Last year's Nobel Prize in Chemistry was awarded to researchers who developed artificial intelligence (AI) for predicting protein structures. AI is rapidly changing the landscape of life sciences research. From protein structure prediction to drug candidate discovery and single-cell analysis, it's now hard to imagine research in biotechnology without AI.
On the 25th of last month, four women researchers from Korea and abroad in the AI and biotechnology fields gathered at Yonsei University's Sinchon Campus. Bonnie Berger, a professor at the Massachusetts Institute of Technology (MIT) and chair of the organizing committee for the 'Computational Molecular Biology Research (RECOMB) 2025' event, along with Kwon Soon-kyung, a professor at Gyeongsang National University and representative secretary of the Korean Academy of Science and Technology (Y-KAST), Min-Kyoung Baek, the first author of a Nobel Prize-winning paper, and Jeong Hyo-bin, an expert in single-cell spatial genomics, discussed the current status of AI applications in biotechnology research, its limitations, and future prospects.
The conversation centered on how the integration of AI and life sciences is transforming research environments. Professor Berger noted, "Since last year, I've started a course on 'Generative AI for Biology,' and although there was no advertising, students flocked to it," adding, "AI has now become central to biotechnology research." Other attendees echoed that the proportion of AI-related papers at the RECOMB 2025 Seoul conference held at Yonsei University had significantly increased compared to previous years.
However, as the conversation progressed, the gap between the rapid technological advancements and the research environment became clearer. While interest in AI has risen, support for necessary education remains inadequate.
Professor Berger stated, "While teaching, I had to personally purchase graphic processing unit (GPU) equipment to provide AI coding practice for students," and added, "I could barely fill the lack of computational resources with free credits from Google Colab." He also mentioned, "One student accidentally mishandled data on Google Cloud, resulting in an $8,000 (about 11 million won) bill."
The situation is similar in Korea. Professor Baek noted, "Currently, the Korean government is expanding its AI research budget, but investment in infrastructure is as important as research funding," and explained, "Without support from supercomputers or GPU-based environments, it’s hard to achieve results."
In fact, the Ministry of Science and ICT conducted a survey last month on the GPU demand from 405 research institutions, excluding large corporations, and found that 45.9% of them responded that they needed GPU resources within six months. Notably, 79.2% of the responding institutions indicated a need for GPUs within a year, showing that most organizations are in urgent need.
Data utilization constraints are also an issue. The reason the protein structure prediction AI has been able to bring about innovation is that experimental data from around the world has been standardized and compiled into the Protein Data Bank (PDB). However, it is difficult to analyze large-scale data using AI.
Professor Jeong stated, "Another major barrier to applying AI to biotechnology is data privacy," adding, "Building AI models requires large-scale training data, but patient-derived genomic data, microbiomes (symbiotic microorganisms), and clinical trial information face strict security regulations, making it challenging to collect big data through cross-border or inter-institutional data sharing."
As an alternative, a 'multi-party homomorphic encryption' technology was introduced. This approach allows each hospital to keep its data while sharing only encrypted intermediate calculation values for AI training. Professor Berger added, "This method has been used in research projects with the UK Biobank and the National Institutes of Health (NIH) in the U.S."
The problem is not limited to infrastructure. There is an absolute shortage of skilled personnel who can connect AI and biotechnology. Professor Baek pointed out, "There are many classes that teach AI alone, but there is almost no education that connects it to biology," emphasizing, "AI needs to be taught to biologists, and biology needs to be taught to AI majors."
Professor Berger also noted, "Until a year ago, we at MIT were still teaching deep learning content from ten years ago," stressing that a new curriculum tailored to the current era is necessary. Deep learning is a field of AI machine learning that uses multilayer neural networks mimicking the structure of the human brain to learn and predict data.
Professor Kwon mentioned, "Even when looking to recruit professors in the field of biotechnology and AI convergence, there are no applicants," stating, "AI experts often leave for large corporations right after graduation with salaries in the hundreds of thousands, leaving academia to face challenges in attracting and retaining talent."
The researchers expressed optimistic outlooks for the next decade in life sciences that AI will transform. Professor Kwon stated, "Whereas until now it has been a molecule-centric science, in the future it will transition to a data-driven science led by AI."
Professor Jeong also predicted that the integration of AI and biotechnology would extend beyond predicting the structures of biomolecules and analyzing individual layers of data, enabling the concept of 'virtual cells' that allows cell-level simulations without animal testing.
Professor Baek highlighted the need for a new integrated AI model for the convergence of AI and biotechnology. He remarked, "Currently, data is analyzed separately, from proteins to cells, tissues, and clinical trial data, but in the future, we will need an AI model that integrates this understanding," indicating that the most important task over the next decade will be to develop an AI model architecture that consolidates this entire system.