Project Title 1: An Integrated Graph Convolutional Network Multimodal Platform (GCN-MP) for Early Detection and Prediction of Late Onset Alzheimer’s Disease
Funding Source: US National Academy of Medicine (NAM) Healthy Longevity Catalyst Award, 2023
Funding: HKD 0.479 M
Led by: Prof. Victor OK Li, Principal Investigator and Dr. Jacqueline CK Lam, Co-Lead (HKU)

Abstract
Alzheimer’s disease (AD) is a leading cause of death worldwide. Early prediction of Late Onset Alzheimer’s disease (LOAD) is critical to timely intervention before irreversible brain damage. Accurate early LOAD prediction is challenging due to the need for a wide range of multimodal assessments, which can be time-consuming, costly, and invasive, preventing those who might develop symptomatic LOAD from timely diagnosis and therapeutic interventions. An AI-driven-multimodal LOAD prediction based on readily available data as inputs to an AI model trained on heterogenous data of high-dimensional modalities/features across different diseases/populations, can be desirable for early LOAD prediction/intervention. Our transformative multimodal platform aims, FIRST, to develop an integrated Graph Convolutional Network Multimodal Platform (GCN-MP) technology, to fuse small datasets of different modalities/features/diseases/populations via a multimodal GCN model for accurate prediction of LOAD onset age; SECOND, to uncover early biomarkers and principal pathways driving LOAD onset. Five novelties have been proposed, including, heterogeneous multimodal inputs, multimodal data fusion, potential causal pathways identification, and integrated multimodal risk scores (IMRS) based on high-saliency complementary modalities. Our work will greatly facilitate clinicians and at-risk AD populations to more conveniently and accurately predict LOAD onset via the IMRS and onset age prediction, based on their available data types, and advance our understanding of LOAD etiology. Our team’s two consecutive grants awarded by NAM, and the numerous works done on AI-driven prediction and estimation and in neurological studies from our members in HKU-Cambridge-AI-for-Neuro-disease-Research-Platform have prepared us to achieve these meaningful goals of AI for Social Good.

Project Title 2: AI-driven causal model to determine upstream definitive genetic markers for early detection of Late Onset Alzheimer’s Disease
Funding Source: US National Academy of Medicine (NAM) Healthy Longevity Catalyst Award, 2022
Funding: HKD 0.479 M
Led by: Prof. Victor OK Li, Principal Investigator and Dr. Jacqueline CK Lam, Co-Lead (HKU)

Abstract
Alzheimer’s disease (AD) is a leading cause of death in China and the fifth leading cause of death worldwide. Early diagnosis is critical to provide timely intervention before irreversible brain damage occurs. 95% of the AD cases occur after age 65 and are considered late-onset AD (LOAD). Although plasma amyloid, phosphorylated Tau, and neurofilaments for individualized risk prediction in mild cognitive impairments (MCIs) are emerging, no causal definitive genetic markers (DGMs) for early detection of LOAD have been established, making accurate early diagnosis and treatment difficult. With the availability of large AD datasets with genetic information and our own Israeli and Hong Kong datasets, it is now possible to analyse the relationship between genetic markers and LOAD. However, (1) traditional statistical or data-driven models only ascertain disease association rather than causation; (2) deciphering disease-driving somatic mutations from inconsequential mutations remains difficult. To overcome these challenges, we will develop an AI-driven causal model to investigate multiple AD pathological pathways and their convergence, utilizing existing AD datasets to efficiently and accurately identify upstream DGMs. To verify the DGMs identified from the AI model, two large ethnically-distinctive LOAD datasets with blood samples from the HK Chinese and Israeli populations developed/accessible by our team will be analysed. Our novelties include the development of an AI-driven causal model which incorporates key AD pathways, somatic mutations, and demographics, to identify what disease-driving biomarkers cause (rather than what associate with) LOAD, paving the way for the early detection and treatment of LOAD.

