Theme 1: Understanding Diseases and Disease Prevention
Project Title: Neuroimmune Mechanisms and Modulation in Alzheimer’s Disease
Project Coordinator: Prof Guojun Bu (HKUST)

Abstract

Alzheimer’s disease (AD), the most common form of age-related dementia, presents an urgent and growing health challenge in Hong Kong. By 2049, one-third of the city’s population will be aged 65 or older, positioning Hong Kong to become the world’s oldest society. Meanwhile, Hong Kong already holds the global record for the highest percentage of centenarians, surpassing Japan. Despite recent scientific advances, there is still no cure or effective way to prevent AD, highlighting the critical need for new treatment breakthroughs. Increasing evidence suggests the immune system is a key player in how AD develops and worsens, though many questions remain about whether immune responses ultimately help or harm the brain over time. To address these questions, a team of leading scientists from Hong Kong’s top universities has launched a cutting-edge research project to investigate the dynamics of immune cells in the brain and throughout the body in AD, with a focus on two major genetic risk factors, APOE and TREM2. Using advanced mouse models and human stem-cell models, as well as patient sample analysis, the team aims to uncover new ways to harness the immune system to protect the brain. Their discoveries will be translated into next-generation therapeutic strategies that could prevent or slow the progression of AD. As we face a rapidly aging society, this project will have a significant impact on addressing one of the most pressing health issues of our time. In addition, this work will position Hong Kong as one of the global leaders in AD research, neuroimmunology, and medical innovation, while contributing to the worldwide fight against this devastating disease.

Theme 1: Understanding Diseases and Disease Prevention
Project Title: Sleep and Circadian Rhythm: Potential Window for Prevention of Mental Disorders in Adolescents
Project Coordinator: Prof Yun Kwok Wing (CUHK)

Abstract

Mental health problems are imposing a significant burden on individuals, families and society. Alarmingly, research indicates that 75% of mental disorders emerge by age 24, underscoring the importance of early detection and intervention. Recent local epidemiological data with approximately 6000 children and adolescents indicates that 5.4% the adolescents experience depression and 6.1% of suffer from anxiety. However, many adolescents remain undiagnosed, highlighting an urgent need for early identification of at-risk adolescents to prevent the progression to more severe mental health conditions.

Most existing prevention programmes adopt a one-size-fits-all approach, which limits their scalability and effectiveness - potentially due to the heterogeneity of patients and illness characteristics. Moreover, the rate of help seeking behaviours among adolescents is low and delay in seeking help is common. To address this, innovative methods are needed to identify features and subtypes at their early stages, thereby guiding the development of effective and personalised prevention programme. In particular, sleep and circadian disturbances, especially insomnia and delayed sleep phase problem, coexist, predate and predict with a 2-3 times increased risk of depression and anxiety. Accurate identification and a comprehensive understanding of the complex relationship between sleep/circadian disturbances and mood disturbances, facilitated by multi-modal digital-biological measurement, will provide important opportunities to prevent the onset of mental disorders.

In this project, we aim to uncover the complex relationship between sleep and circadian disturbances and mental health in adolescents through a two-phase study. Phase I will leverage cutting-edge AI-digital-wearable technology and advanced AI analytics to develop precise models for predicting depression and anxiety from sleep and circadian perspectives. This deep phenotyping will enhance understanding of the underlying mechanism linking sleep and circadian disruptions to mental health outcomes. Phase II will implement a personalised, digitally delivered intervention focused on improving sleep and circadian health to prevent the development of mood disorders. This transdiagnostic approach will be further adapted to tailor personalised features. In addition, the study will explore the biological changes and cost-effectiveness analysis related to digital interventions.

The study could have great potential to revolutionise mental health prevention strategies, offering critical insights for future healthcare policies and implementation. By prioritising sleep and circadian rhythms during this vital developmental stage of adolescents, we aim to create personalised AI-driven approach that will have the potential to fundamentally transform the delivery and management of mental health care and preventive medicine.

