DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Design of targeted photosensitizers for applications in detection, bioimaging and phototheranostics
No full textSquaraines are a class of polymethine dyes easily recognized by their four-membered ring core with two flanking electron-donating groups. Their intense absorption and emission profiles in the far-red to near-infrared (NIR) region make them highly attractive photosensitizers. Utilized broadly in applications like organic photovoltaic cells, bioimaging, chemosensors, and phototherapy, squaraines offer great flexibility and versatility in terms of their synthesis, photophysical properties and functional modifications. Phototheranostics is an increasingly popular trend whereby photosensitizers are employed for simultaneous fluorescent detection of specific therapeutic targets and treatment through light-induced generation of heat or reactive oxygen species (ROS). In the first chapter, key working principles of how squaraines can be utilized in the field of detection and phototheranostics will be introduced. Delving further, we explored the application of aggregation-induced fluorescence quenching in the second chapter by developing a thermoswitchable squaraine-based sensor for the rapid detection of Gram-negative bacteria. The third chapter focused on manipulating the organelle specificity of squaraine dyes through the use of an affinity ligand to achieve precise targeting and photodynamic treatment of cancer cells. In the fourth chapter, we further built on this concept and once again tuned the affinity of squaraines to target the cytoskeleton. Tubulin was successfully tagged and the microtubular network within the cancer cells was disrupted during photodynamic therapy. Finally, we provide an outlook on the bright prospects of squaraine dyes with possible future development to the work laid out in this dissertation. Ultimately, this research hopes to strengthen the squaraine toolkit for phototheranostic applications.Doctor of Philosoph
Machine learning-driven interfacial characterization and dielectric breakdown prediction in polymer nanocomposites
No full textThe development of polymer nanocomposites has emerged as a promising approach for achieving higher-density energy storage. However, challenges in directly characterizing the interface between the matrix and nanoparticles, a pivotal factor for performance enhancement, have led to a shortfall in effective modeling methods. In this work, we propose a novel interfacial modeling approach that quantitatively describes the continuous transition of dielectric properties across the interface, capturing the inhomogeneous nature observed experimentally. A finely tuned Polynomial Chaos Neural Network (PCNN) with a determination coefficient exceeding 0.999 is developed to elucidate the relationship between model parameters and nanocomposite permittivity. The finite element model employing the proposed interface model demonstrates improved accuracy in predicting the permittivity of various nanocomposite systems with a physical insight into the interface. Built upon the interface model, a developed phase field model is then incorporated to investigate the dielectric breakdown mechanism in nanocomposites, highlighting the interface's capacity to repel the breakdown path. 3D phase field simulations on electrical treeing successfully forecast the electrical tree structures in pure epoxy and nanocomposites with new insights into the dielectric breakdown. This research addresses a crucial need in the numerical modeling of nanocomposite interfaces and their role in dielectric breakdown analysis, providing a valuable tool for the design of next-generation dielectric materials with improved energy storage capabilities.Energy Market Authority (EMA)Nanyang Technological UniversityNational Research Foundation (NRF)Submitted/Accepted versionThis work was supported by SP Group, the National Research Foundation, Singapore, the Energy Market Authority, under its Energy Programme (EMA-EP010-SNJL-002) and Nanyang Technological University. W.H acknowledges the support of the National Natural Science Foundation of China (Grant No.12302146) and the China Postdoctoral Science Foundation (Grant No.2022T150086)
Mechanism decoding and de novo design of drug molecules with deep learning for targeted therapy
No full textMolecular design is a paramount mission in drug discovery process. Successful drug design strategies depend on two tasks: decoding of drug mechanism of action; and development of algorithm for targeted design of drug molecules, both of which remain obscure currently. Intuitively, decoding mechanism of drugs involves understanding the roles of both biological features from drug targets and chemical features from drug molecules in the interaction between drugs and their targets. Based on features extracted from potential biological targets, targeted drug design algorithm learns to generate molecule structures accordingly. Therefore, in this thesis, we divide our research on the two tasks into three chapters: (1) identification of salient multi-omics features in predicting drug responses; (2) investigation on salient chemical structures and their interaction; and (3) development of deep generative models for molecular design.
