Exclusive Interview with Professor Ma Lijia, Founder of Westlake Yungu Therapeutics: Pioneering AI-Driven AAV Platforms for In Vivo CAR-T and CNS Therapies

When a cup of light iced Americano and a cup of hot latte were placed in front of Professor Ma Lijia, she hesitated for a moment and chose the hot latte. "Given your regular fitness routine, the iced Americano would suit you better." Communicating with Teacher Ma always feels like a breath of fresh air; she seems to constantly try to think from the other person's perspective, which is very comforting.

Saying it's an interview, it's actually just like usual, having a chat with Teacher Ma. Every time we chat, I end up with some unexpected insights. This time, Teacher Ma started by talking about AI technology.

VCBeat: Mr. Ma, could you briefly introduce your personal background and Westlake Genetech?

Marijia Ma:I am actually half a "code farmer" by origin (Mr. Ma took a sip of his latte and smiled), with a PhD in Bioinformatics. It was during my studies and practice in bioinformatics that I realized the understanding of life could be digitized, and this perspective would become increasingly evident with the development of biotechnology. During my doctoral studies, my supervisor required me not only to understand the already digitized data itself but also to comprehend the process of digitization. Not only did I need to understand why data needed to be digitized, but I also had to know how to achieve data digitization. Therefore, during my postdoctoral period, my research direction expanded to functional genomics. In those four years, I participated in multi-center large-scale functional genomics projects organized by the NIH, establishing new methods to describe many biological processes using sequencing data, which laid a very solid foundation for my subsequent research and entrepreneurship.

Hangzhou is one of the highlands of digitalization, and Westlake University is one of the highlands in the field of life sciences. The reason for choosing Westlake University is not only because of its highly innovative atmosphere but also due to the vast prospects for the intersection of digitalization and life sciences here.

Westlake Genetech is a very interesting company. We position ourselves as a small and beautiful platform company, utilizing our technical advantages at the intersection of AI and gene and cell therapy to establish product superiority. For example, by using self-produced large-scale data and AI modeling to engineer modifications to AAV capsid proteins, then applying the modified capsids in various innovative drug development scenarios.

VCBeat: The engineering modification of capsid proteins is very interesting. As we understand, some large enterprises and Biotech companies are also working on this. What are the advantages of Westlake Genetech in this field?

Marijia Ma:Before discussing the advantages, I would like to talk about why we need to modify the capsid and how to do it more efficiently, which was constantly reinforced during my Ph.D. studies.

VCBeat: (While Teacher Ma takes a sip of coffee) This kind of Why-What-How thinking is essential in many large multinational corporations.

Marija:Yes, "code farmers" must have strong logic, and logic is also an essential requirement for successful scientific research and entrepreneurship. We all know that the three key elements of gene therapy are Delivery-Delivery-Delivery. Good gene fragments or gene editors are futile if they cannot be accurately delivered to the corresponding targets in the body. We were both at this year's CSGCT conference, where Mark Kay repeatedly emphasized this point in his report. Similarly, when I gave talks at ASGCT in May this year, it was a consensus in the industry over the years. We all understand the importance of delivery, which answers the Why question. Then, what makes a good delivery, and how to achieve good delivery? AI can be of great help here.

We generate a large amount of data to train AI to help us answer the "What" questions. It is generally believed that the discovery of AAV capsid proteins mainly involves the following approaches:

● Natural discovery: That is, searching for naturally occurring AAV in the natural world;

● Rational Design: Genetically engineering the capsid protein based on domain knowledge;

● Directed Evolution: That is, enriching AAV variants with tissue targeting through pressure screening, such as using random mutations (e.g., error-prone PCR, DNA shuffling) to generate a mutant library for screening;

● Computer-aided design: Finding novel AAV variants by simulating the secondary and tertiary structures of mutant proteins through computers.