Theme 2: Developing a Sustainable Environment
Project Title: Identification and Exploration of Effective and Eco-friendly Disinfectants from Halophenolic Dbps/Compounds
Project Coordinator:Prof Xiangru Zhang (HKUST)

Abstract

The increasing prevalence of diseases caused by pathogenic microorganisms has driven up the use of chemical disinfectants in personal care products (such as hand sanitisers, detergents, soaps, cosmetics, shampoos and lotions), in industrial products (such as adhesives, paints, lubricants, textiles, pulp, bedding, ink, and medical scrubs), and in sanitation of surgical instruments, hospitals, households, shopping malls, restaurants and streets. The COVID-19 pandemic has further raised public awareness of personal and environmental disinfection, boosting global sales of disinfectants by over 50%. However, as chemical disinfectants generally end up in natural aquatic environments through wastewater effluent and urban runoff, extensive use of these chemicals can trigger secondary disasters in aquatic ecosystems. A typical example is chloroxylenol, a world-widely used disinfectant, which is found in 17-56% of antiseptic detergents, household disinfectants and hand sanitisers in the US, UK, China (mainland), and Hong Kong. The widespread use and high chemical stability of chloroxylenol can cause ecological damage to water bodies. There is thus an urgent need for Novel Effective Ecofriendly Disinfectants (NEED).

Inspired by the findings that chloroxylenol has a chemical structure similar to numerous halophenolic disinfection byproducts (DBPs) we have previously identified and that some DBPs can be readily photodegraded and detoxified in aquatic environments, we propose to identify eco-friendly disinfectants from halophenolic DBPs/compounds to address disinfectant-associated ecological risks. In this project, we will 1) systematically study the disinfection efficacy (against typical bacteria, fungi and viruses) and environmental fate/impacts of pre-screened halophenolic compounds by benchmarking against chloroxylenol, 2) provide new mechanistic insights into the pathogen inactivation and environmental degradability of halophenolic compounds, and 3) construct an AI-based predictive model for the disinfection efficiency and degradability of halophenolic compounds, and apply the model to further develop optimal disinfectants. This project has gathered the world’s top scientists in disinfection, analytical chemistry, synthetic chemistry, microbiology, ecology and engineering to develop ground-breaking novel effective eco-friendly disinfectants, with the goal of providing science-based solutions to minimise environmental burdens while promoting public health. The outcomes of this project will significantly benefit academia, the public, the government and the industry and will contribute to sustainable development in Hong Kong and globally.

Theme 2: Developing a Sustainable Environment
Project Title: Chemical Weather Observations of Carbon Chemistry in Greater Bay Area
Project Coordinator: Prof Jianzhen Yu (HKUST)

Abstract

Physical weather data are widely available and have long been used to inform our daily life decisions. Unlike physical weather parameters, chemical weather components are numerous and of extreme diversity, making them much more challenging to monitor. Chemical weather data that has been routinely monitored are mainly limited to the regulated pollutants, such as carbon monoxide, ozone, particulate matter (PM) mass, etc. However, monitoring for organic molecules are still largely deployed on an intensive field campaign basis. Organic materials, which make up a substantial fraction of the atmospheric chemical composition, are “fuels” for atmospheric oxidation chemistry. Their abundance and composition are intricately embedded in the chemistry that produces secondary pollutants such as ozone and a significant fraction of PM. The air quality and climate change issues facing the Greater Bay Area (GBA) are constantly evolving as a result of shifts in energy supplies for electricity production and transportation, as well as regional and national pollution control policies. In addition to monitoring the few criteria pollutants, it is critically important that we have a regional network of observing chemical weather data for tracking and understanding carbon chemistry in the ambient atmosphere to address the persistent secondary pollution and episodic pollution events in the GBA. In this project, we will expand and build upon existing measurement capacities in GBA to establish and demonstrate unified measurement platforms for key chemical weather components of carbon chemistry, including carbonaceous molecules in both gaseous and PM phases. The measurements are designed to enable the study of air toxics, episodic ozone and PM events as well as their long-term trends in the GBA.