In Chapter 2, to elucidate the biological activities caused by drugs, we depict the biological profiles of cells with multi-omics data and leverage them to predict drug responses. Complex diseases, such as cancer and neurodegenerative disorder, are often caused by factors of various biological levels (e.g., gene mutation and protein misfolding), which renders it difficult to demystify the pathological reason of the disease progression. Multi-omics technology enables researchers to investigate the cause of such diseases from multiple levels. However, multi-omics data are intrinsically heterogeneous and high-dimensional. To overcome such difficulties, we propose an attention-based deep neural network to predict cancer drug responses. We first reduce the dimensionality of the multi-omics data with embedding layers, then leverage the attention mechanism to integrate the latent multi-omics features. By looking into the attention weights assigned by the model, we found that gene mutation and proteomics features gave rise to the prediction to a great extent.
In addition to integrating multi-omics features with the attention mechanism, we explore the network-based approach, which converts the high-dimensional data into patient similarity networks (PSN), on vital status prediction of neuroblastoma patients. PSNs constructed with different omics data are merged by the similarity network fusion algorithm, and topological features are then extracted from the PSN, including centrality features and modularity features. Centrality features reflect the node’s importance in the network, and the modularity features illustrate the membership of each node in the modules identified by clustering algorithm. We predict vital status of neuroblastoma patients with deep neural networks and explain the feature importance with integrated gradients. Our results show that modularity features generally contribute to the prediction more than centrality features.
In Chapter 3, we investigate the association between drug molecular structures and drug responses. We represent drugs with molecular graphs, which preserve more structural information than 1D descriptors such as simplified molecular-input line-entry system (SMILES) or molecular fingerprints, and utilize graph neural networks (GNN) to comprehensively model the molecule structures. We also incorporate gene expression and gene mutation profiles of cancer cell lines for drug response prediction and integrate the learned latent drug and cell line features with cross-attention mechanism. We further analyze the model’s interpretability with GNNExplainer and integrated gradients. Pathway analysis is conducted based on the gene saliency scores. Our model successfully captured the mechanism of action of the drugs and predicted the drug responses accurately.
In Chapter 4, we step into design of potential drug candidates with desired biological and chemical properties. The rapid development of deep generative models has brought forth an unprecedented opportunity of generating molecular structures from scratch, named \textit{de novo} design, rather than screening the vast chemistry library. Inspired by the generative model, we develop a constrained graph latent diffusion model to generate hit candidates under the guidance of transcriptomic profiles. We first train a variational autoencoder to model the latent space of drug-like molecules, and then train the diffusion model to manufacture the latent representations. This circumvents operating diffusion models directly on the topological structures of molecules, which is difficult and costly. In order to generate drug candidates with specific biological targets, we incorporate gene expression changes that are expected to be caused by the generated molecules as constraints. Our method exhibited outstanding performance in a series of constrained and unconstrained generation tasks.
In addition to drug design guided by biological targets, we explore pharmacophore-guided drug design, which induces the generation of molecules with characterized chemical structures that can form specific types of interactions with other biomolecules. Specifically, we devise a geometric diffusion bridge to align the distributions between 3D pharmacophore arrangements and molecular structures. Diffusion bridge extends the general diffusion model's ability of mapping the data distribution to the prior distribution into aligning arbitrary distributions. We equip the diffusion bridge with SE(3)-equivariant dynamics to transform between the point clouds representations of the molecule and the associated pharmacophores. The generated molecules exhibit high recovery rates of desired pharmacophores and high binding affinity with potential target proteins. This demonstrates our model's ability of transforming indispensable chemical structures into potential drug candidates.
In sum, in this thesis, we have explored decoding of drug mechanism via explainable deep learning, and generation of de novo drug designs with generative models. Our approaches tackle the difficulty of interpreting deep neural networks in pharmaceutical research and provide a feasible way of generating hit candidates with desired biological activities expressed by identified salient features. We believe our research will expedite the drug discovery process drastically.Doctor of Philosoph
Utilization of fingerprint dusting powder in tracking fingerprint aging using MALDI-TOF-MS
No full textTracking of fingerprint aging will provide valuable forensic value for crime investigation as trust in traditional fingerprint evidence has been declining over the years and a practical standard procedure has not been established. Chemical analysis of fingerprints by matrix-assisted laser desorption/ionization mass spectrometry has been utilized in previous research and has been shown to be capable in tracking fingerprint aging. However, the usage of conventional matrices in current protocol is undesirable as they disrupt the integrity of fingerprint samples. Hence, this project aims to validate the usage of fingerprint-dusting powders in substitution of conventional matrices in the forensic study of fingerprint aging by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS).
