DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Additive effects of natural plant extracts/essential oils and probiotics as an antipathogenic topical skin patch solution for acne and eczema
No full textThis work leverages the additive antipathogenic effects of natural extracts/essential oils (EOs) and probiotics for the treatment of acne vulgaris associated with Cutibacterium acnes (C. acnes) and eczema complicated by secondary infections with Staphylococcus aureus (S. aureus). Six probiotic strains and various extracts/EOs were evaluated in a large screening to evaluate their potential against both pathogens. Lacticaseibacillus paracasei PCB003 was able to inhibit the growth of both pathogens. For extracts/EOs, Oregano EO had the best antipathogenic effects on both pathogens and did not show any adverse impact on the growth of probiotics, making it suitable for simultaneous use. Using Lactiplantibacillus plantarum PCB011 as a probiotic model, five material formulations were assessed for their suitability to protect probiotic cells within freeze-dried topical patches. Alginate and trehalose (ALG+TRE) and thermoplastic starch (TPS) had the highest probiotic survivability, with ALG+TRE chosen as the final patch material as it was more robust. PCB003 and PCB011 were individually incorporated into the ALG+TRE freeze-dried matrix to form a 6 mm patch; both ALG+TRE (PCB003) and ALG+TRE (PCB011) patches, when used individually, successfully inhibited C. acnes growth by 4.7 and 6.0 mm, respectively, surpassing the performance of commercially available acne patches. The additive effect with 30% Oregano EO further improved pathogen inhibition. For S. aureus, the incorporation of 30% Oregano EO to ALG+TRE (PCB003) increased the size of the inhibition zone more than 10-fold. For C. acnes, the ALG+TRE (PCB003) patch with 30% Oregano EO demonstrated an inhibition zone of 16.3 mm, and the ALG+TRE (PCB011) patch with 30% Oregano EO achieved a 14.3 mm inhibition zone. Genomic analysis confirmed that PCB003 and PCB011 lack antimicrobial resistance determinants, ensuring safety. This study successfully combined probiotics and natural agents to create effective dermatological antipathogenic patches.Ministry of Education (MOE)Singapore Food AgencySubmitted/Accepted versionThis work was supported by the Ministry of Education (RG79/22, RG29/24, MOE-T2EP30223-0001, NGF-2023-14-005) and the Singapore Food Agency (SFS_RND_SUFP_001_06)
Impact of heat pumps and future energy prices on regional inequalities
No full textThe adoption of heat pumps to displace the use of gas for domestic heating is a major component of the strategy to reduce emissions in the UK. This study examines the impact of adopting heat pumps on regional inequalities in the UK. An index is used to assess how variations in household fuel costs could affect regional disparities across different future price scenarios. The findings reveal that, at 2019 prices, most households would face higher heating costs with heat pumps. However, following the 2022 energy price shock, heat pump adoption would lead to lower heating costs for most households compared to gas heating. The effect is sensitive to the electricity-to-gas price ratio, with regions experiencing high fuel poverty being most vulnerable to negative impacts. By mapping these geospatial effects, the study enables the forecasting of future inequality trends, providing insights for informed policy development. The results suggest that, under appropriate price structures, heat pump adoption could contribute to both decarbonisation and reduced social inequality. An example mechanism for financial support to mitigate the impact of adopting heat pumps on inequality is demonstrated. This study highlights the novel capability of The World Avatar (TWA) approach to integrate cross-domain data sets, combining energy policy with social equity goals. By forecasting future inequality trends based on energy price scenarios, the study provides a route to valuable insights to support informed policy development, highlighting how the adoption of heat pumps can influence regional inequalities and emphasising the need for targeted interventions to support vulnerable regions.National Research Foundation (NRF)Published versionThis research was supported by the National Research Foundation, Prime Minister’s Office, Singapore under its Campus for Research Excellence and Technological Enterprise (CREATE) programme. Part of this work was also supported by Towards Turing 2.0 under the EPSRC Grant, United Kingdom EP/W037211/1. The authors would further like to thank and acknowledge the financial support provided by the Cambridge Trust. Markus Kraft gratefully acknowledges the support of the Alexander von Humboldt Foundation, Germany
Deep learning methods for predicting binding affinity of antibody-antigen: a comparative study of embedding techniques and model architectures
No full textAntibody-antigen interactions are critical to immune responses and play a pivotal role
in therapeutic development, diagnostics, and vaccine design. Accurate prediction of
binding affinity between antibodies and antigens is essential for accelerating
antibody discovery and advancing our understanding of immune system dynamics.
