Cracking the code of diseases—Artificial intelligence meets genomics

Artificial intelligence (AI) technologies are ubiquitous, affecting our lives in numerous ways. In healthcare, AI helps improve the efficiency and accuracy of medical services. With algorithms for machine learning and automation, doctors and clinicians are now able to perform tasks more effectively, from detecting diseases, making diagnoses, and predicting the onset and progression of diseases, to providing personalised treatment. These exciting health technologies may look new to us. However, AI in healthcare is actually a field in which computer scientists and researchers have been working for decades, long before its rise to prominence in recent years.

Prof. ZHANG Weixiong, Chair Professor of Bioinformatics and Integrative Genomics in the Department of Health Technology and Informatics, has been deciphering and decoding the complex genetic and biological systems that underlie human disorders through computational and mathematical analysis and modelling. He joined PolyU in 2022 under the Global STEM Professorship Scheme of the HKSAR, after spending more than 20 years at Washington University in St. Louis (WashU) in the United States (US). This year, Prof. Zhang took on an additional role as Associate Director of PAIR. 

This new chapter at PolyU, according to Prof. Zhang, is an opportunity to “really do something he has sought for a long time”—turning his innovations into clinical applications.

Prof. ZHANG Weixiong

When the research journey comes full circle

Prof. Zhang, you lived in the US for 30 years after graduating from Tsinghua University in the early 1980s. What attracted you to move to Hong Kong and join PolyU?

It has been a long journey for me to come to Hong Kong and to get into life science research. Thirty years ago, when I earned my PhD in Computer Science from the University of California, Los Angeles, I received job offers from universities in Hong Kong and Singapore. But I decided to stay in the US for my career due to family reasons.

I became a professor in computer science and genetics at WashU. I have always been interested in life science research and particularly interdisciplinary research on computational genomics and computational systems biology. Back then, my research works were strongly focused on the computational side. I never got the chance to get my feet wet in real biology research as I did not have my own molecular biology laboratory for bench work. The offer from PolyU included a “wet lab”. This was a good opportunity for me to really do something that would have been impossible before I joined PolyU. With the additional support of the Hong Kong Global STEM Professorship from the HKSAR government and the Hong Kong Jockey Club, I have set up an integrated interdisciplinary Genomics and AI research laboratory.

I have spent more than 20 years working in the interdisciplinary field of computational biology. I have already developed some primary computational tools with big data and AI. However, these tools have not been applied to real-life settings for clinical applications. There are several objectives and goals that I have always wanted to achieve in my career. At PolyU, I can see the opportunities to realistically achieve some of these goals. My lab is now working on projects involving psychiatric disorders and cancers by combining computational approaches and experimental biology techniques.

On a mission for better health: Turning medical problems into computational problems

What are some scientific questions that you are eager to solve? 

There are a few outstanding scientific problems that I want to solve. Studying complex diseases (e.g., cancers, mental disorders and cardiovascular diseases) is one of them. In a scientific sense, these diseases are often determined by the interaction and correlation of a group of genes, rather than a single gene. Right now, there is no good tool for studying complex diseases. The problem needs to be treated as a network of problems. The medical problem now becomes a computer science problem.

Psychiatric disorders like depression and schizophrenia, for example, are complex diseases. The diagnoses of these diseases are based on clinicians’ and doctors’ judgements of the patients’ symptoms, including mood, social interaction, behaviours and other self-descriptive information that may be signs of hallucination or suicidal thoughts. The judgement and reporting can be subjective. This can make diagnosis very difficult and inaccurate.

So, instead of looking at and focusing on symptom data, there is a need for objective tools similar to lab tests. This is like taking a lab test when we have the flu. But in the case of psychiatric disorders such as depression, I want to develop tools that resemble blood tests or brain imaging, and to see if I can find some biomarkers that are indicative of depression development.

Seizing opportunities in research: Act fast, work hard, build a team

In July 2023, you were awarded HK$37 million in funding from the Strategic Topics Grant (STG) 2023/24 by the Research Grants Council (RGC) for a five-year health tech project focused on psychiatric disorders. This project received the highest funding allocation among all funded projects in this round of STG competition. Can you describe your experience with the STG?

Obtaining a large amount of funding just about a year after my arrival in Hong Kong seems very difficult to achieve. Many of my colleagues and friends wonder how I managed this. Indeed, I knew few people in Hong Kong at that time, and the proposal development took place during the COVID period when connection and communication were affected.

The entire process, from writing the proposal and finding collaborators, to going through university-level internal hurdles like interviews and screening, happened within three months after the call for proposals. I was very excited when I learnt about the STG because it covers AI in healthcare, an area I have been working on for many years. Therefore, I quickly pulled my thoughts together in order to write and refine the plan, and identified the specific diseases to be treated, the fundamental scientific questions to be addressed, and the advanced technological solutions needed. I then reached out to potential collaborators and built a research team in a short period of time.

