVA with proteins and protein immobilization on created adsorption points on PVA. The latter is not yet done due to time constraint. However, preliminary data on this was shown. We hypothesize that biochemical modification helps to improve endothelial cell adhesion to PVA. References 1. Chaouat, M., et al., Advanced Functional Materials, 2009. 18: p. 2855-2861. 2. Urban, P. and E. De Benedetti, Lancet, 2007. 369: p. 619-621. 3. Meinhart, J.G., et al., Ann. Thorac. Surg., 2001. 71: p. S327-S331. 4. Seifalian, A.M., et al., Artif. Organs, 2002. 26: p. 209-210. 5. Thebaud, N.B., et al., J. Mater. Sci.: Mater. Med., 2007. 18: p. 339-345. 6. Thomas, A.C., G.R. Campbell, and J.H. Campbell, Cardio Vasc. Pathol. , 2003. 12: p. 271-276. 7. Saxena, S.K., 61st JECFA, 2004.
3	Parallel Speaker: Azmall Fraiszudeen Title: Development of Microfluidic Platform with Nanotopography for 3D micro Muscle tissue Engineering Supervisor: Professor Leo Hwa Liang Department: Department of Bioengineering, Faculty of Engineering Abstract: Microfluidic devices are enabling platforms that can be useful for studies related to high-throughput screening. These devices are cheap, manipulate fluid flows, minimize costly reagent consumption, and enable high-throughput experimentation in a controlled manner. The microfluidic setup can also enhances mass transfer of nutrients, gases and metabolites, provide a spatiotemporal control of delivery of signaling molecules, and create mechanical strain in the physiological range. We hope to develop a microfluidic platform that incorporates topographical cues which have been shown to be beneficial for engineering of certain cell types, like cardiomyoctyes and skeletal muscle cells and cultivate 3D tissues which mimic physiological conditions. One of applications of this biochip include drug testing as Drug induced cardiotoxicity is a leading cause of drug withdrawal from the US market. Early detection of candidate compounds that are cardiotoxic during the in vitro stage would help reduce the incidence of drug induced cardiotoxicty and also save costs for pharmaceutical companies .
4	Parallel Speaker: Wong Zheng Hao Samuel Title: Molecular characterization of gulonolactone oxidase in lungfishes and freshwater stingrays Supervisor: Professor Ip Yuen Kwong Department: Department of Biological Sciences, Faculty of Science Abstract: This study aimed at the molecular characterization of a gene involved in ascorbic acid synthesis, gulonolactone oxidase (gulo), in five species of extant lungfishes, one from Australia, one from South America and three from Africa, and three species of freshwater stingrays, one from Amazon and two from Asia. Phylogenetic analysis revealed that lungfishes formed a clade separated from other species, and indicated a possible correlation between their positions in the phylogenetic tree and their aestivation behaviour. More importantly, results indicated for the first time that gulo could be expressed in non-kidney organs, such as in the brain and lung of Protopterus aethiopicus and Protopterus dolloi, suggesting that ascorbic acid might have important functions in these organs during aestivation. Results reported herein also showed a sister group relationship between two Asian euryhaline freshwater stingrays, Himantura signifer and Himantura oxyrhynchus indicating their divergence from the stenohaline freshwater Potamotrygon motoro. Unlike the African lungfishes, the expression of gulo in the three stingrays was kidney-specific. Since the mRNA expression of gulo in the kidney of H. signifer was down-regulated after exposure to brackish water, this euryhaline stingray might experience greater osmotic and oxidative stress in freshwater than in brackish water.
5	Parallel Speaker: Liu Simin Title: Incorporating Continuum Models of the Solvent into Molecular Energy-based Fragment Molecule Calculations Supervisor: Associate Professor Bettens, Ryan Philip Anthony Department: Department of Chemistry, Faculty of Science Abstract: In this paper, we are interested in assessing the reliability of a proposed fragmentation method by Bettens and Lee in calculating the energy of large biomolecular systems such as polypeptides. For this purpose, secondary structures β-strand, α-helix and 310- helix of polyglycine systems containing 1 to 13 residues were constructed and subjected to full energy calculation in using the Gaussian03 suite of computer programs at the B3LYP level of theory and 6-31G(d) basis set. All calculations were performed with SCRF=PCM, which uses a continuum model to mimic the solvent (water). Similar parameters were used for the fragmentation approach and the energies from both approaches were compared. The largest magnitude of error derived from the proposed fragmentation approach was found to be within an error margin of 4 mHartrees, 10.9 kJ/mol, and is reasonably acceptable. The energies calculated from both approaches were further used to explore structural stability of secondary structures and hydrogen cooperativity in helices. Finally, linear correlation was employed to study the transitional pathway from the 310-helix to the "-helix in water.
6	Parallel Speaker: Le Anh Phuong Title: Expression Analysis of Chondroitin Sulfate Proteoglycan-5 in Phyllodes Tumor Supervisor: Associate Professor Yip Wai Cheong, George Department: Department of Anatomy, Faculty of Science Abstract: Chondroitin Sulfate Proteoglycan-5 (CSPG5) is a part time proteoglycan predominantly expressed in the central nervous system. It is actively involved in cellular differentiation and brain development. It was also found active in context of tumorigenesis; however, little is known of its role in breast tumor. Breast phyllodes tumor is a rare type of breast lesion that has the potential to become malignant and lethal. Its clinical behavior is difficult to predict. This study attempts to evaluate expression of CSPG5 on 342 phyllodes tumor cases by immunohistochemistry. Statistical tests were implemented to investigate diagnostic and prognostic values of this protein. CSPG5 was found in diffuse stroma and stromal cells. Marked immunohistochemical expression of CSPG5 reflected poor clinical outcome. Correlation between this protein expression and tumor size, age, tumor grade, biochemical factors such as p53, ER, MiB1 suggests that CSPG5 may participate in malignant tumor growth. The findings proposed Chondroitin Sulfate Proteoglycan-5 as an adjunct to existing morphological and biological markers in diagnosis and prognosis of phyllodes tumor.
7	Parallel Speaker: Zhang Biyan Title: Investigation of X-linked inhibitor of apoptosis associated factor 1 (XAF1) in breast cancer Supervisor: Associate Professor Yip Wai Cheong, George Department: Department of Anatomy, Faculty of Science Abstract: Breast cancer is the main cause of female mortality from cancer. Invasive ductal carcinoma (IDC) makes up 80% of breast cancer. Heparan sulphate 6-O sulphotransferases (HS6ST) are responsible for catalyzing 6-O sulphation on heparan sulphate chains, which have been proven to interact with many growth factors resulting in cancer progression. Previous work from HS6ST3 silencing revealed up-regulation of XAF1, a pro-apoptotic protein. This study aims to investigate the expression and role of XAF1 in IDC as well as its association with HS6ST3 at the protein level using immunohistochemistry on clinical IDC samples. Results revealed significant association between HS6ST3 and XAF1. In addition, low XAF1 is shown to correlate with pro-malignant phenotypes such as large tumour size (P=0.035) and high tumour staging (P=0.013). Low XAF1 is also predictive of shortened survival after recurrence with a hazard ratio of 2.364. Taken together, the results suggest that XAF1 plays a tumour-suppressive role in IDC and may serve as a potential prognostic marker for this disease.