However, there is a problem here: the sequence space that needs to be explored is extremely vast. Take the AAV2 CAP gene as an example, where all 735 amino acids could potentially become targets for modification. Assuming we only need to select 10 amino acids for modification, and without considering insertions or deletions, there would be approximately 20 to the power of 10, or 10 trillion possibilities (2 to the power of 10 is 1024, followed by ten zeros). This far exceeds the testing capacity of current experimental capabilities, so it is essential to leverage AI technology to improve efficiency. Simply put, we need to use AI technology to effectively prioritize, making rational design "smarter" under the guidance of AI. Similar drug design approaches have already been practiced for many years in the field of small molecule pharmaceuticals.

CGT and small molecules also have some differences. We know that the core components of CGT drugs are all sequence-based, whether they are nucleic acid sequences or amino acid sequences (Editor's note: CAR sequence design in CAR-T therapy, design of replacement genes in gene therapy, etc.). Sequences are quite similar to computer binary coding. Binary uses 0 and 1 to encode everything, while the core components of CGT use the permutations and combinations of the four bases A, G, C, and T to encode everything. By coincidence, sequencing technology, which can read out the A, G, C, and T bases, is currently the most suitable technology in the entire biomedical field for generating high-quality, reliable data at low cost and high throughput. With this data foundation, AI technology can play a significant role in the CGT field, and has tremendous potential.

VCBeat: As you said, delivery technology is indeed the top priority for CGT and almost all innovative drugs. How do you evaluate the several major delivery systems that are currently popular?

Marija:It's hard to evaluate other delivery systems since I haven't conducted in-depth research on them, but I have come across some in papers. Currently, the more common delivery systems include AAV, lentivirus, LNP, and VLP. Compared with other delivery systems, AAV is relatively mature. So far, eight AAV-based products have been approved by the FDA for marketing, including those for local ocular administration and systemic intravenous delivery. The safety and efficacy of AAV have been fully validated. As for LNPs, most clinical data currently available are from vaccines. Whether long-term safety for intravenous injection and non-liver targeting are feasible still requires further study. There may be some debate in the industry regarding the safety of lentiviruses. Although I believe that current safety incidents are largely related to molecular design, theoretically, non-site-specific genomic insertion does carry certain risks, which require more data to verify. Therefore, I think AAV is a relatively mature delivery technology with room for modification and innovation.

Of course, we also see a limitation of the current AAV delivery technology: all the indications targeted by commercially available products are rare diseases, which brings some challenges to the market size of AAV gene therapy drugs. If AAV can be used as a vector to provide solutions for major diseases, it will not only greatly benefit the development of the AAV industry but also allow more patients to benefit.

VCBeat: Then, in your opinion, what major indications might the AAV delivery technology be suitable for?

Marijia Ma:I would like to share a set of data first. Through the engineering transformation of AAV delivery vectors, we have achieved a technological breakthrough in the in vivo delivery of CAR sequences, which can convert 5-15% of T cells into CAR-T cells and achieve B cell killing. This success demonstrates the exciting prospects of innovative AAV vector-based in vivo CAR-T technology for treating autoimmune diseases and B-cell tumors. At the same time, this set of data may also provide some new ideas and assistance for the breakthroughs in in-vivo CAR-T technology. For another example, using the same technical platform, we have currently identified variants with far superior blood-brain barrier penetration compared to wild-type AAV. Based on the non-human primate efficacy and safety data we possess, besides being used to treat rare diseases such as spinal muscular atrophy and chorea, taking one step further will lead us into the vast universe of treating major CNS diseases including AD and PD.

Of course, AAV in terms of CMC has become very mature with nearly three decades of extensive experimentation. Lentivirus traditionally uses adherent culture processes, and its scalability has always been a challenge. Recently, some technologies adopting suspension culture have emerged, but the expression levels remain around 1e+8 vg/mL. Based on data I’ve seen both domestically and internationally, AAV has already achieved an upstream expression level of 2e+12 vg/mL and a downstream purification efficiency of 60%. With this comparison, under the premise that efficacy data is confirmed and safety is guaranteed, the accessibility of the drug can be significantly improved.

I have always believed that truly innovative technologies will gain recognition from both academia and industry. In some disciplines of life sciences, such as the field of genomic medicine where I am located, the gap between technological breakthroughs and practical applications is rapidly shrinking. We are also eagerly looking forward to sharing our data and ongoing peer-reviewed research results with everyone as soon as possible. Meanwhile, we highly anticipate more exchanges and interactions with gene and cell therapy companies, especially those enterprises focusing on CAR-T and CNS fields, to explore potential collaboration opportunities.