This study will focus on a fingerprint dusting powder known as Silver/Black a commercial product manufactured by Sirchie. Silver/Black was characterised and their performance by MALDI-TOF-MS in fingerprint aging was studied. Silver/Black was selected due to the presence of stearic acid in it that could function as a proton donor that may enhance ionization efficiency. Silver/Black was found to be able to detect biomarkers from fingerprints at a satisfactory quality in ambient/accelerated aging over 28 days. Sunlight simulation was also conducted that stimulated sunlight exposure of 28 and 92 days respectively. Silver/Black was able to produce satisfactory results in ambient aging with peak m/z 297.5 showing consistent decreasing trend when compared from Fresh to Day 28 with statistical significance. Silver/Black also produced promising results in sunlight simulation aging with peak m/z 341.5 showing statistical significance and a consistent increasing trend. Therefore, Silver/Black can be utilized as a matrix in future experiments in the tracking of fingerprint aging using MALDI-TOF-MS.
Future work will focus on the optimization of fingerprint dusting powders, verification of current findings by utilising more robust techniques and optimizing experiment conditions as well as sample preparationBachelor's degre
MemPal: leveraging multimodal AI and LLMs for voice-activated object retrieval in homes of older adults
No full textOlder adults have increasing difficulty with retrospective memory, hindering their abilities to perform daily activities and posing stress on caregivers to ensure their wellbeing. Recent developments in Artificial Intelligence (AI) and large context-aware multimodal models offer an opportunity to create memory support systems that assist older adults with common issues like object finding. This paper discusses the development of an AI-based, wearable memory assistant, MemPal, that helps older adults with a common problem, finding lost objects at home, and presents results from tests of the system in older adults' own homes. Using visual context from a wearable camera, the multimodal LLM system creates a real-time automated text diary of the person's activities for memory support purposes, offering object retrieval assistance using a voice-based interface. The system is designed to support additional use cases like context-based proactive safety reminders and recall of past actions. We report on a quantitative and qualitative study with N=15 older adults within their own homes that showed improved performance of object finding with audio-based assistance compared to no aid and positive overall user perceptions on the designed system. We discuss further applications of MemPal's design as a multi-purpose memory aid and future design guidelines to adapt memory assistants to older adults' unique needs.Published versio
文图学视角下的《仙剑奇侠传》IP研究——以《仙剑三》和《祈今朝》为例 = A study of “Chinese Paladin” IP from the perspective of text and image studies: taking “Chinese Paladin 3” and “Sword and Fairy 6” as examples
No full text仙侠剧源于《仙剑奇侠传》。自该剧于2005年播出后，市面上便掀起了仙侠热潮。尽管仙侠剧层出不穷，但获得高评价的作品却屈指可数，而《仙剑奇侠传三》（以下简称《仙三》）正是其中之一。它不仅获得了观众的认可，还连带着把剧里的主演都推向了更高的知名度。
由于《仙剑奇侠传》是仙侠剧的开始，与之比较，有失公允，因此，本文将以文图学的视角分析其他《仙剑奇侠传》IP电视剧——2009年播出的《仙三》以及2024年播出的《祈今朝》（又名《仙剑奇侠传六》），探讨前者高口碑的原因，是否与游戏相关，以及为何后者未能延续前者的成绩。
研究发现《仙三》电视剧的成功与游戏关联度较低，它的高口碑得益于画面暗示、CP感的塑造以及音乐的安排。《祈今朝》在画面、叙事节奏、剧情上，都存在一定的缺陷，因此未能达到《仙三》的成就。
关键词：文图学 《仙剑奇侠传三》《祈今朝》 仙侠电视剧 电子游戏
Xianxia drama originated from Chinese Paladin. After its release in 2005, many Xianxia dramas entered the market. Although numerous Xianxia dramas exist, only some received widespread recognition with Chinese Paladin 3 being one of them. Not only was it well received by audiences, but it also significantly increased the popularity of its lead actors.
Since Chinese Paladin is the origin, a direct comparison may be unfair. Hence, this paper will analyse other dramas within the Chinese Paladin IP through the lens of Text and Image Studies: Chinese Paladin 3, broadcasted in 2009 and Sword and Fairy 6, broadcasted in 2024. It will explore the reasons behind the former’s strong reputation, its connection to the game and why the latter failed to achieve the same result.
Research reveals that Chinese Paladin 3’s success is less related to the game, and its high reputation is due to its implicit visual storytelling, engaging character pairings and the arrangement of music. Conversely, Sword and Fairy 6 was hampered by visual flaws, narrative pacing and weak plot, preventing it from replicating the same success.