This study explores deep learning methods for predicting antibody-antigen binding
affinity, focusing on employing various deep learning architectures and evaluating
different antibody-antigen representation techniques by leveraging transfer learning
with large language models. A comprehensive comparative analysis is conducted to
assess the performance of various embedding techniques and deep learning
architectures, including transformer-based models, in modelling the binding affinity
between antibody-antigen pairs. The research aims to establish fundamental
benchmarks through unified datasets and standardized evaluation metrics to provide
a solid foundation for future research in this area. Results from the proposed models
demonstrate superior performance, surpassing existing state-of-the-art approaches
in binding affinity prediction. By offering valuable insights into the effective
application of deep learning to this complex task, this work lays the groundwork for
future innovations in immunology, therapeutic design, and vaccine development.Bachelor's degre
Robust loop closure detection and relative localization for multi-robot SLAM
No full textMulti-Robot Simultaneous Localization and Mapping (MR-SLAM) has gained increasing
attention due to its potential in enhancing efficiency, coverage, and robustness in autonomous
navigation and large-scale mapping. Unlike single-robot SLAM, MR-SLAM introduces additional
challenges such as inter-robot data synchronization, cross-robot loop closure detection, and
maintaining a globally consistent map. Among these challenges, loop closure detection plays a
crucial role in reducing cumulative drift and ensuring map accuracy. However, existing loop closure
detection methods struggle with issues such as sensor noise, viewpoint variations, repetitive
environments, and dynamic objects, leading to false positives or missed loop closures.
This dissertation explores loop closure detection techniques in MR-SLAM, focusing on
LiDAR-based, visual-based, and hybrid SLAM approaches. The study provides a comparative
analysis of state-of-the-art loop closure detection algorithms, including RING++, DiSCO, and Scan-
Context, evaluating their robustness, computational efficiency, and accuracy in multi-robot
environments. Additionally, LiDAR-based SLAM methods (A-LOAM, Hector-SLAM, Gmapping)
and visual-based SLAM techniques (ORB-SLAM3, VINS-Fusion, Bag-of-Words) are assessed to
understand their effectiveness in handling real-world challenges.
The experimental results indicate that RING++ outperforms other loop closure methods in
large-scale and feature-rich environments, demonstrating superior rotational and translational
invariance. DiSCO exhibits high accuracy in structured outdoor settings but is computationally
intensive. ORB-SLAM3 and VINS-Fusion, leveraging multi-sensor fusion, achieve better
robustness in visual SLAM by integrating camera, IMU, and LiDAR data. However, traditional
methods still face difficulties in dynamic and highly repetitive environments, where false loop
closures often occur.
To address these challenges, this dissertation discusses potential hybrid approaches that
combine deep learning-based place recognition, semantic SLAM, and adaptive loop closure
validation for improved reliability. Furthermore, it explores the feasibility of decentralized SLAM
architectures, where multiple robots autonomously detect and validate loop closures without relying
on a central processing unit.
This research contributes to the advancement of multi-robot SLAM technology by providing
insights into effective loop closure detection strategies, highlighting the trade-offs between accuracy,
efficiency, and real-time performance. The findings have significant implications for autonomous
navigation, search and rescue operations, smart city infrastructure, and large-scale robotic
deployments, where robust multi-robot SLAM systems are essential for real-world applications.