VCBeat: We know that many large MNCs are engaged in capsid screening and modification or directed evolution. I am particularly curious about how large a team Westlake Genetech has to carry out this work?

Marijia Ma:In fact, Westlake Genetech has a very compact team size. We only have three core teams: the early R&D team, also known as the biology team, the AI team, and a very small CMC team, with a total of less than thirty people. One thing I am particularly proud of is that the communication between our teams is seamlessly connected.

VCBeat: Is this because you play an excellent role as a lubricant in the process?

Mali Jia:In the initial stage, I definitely needed to do the "translation" work because IT people are not familiar with biology, biology people are not familiar with IT, the research team does not understand CMC, and CMC people do not understand research. Fortunately, I am quite familiar with both IT and biology, and at the same time, I am continuously learning about CMC as we move forward. Therefore, I particularly understand that the languages of different teams are different, and the difference in language is only because of certain deviations in the perspective of looking at things or the goals to be achieved or the understanding of the goals. So, in the beginning, I spent some time letting the teams understand each other and learn from each other. What made me particularly gratified was that after a period of time, I found that communication and coordination between the teams had become very smooth, and I was able to focus more energy on scientific and technological innovation. The innovations that Westlake Genetech wants to achieve and the innovations we think we can achieve must meet three criteria: first, clinicians and patients must approve; second, the payer must approve; and third, CMC must be feasible. Therefore, although we have many innovative ideas, if we find that an idea cannot be implemented, we will not rush into it easily. Thus, the Westlake Genetech team and I are still very rational. What we have always adhered to is to solidify the scientific foundation and then find the correct application scenarios. As mentioned in the previous example, using AAV for in-vivo CAR-T is an innovation point that we believe has great prospects.

Of course, there is also a little "supervisor" at Westlake Genetech (at this moment, Teacher Ma's face showed more warmth). My eldest son often comes up with some wonderful ideas inadvertently and then discusses them with me. He has grown up in the laboratory since he was a child, whether during my postdoctoral stage in the United States or after I returned to China. He loves to hang out with me in the lab, and even more, he loves to ask "why" persistently. I don't know whether he is more influenced by me or if we influence each other. When I am conducting scientific reasoning or looking at some data, I also ask myself many "whys" first, until I find the root cause of the problem before drawing a conclusion. This, I think, is also one of the reasons that urges us to solidify our scientific foundation (big laugh).

The coffee in front of each of us had almost run out, and our conversation was coming to an end. As VCBeat turned off the voice recorder, a question suddenly came to mind.

VCBeat: Mr. Ma, regarding Westlake Genetech, what is your vision for its future positioning? Do you hope to start from scratch in drug development all the way to product commercialization?

Marijia Ma:(Facing an offhand question from VCBeat, Teacher Ma still answered naturally.) "Our positioning is very clear. We know what we are good at. What we excel in is identifying the frontier technologies with the most potential for implementation from scientific discoveries and source technologies, and creating products that are half a step to a full step ahead of the market. Therefore, we advocate that people should do what they are good at. For example, we leave processes and applications to reliable partners. Hence, Westlake Genetech maintains a very open attitude towards cooperation. We focus on original innovation and once our product reaches a certain stage, we will definitely seek industry collaboration. We believe that only through cooperation, and by everyone playing to their strengths, can we successfully accomplish such a complex task as CGT and innovative drug development."

More than an hour passed in the blink of an eye, and VCBeat felt as if the discussion had just begun. We agreed with Teacher Ma that after more data on innovative technologies from Westlake Genetech are disclosed, we hope to have a deeper conversation and discussion with Teacher Ma. We deeply understand that it took physics several centuries to evolve from experimental science to theoretical science, then move into industrialization and change the world. However, biology's transition from scientific discovery to industrial application is much faster, especially with the empowerment of generative AI, which will further accelerate this process. We all look forward to more scientifically validated biotechnologies entering the industry as soon as possible, benefiting humanity and the world.