Keywords：Text and Image Studies, Chinese Paladin 3, Sword and Fairy 6, Xianxia Drama, Video GameBachelor's degre
Shake the habit
No full textShake The Habit is a communication campaign aimed at promoting healthier eating habits among young adults, with a specific focus on keeping sodium intake in check to prevent hypertension and other cardiovascular risks in the long run. Recognising the challenges young adults face in making healthier choices at a stage where lifelong habits take shape, the campaign targets university students aged 18 to 25 in Singapore – a demographic transitioning into independent nutrition and health management.
The primary objectives of the campaign were to increase awareness of the dangers of excessive salt consumption, encourage healthier eating habits, and provide practical tools for young adults to make informed dietary choices. Running from October 2024 to February 2025, Shake The Habit adopts a more interactive and relatable strategy, leveraging a mix of digital content, campus activations, and rewarding partnerships, to reinforce learning and building of positive dietary habits. The campaign’s impact was evaluated through a post-campaign survey and results of offline and digital outreach exceeded the campaign’s target objectives, demonstrating significant growth and engagement.Bachelor's degre
Chatbot companion for elderly
No full textThe rapid growth of the aging population has underscored the need for innovative solutions to
address challenges related to social isolation, mental health, and accessibility among the elderly.
There are around 23% of elderly aged 60 and above reporting to experience loneliness in
Singapore.
Therefore, this project aims to explore the design and development of a chatbot companion for
elderly users, focusing on fostering companionship, promoting mental well-being, and providing
practical assistance. The chatbot integrates natural language processing (NLP) and advanced
machine learning techniques to deliver intuitive and empathetic interactions as well as initiating
conversations with the elderly.
Key features include daily reminders for medication and appointments, conversational
engagement on various topics, lottery result reporting, weather forecast, latest news and so on.
By utilising DeepSpeech, a speech-to-text engine and 01.AI’s Yi-1.5 Chat LLM, a chatbot that is
capable of providing companionship to the elderly will be developed.Bachelor's degre
Highly integrated hybrid inductive and capacitive power transfer system with asymmetrical printed-circuit-board-based self-resonator
No full textThis paper presents a highly integrated and compact hybrid wireless power transfer (WPT) system with asymmetrical printed-circuit-board (PCB) based self-resonators. The PCB-based self-resonant coupler consists of four PCB-coil plates with two different sizes, which can work as the transmitter/receiver for IPT as well as the capacitive plates for CPT. With a typical stacked four-plate configuration, both inductive and capacitive mutual couplings are achieved between transmitter and receiver, contributing to a highly compact and integrated self-resonant hybrid WPT system without any external compensation components. Detailed theoretical analysis and system modeling are provided based on the two-port parameter theory and a 300W hybrid WPT prototype is implemented with an asymmetrical coupler consisting of 210-mm and 140-mm PCB-coil plates. The implemented hybrid WPT system is tested at 80mm, 60mm, 37mm, and 12mm with self-resonant working frequencies of 3.845MHz, 3.75MHz, 3.57MHz, and 3.19MHz, respectively, and the system performance in terms of output current property, power transfer capability, DC-DC efficiency, and misalignment tolerance are evaluated in details, which demonstrate a peak DC-DC efficiency of 87.3% with 155.7W at 12mm and 86.7% with 237.5W at 37mm, validating the effectiveness of the designed hybrid WPT system.Agency for Science, Technology and Research (A*STAR)Ministry of Education (MOE)Submitted/Accepted versionThis work was supported in part by A∗Star MTC Young Individual Research Grant under Grant M23M7c0115 and in part by the Ministry of Education Academic Research Fund Tier-1 under Grant RG134/23
Spectral super-resolution for real-world applications
No full textSpectral super-resolution (SSR) aims to recover hyperspectral (HS) information of the scene from RGB images, which is important for the downstream tasks in various fields, including medical imaging, remote sensing and food science. However, existing methods failed to consider the physical imaging process and ignored the relationship between the spectral and spatial dimensions. In addition, the practical application of SSR in real-world settings remains challenging and insufficiently explored. To this end, this dissertation proposed a multidimensional spectral fusion network (MSFNet). In particular, this framework proposed an adaptive spectral fusion module (ASFM) to simulate the reverse process of the transformation from HSI to RGB images. By utilizing the attention mechanism, adaptive weights are assigned to different wavelength ranges. Moreover, a multidimensional dependency attention module (MDAM) is proposed to capture the relationship between the spectral and spatial dimensions. The experiments show that the proposed MSFNet achieves superior performance compared to other state-of-the-art methods on both the ARAD-1K dataset and a real-world dataset.Master's degre