Future research directions include deep learning-enhanced place recognition, GPU-accelerated loop
closure detection, and decentralized MR-SLAM frameworks to further improve performance in
complex environments.Master's degre
Click with caution: investigating the antecedents of protective behaviour towards online scams among college students
No full textYoung adults, particularly college students, are falling victim to online scams more frequently, which exploit their digital habits and financial motivations. This study examines scam vulnerability and protection behaviours among college students in Singapore using a mixed-methods approach. Through findings from six focus group discussions, e-commerce, job, and phishing scams were identified as the most prevalent among college students, often leveraging social engineering tactics such as social proof, urgency, and impersonation. The findings also suggest key factors that influence college students’ susceptibility to scams, including identified financial pressure, impulsive online behaviour, limited scam knowledge, overconfidence, trust in social circles, and emotional reactivity. Based on the qualitative findings, we developed a survey to quantitatively examine factors influencing college students' intention to protect themselves against online scams. Guided by Protection Motivation Theory, the survey findings show that threat severity, response efficacy, and self-efficacy significantly predict protection intentions, while perceived susceptibility negatively impacts them. The relationship between both online scam knowledge and fear with behavioural intention is partially mediated by threat appraisal. The study serves as a guide to understand both exploratory and theoretical insights of the target demographic, along with practical implications to inform the development of future interventions regarding scam protection.Bachelor's degre
Reviving the Lion City's roar: an empirical analysis of Singapore's stock market underperformance and the role of regulatory reforms on stock market success
No full textSingapore, a global financial hub, faces a declining stock market (SGX) with de-listings consistently outnumbering new listings. This is deemed to be a critical national issue, as the Singapore government established a task force to explore ‘bold regulatory changes’ to revive the stock market. This study analyses SGX’s performance across 71 countries (markets) using key stock market success measures: 1) market capitalisation (% of GDP), 2) turnover ratio, and 3) number of listings (normalised by population) from 2007 to 2022. We conducted a cross-sectional analysis and longitudinal analysis by constructing a composite Stock Market Performance Index (SMPI) via Principal Component Analysis (PCA) to rank stock markets’ performances. Findings show that while SGX performs relatively well in market capitalisation, it underperforms in turnover ratio and number of listings. The study is then extended to analyse the effect of specific regulatory measures on the 3 success measures from 2007-2022. We do so by conducting a random effect panel regression with Driscoll-Kraay standard errors (DKSE) across 31 markets, examining the effects of 1) short-selling freedom, 2) pre-IPO profitability, 3) national price-to-earnings ratio, 4) the existence of an independent national securities and exchange commission, and 5) stock market connectivity. Results show that short-selling freedom significantly affects market capitalisation and turnover ratio, pre-IPO profitability affects market capitalisation, national price-to-earnings ratio affects all 3 measures, and the existence of an independent national securities and exchange commission affects market capitalisation. Surprisingly, the stock connect programs show no significance for all measures in our results. These insights inform policy recommendations for revitalising SGX’s market vibrancy and listing attractiveness.Bachelor's degre
Gut microbes modulate the effects of the flavonoid quercetin on atherosclerosis
No full textGut bacterial metabolism of dietary flavonoids results in the production of a variety of phenolic acids, whose contributions to health remain poorly understood. Here, we show that supplementation with the commonly consumed flavonoid quercetin impacted gut microbiome composition and resulted in a significant reduction in atherosclerosis burden in conventionally raised (ConvR) Apolipoprotein E (ApoE) knockout (KO) mice but not in germ-free (GF) ApoE KO mice. Metabolomic analysis revealed that consumption of quercetin significantly increased plasma levels of benzoylglutamic acid, 3,4 dihydroxybenzoic acid (3,4-DHBA) and its sulfate-conjugated form in ConvR mice, but not in GF mice supplemented with the flavonoid. Levels of these metabolites were negatively associated with atherosclerosis burden. Furthermore, we show that 3,4-DHBA prevented lipopolysaccharide (LPS)-induced decrease in transendothelial electrical resistance (TEER). These results suggest that the effects of quercetin on atherosclerosis are influenced by gut microbes and are potentially mediated by bacterial metabolites