Project Title 3: A directed graph neural network-based drug-repurposing approach to identify a lead combination of effective drugs for Alzheimer's Disease
Funding Source: US National Academy of Medicine (NAM) Healthy Longevity Catalyst Award, 2021
Funding: HKD 0.479 M
Led by: Prof. Victor OK Li, Principal Investigator and Dr. Jacqueline CK Lam, Co-Lead (HKU)

Abstract
Worldwide, around 50 million people are suffering from Alzheimer’s Disease (AD) and related forms of dementia, resulting in 28.8 million disability-adjusted life-years (DALYs), posing a significant threat on human longevity and quality of life globally. To date, no effective disease-modifying treatment or preventative therapies have been found, while the search for effective drug candidates is lengthy and data-constrained. To address this challenge, we propose a novel, ground-breaking AI-driven Directed Graph Neural Network (GNN)-based Drug-Repurposing Approach, capitalizing on the association of somatic mutations in AD pathology and the identification of 272 very long gene targets. Our approach will embed the 272 protein pathway data, and make use of relevant available big genetic and drug datasets, to determine a lead combination of effective drug candidates that interacts with mutation phenotype either directly or through network-based actions. Our novelties include: (1) a directed GNN drug-repurposing approach to identify drug candidates; (2) domain-specific somatic mutations/genes incorporated into the biomedical graph to determine a lead combination of candidate drugs that interacts with somatic mutation phenotype either directly or through network-based actions; (3) domain-specific genetic directed pathways and long genes incorporated; (4) knowledge of co-morbidities of AD incorporated; (5) a lead combination of effective drugs, instead of single drugs investigated; (6) a causal model integrated to validate the lead combination of candidate drugs and confirm its impacts on genes, proteins, and behaviours, associated with AD. This longevity- and quality-of-life-driven AI drug-repurposing study will significantly accelerate the process and precision of AD drug identification.

Project Title 4: Big Data for Smart and Personalized Air Pollution Monitoring and Health Management
Funding Source: RGC Theme-based Research Scheme 7th Round
Funding: HKD 50 M (USD 6.3 M)
Led by: Prof. Victor OK Li, Project Coordinator and Dr. Jacqueline CK Lam, Co-PI (HKU)

Abstract
We are all entitled to live with dignity in a clean environment. With big data technologies, it is possible to collect complex, heterogeneous, high resolution, personalized, and synchronized urban air pollution, human activity, health condition, well-being, and behavioral data, enabling the generation of smart (real-time and interactive), personal alert and advice to improve the health and well-being of individual citizens, creating new business opportunities and competitive advantage for the IT and health industry in HK and beyond. There are five major challenges. FIRST, urban air quality data is sparse, rendering it difficult to provide timely personalized alert and advice. SECOND, collected data, especially those involving human inputs, such as health perception, are often missing and erroneous. THIRD, data collected are heterogeneous, and highly complex, not easily comprehensible to facilitate individual or collective decision-making. FOURTH, the causal relationships between personal air pollutants exposure (specifically PM(2.5,1.0) and NO2) and personal health conditions, and health (well-being) perception, of young asthmatics and young healthy citizens in HK, are yet to be established. FIFTH, one must determine if information and advice provided can effect behavioral change. To overcome these challenges, our FIRST novelty is to develop a big data framework based on deep learning to estimate smart personalized air quality. Our SECOND novelty includes the deployment of mobile pollution sensor platforms to substantially improve the accuracy of estimated and forecasted air quality data, and the collection of activity, health conditions and perception data, accounting for human in the loop. Our THIRD novelty is the development of visualization tools, and comprehensible indexes which correlate personal exposure with four other types of personal data, to provide timely, personalized pollution, health and travel alerts and advice. Our FOURTH novelty is determining causal relationship, if any, between personal pollutants, PM(2.5,1.0), NO2 exposure and personal health conditions, and also personal health perceptions, based on clinical experiments of 250 young asthmatics and 250 young healthy citizens in HK. An exposure model is developed, trained and verified with real data collected by 250 young asthmatics to further conduct population-based time series health study on 90% of asthmatics in HK. Our FIFTH novelty is an intervention study to determine if smart data, presented via our proposed system, will induce personal behavioral change. Our novel big data technologies and analytical approaches create a unique framework for personalized air pollution monitoring and e-health management, easily transferrable to and applicable in other domains and countries.