The unified measurement capacities to be established through collaborative efforts will represent a significant step-change from the currently available air quality monitoring network. The new chemical weather data sets, continuous and long-term in temporal coverage and covering multiple sites, including Hong Kong, Macao, Guangzhou, and Shenzhen, will enable a wide range of exciting and impactful research opportunities, such as advancing fundamental process-level understanding of ozone and PM in response to changes in energy and land use, constraining uncertainties in air quality and regional climate change modeling, and unravelling specific organic constituents and properties responsible for health impacts. The project will include monitoring optimisation, data processing and evaluation, pollution process analysis, and modeling. The ultimate goal is to achieve more cost-effective protection of the public health and environment of the GBA.

Theme 4: Advancing Emerging Research and Innovations Important to Hong Kong
Project Title: Collaborative Generative AI (Co-GenAI)
Project Coordinator: Prof Hongxia Yang (PolyU)

Abstract

Generative Artificial Intelligence (GenAI) has recently attracted significant attention, driven by the rapid emergence of advanced tools and applications. GenAI encompasses a broad spectrum of technologies, including Large Language Models (LLMs), Multimodal Large Language Models (MLLMs), and diffusion-based models. As technologies continue to evolve, GenAI is poised for transformative growth, fundamentally reshaping both industry and academia.

The current monopoly on GPU resources significantly hinders the development of GenAI. At present, very few AI researchers/end-users have the opportunity to participate in the pretraining stages of GenAI, which are crucial for determining a model’s capabilities for generating their own GenAI models. To overcome this challenge, we are developing a collaborative GenAI system called, Co-GenAI, evolving through the integration of several hundred domain-specific models to create a foundation model aimed at achieving Artificial General Intelligence (AGI) with far less centralised computational demand. More specifically, we will carry out the following tasks: 1) Develop the Domain-Adaptive Continual Pretraining (DACP) infrastructure to enhance GenAI models by continually pretraining on domain-specific, unlabeled data, effectively adapting to target domain distributions; 2) Design a robust and generalisable ranking methodology that combines industry-standard evaluations, more secured in-house domain benchmarks, and precise loss-based scores from the foundation model to derive robust final rankings; 3) Implement advanced ‘model-over-model’ methodology for merging heterogeneous top-ranked domain-specific models and leveraging high-quality inputs and scalable ranking algorithms to support diverse applications so that GenAI models can be collaborative as a whole; 4) Demonstrate the project’s effectiveness through launching the Co-GenAI platform tailored to enhance diverse fields and collaboration, ultimately creating a versatile platform for next-generation AI applications. Through this innovative approach we aim to democratise AI development, making it more accessible and less dependent on massive centralised computational resources, thereby fostering greater innovation and diversity in the field. To evaluate Co-GenAI, we will implement and deploy the system across a wide range of applications in collaboration with our industry partners, including Cyberport, Hong Kong Science and Technology Park, Alibaba, and leading hospitals such as Huashan Hospital, Shandong Cancer Hospital and Institute, and Sun Yat-sen University Cancer Center.

The uniqueness and novelty of this project lie in its strategic identification of the most critical scientific challenges and its systematic approach to develop the promising Co-GenAI paradigm, including the easy-to-adapt training infrastructure, novel methodologies, and a real-world GenAI ecosystem. Timely and forward-looking, the project holds strong potential to position Hong Kong as a leader in next-generation GenAI paradigm. Backed by a team of world-renowned researchers with extensive expertise, we are confident in the project’s success and far-reaching impact.

Theme 4: Advancing Emerging Research and Innovations Important to Hong Kong
Project Title: Next Generation EDA
Project Coordinator: Prof Evangeline F.Y. Young (CUHK)

Abstract

The VLSI industry has advanced rapidly, moving beyond 5nm and pushing into 3nm technology. Designing such complex chips with billions of transistors and wires relies heavily on powerful electronic design automation (EDA) tools. As circuits grow more complex, there is an urgent need for fast, smart and scalable EDA solutions powered by AI and modern hardware. Modern large-scale heterogeneous integration ICs, such as 3D/2.5D ICs, chiplets, and advanced packaging also demand new EDA approaches. Our objective is to explore and develop Large Circuit Models, AI-native EDA, GPU/CPU-Accelerated EDA, and their applications on heterogeneous integrated ICs.