Finding hidden patterns in diseases: Harnessing AI in computational genomics

The health tech project aims to develop AI, genomic and biomedical technologies to support objective diagnosis and personalised therapy for major psychiatric disorders, including depression, schizophrenia and bipolar disorder. The project supports AI-based, data-driven diagnosis and personalised therapy. How can this be achieved?

The overall scope of the project is very broad. On the scientific front, it touches upon the life science side, such as genomics, molecular biology, epigenetics and medicine, as well as the technological side, such as AI and biotech tools. In brief, the project comprises three main aspects.

For the first aspect, the research team will collect a lot of data, including DNA variations, genomic variations, brain images and much more, and will develop AI tools to analyse these gigantic amounts of data and find patterns in them. We will then use the DNA signatures and image signatures identified to help clinicians perform more accurate diagnoses.

For the second aspect, we will build animal models of depression, to understand the causes, mechanisms, onset, progression and inheritance. In mouse models, we will introduce stresses and investigate their effects on disease development. One area is physical stress, such as disruptions to the circadian rhythm by not letting the animals go to sleep on a regular cycle, and starvation by giving them only a little food. Another is social stress, such as putting animals into fear-eliciting situations, which can cause symptoms of depression.

We will perform studies at a miniature cellular level, such as conducting single-cell sequencing to examine the gene expression and biomarkers in these animals, as well as brain imaging to identify functional impairments to the brains. We will also seek to understand how these psychiatric disorders are passed on from one generation to the next. To achieve this, we will breed mice that exhibit depression symptoms, and see if these molecular markers get passed on to the second and third generations of offspring. Data collection and analysis are a costly process and require high-performance computers to support the analysis of a large quantity of biological and imagery data.

We are interested in the effects of stress. As I have mentioned, psychiatric disorders are complex diseases determined not only by individual genes (e.g., DNA mutation that can cause problems), but also by their interactions with the environment, which is known as the “epigenome”. Enzymes in the cell can modify the DNA in some sense. For example, if a person is put under stress at an early stage of life, the P11 gene (a gene that affects serotonin neurotransmission and is implicated in depression) is affected. The DNA sequence that drives the expression of P11 gene becomes methylated, so the gene expression of P11 is affected. In this case, the P11 gene is not mutated at the DNA level, but its chemical property gets modified, which can also contribute to psychiatric problems.

From scientific discoveries to applications, from animal to human translation

How will the research findings be used for practical applications in frontline healthcare settings?

This relates to tool development and clinical trials, the third aspect of the project. Currently, repetitive transcranial magnetic stimulation (rTMS) therapy, a form of physical brain stimulation therapy, is used to treat depression, in addition to psychiatric medication. rTMS has been approved by the Food and Drug Administration (FDA) and has gone through many years of clinical applications.

Based on previous findings, we will develop an AI-based system that can automatically generate recommendations for personalised rTMS therapy. It is more or less like a ChatGPT tool for doctors. We will run clinical trials in collaboration with hospitals based in Hong Kong and Mainland China. 

One common question is whether this kind of AI system will eventually replace doctors. In my view, this will not and should never happen. The system we are trying to build basically looks at brain images, picks up the subtle differences that doctors may miss or overlook, and circles the particular brain regions that require doctors’ special attention. This system has the potential for commercialisation.

New mission to drive PolyU interdisciplinary research

You were appointed as the Associate Director of PAIR with effect from August 2024. Can you describe some of your experiences at PAIR so far?

I have had opportunities to represent PAIR on several occasions in Mainland China and have helped to establish new research institutes there. PAIR has the mission of broadening the scope of research and achieving impact. One way to do so is to collaborate with individuals and organisations in Mainland China and the Greater Bay Area.

There are a lot of opportunities for PAIR research in life science and healthcare areas. We can collaborate with hospitals and professionals in Mainland China, given their large patient bases. There are many external collaborations going on at PAIR already. This is the right move.

Breaking down disciplinary walls: Leveraging AI and data science to drive interdisciplinary research

Do you have any further suggestions on PAIR developments in terms of collaboration, research and talent development?

At this time, I see opportunities for different PAIR research centres and institutes to collaborate. By collaborations, I mean not only those involving different units under the health theme (e.g., food and ageing), but also those involving units across different themes, including AI and the Internet of Things.

In my view, AI plays a fundamental role in research. Almost all fields of research involve collecting data, and conducting research really boils down to data science—how we analyse data and unveil the real story underneath.

PAIR can take a leadership role in encouraging the integration of AI in interdisciplinary research. One possibility is through the PAIR Young Fellowship Scheme. The young fellows will work in an interdisciplinary research field and be co-supervised by faculty members from completely different fields, say, health and data science. This helps create bridges to different PAIR units to promote truly cross-disciplinary research.

In the bigger picture, there is also a need for university-level measures to tear down the walls between academic units and create a collegial environment so that people from different faculties, schools and departments can collaborate more easily. One possibility is the offering of joint academic appointments. This may add new challenges from management and administration perspectives. However, in interdisciplinary research, there should not be these kinds of barriers. After all, it is part of PolyU’s culture that the different units at the university need to work on together.