derived from the flavonoid.Ministry of Education (MOE)Ministry of Health (MOH)Nanyang Technological UniversityNational Medical Research Council (NMRC)Published versionThis work was partly supported by grants from NIH HL144651 (FER), and NIH HL148577 (FER). This work was also supported by a grant from a Transatlantic Networks of Excellence Award from Foundation Leducq (17CVD01; to F.B. and F.E.R.). K.K. was supported by the Ministry of Education (Singapore) under its Academic Research Fund Tier 1 (RS10/22), the Singapore Ministry of Health's National Medical Research Council under its CS-IRG-NIG (MOH-001379), and Vascular Research Initiative, LKC Medicine, Nanyang Technological University. F.B. is the Torsten Söderberg Professor in Medicine and a Wallenberg Scholar. T.-W.L.C. was supported by the National Institutes of Health, under Ruth L. Kirschstein National Research Service Award T32 HL007936 from the National Heart Lung and Blood Institute to the University of Wisconsin-Madison Cardiovascular Research Center
CheckGPT
No full textRecent advancements have led to the widespread adoption of generative artificial intelligence (GAI) chatbots such as ChatGPT, Gemini, and Perplexity. Despite their articulate and coherent responses, several concerns about using GAI chatbots have been raised. In particular, the issues surrounding inaccuracies produced in their responses. Students — key users of GAI chatbots — are typically vulnerable to accepting inaccurate chatbot responses. However, both global and local initiatives have largely overlooked this issue.
In response, CheckGPT was developed as a communications campaign to equip students with practical strategies to avoid inaccurate GAI chatbot responses. The campaign ran from 3 November 2024 to 8 March 2025, and targeted students in post-secondary educational institutions (PSEIs) and secondary schools. Central to the campaign was the self-developed 4Cs framework, which guides students in crafting better prompts and verifying AI-generated content.
Key efforts included a custom GPT, school roadshows, and a GAI chatbot guidebook. The campaign reached over 12,175 students and attained media coverage amounting to S$231,480. The custom GPT achieved an adoption rate of approximately one in four students reached, reflecting its relevance and functionality. Beyond student engagement, the 4Cs framework was recognised by various educational institutions, and was integrated into an institution-wide GAI literacy course. By addressing a persistent gap in GAI literacy education, CheckGPT leaves behind lasting, accessible resources to support students in navigating the challenges presented by GAI chatbots.Bachelor's degre
Towards a scalable and robust federated learning system
No full textThis project enhances a prior peer-to-peer federated learning (P2PFL) prototype by addressing key challenges in scalability, transparency, and robustness.
On the development side, the native Android application is extended with an improved asset management system for handling models and datasets, a scalable HTTP-based model transfer protocol, and a centralised web-based registry to facilitate model discovery. To improve traceability and trust, a graph-based model lineage visualisation system is implemented using Neo4j.
From a research standpoint, the project examines three types of model poisoning attacks—random label, flip label, and backdoor—to determine the minimum effective poisoning threshold. These attacks are evaluated against three defence strategies: performance-based (RONI), weight-based (Dim-Krum), and representation-based (activation clustering). Experimental findings reveal that no single method offers complete protection, but a hybrid strategy combining performance and representation-based defences provides the most resilient safeguard. Collectively, these contributions advance the scalability and security of decentralised federated learning systems in real-world applications.Bachelor's degre
DINK lifestyle in pro-natalist Singapore
No full textThe dual-income, no-kids (DINK) lifestyle has gained prominence among young, highly
educated Chinese heterosexual couples in Singapore. Using a qualitative research approach, in-depth
interviews with six couples were conducted to examine their motivations, lived experiences, and
societal perceptions within the context of Singapore’s persistent pro-natalist policies. Findings reveal
that financial pressures, career aspirations, and lifestyle priorities collectively outweigh state-driven
incentives for childbearing. Pro-natalist policies fail to address deep-seated structural barriers, such as
high living expenses and stressful work culture, which deter young, highly educated Chinese
heterosexual couples from having children. Interestingly, these couples experience minimal stigma,
reflecting a “deviant departure”, in which the DINK lifestyle is increasingly normalised rather than
perceived as deviant. This study highlights the misalignment between state policies and evolving
lifestyle aspirations, contributing to broader discussions on fertility trends and family planning in
Singapore.Bachelor's degre