AI Protein Design: Stable Intrabodies from Science Tokyo | AcademicJobs

Science Tokyo Leads AI Protein Design for Intrabodies

New0 comments

Be one of the first to share your thoughts!

Add your comments now!

Have your say

Engagement level

See more Research Publication News Articles

a container of protein powder next to a spoon
Photo by Aleksander Saks on Unsplash

Breakthrough in AI-Driven Intracellular Antibody Design at Institute of Science Tokyo

In a groundbreaking advancement for biomedical research, scientists at the Institute of Science Tokyo (Science Tokyo) have harnessed artificial intelligence to create stable intracellular antibodies, known as intrabodies. These engineered proteins function effectively inside living cells, addressing long-standing challenges in studying cellular processes. Led by Professor Hiroshi Kimura from the School of Life Science and Technology, the team published their findings in Science Advances on January 2, 2026, demonstrating a pipeline that converts conventional antibody sequences into functional intrabodies with remarkable efficiency. 68 69

This innovation not only triples the number of available probes for histone modifications but also paves the way for new diagnostics, imaging techniques, and therapies. As Japan's higher education landscape evolves, Science Tokyo's interdisciplinary approach—born from the 2024 merger of Tokyo Institute of Technology and Tokyo Medical and Dental University—exemplifies how university collaborations drive cutting-edge science. 66

The Rise of Science Tokyo: A New Powerhouse in Japanese Higher Education

Established on October 1, 2024, the Institute of Science Tokyo emerged from the strategic merger of two prestigious institutions: Tokyo Institute of Technology (Tokyo Tech), renowned for engineering and science, and Tokyo Medical and Dental University (TMDU), a leader in medical and dental research. Located primarily in Yokohama, Science Tokyo fosters cross-disciplinary initiatives in areas like life sciences, AI, and disaster response, aiming to tackle global challenges through visionary research. 56 57

Professor Kimura's lab, part of the Cell Biology Center in the Institute of Integrated Research, specializes in tracking histone modifications and RNA polymerase dynamics using fluorescent probes called mintbodies. This latest work builds on years of expertise in live-cell imaging, positioning Science Tokyo as a hub for AI-biotech fusion in Japan. For aspiring researchers, opportunities abound in such dynamic environments—check out research jobs and postdoc positions to join similar teams. 46

Institute of Science Tokyo campus in Yokohama, hub for AI and life sciences research

Challenges in Developing Intracellular Antibodies

Traditional antibodies excel at binding targets outside cells, such as in blood or tissues, but falter inside the cytoplasm. The cell's reducing environment breaks disulfide bonds essential for antibody structure, leading to misfolding, aggregation, and loss of function. Historically, success rates for converting antibodies to intrabodies hovered at 5-10%, making them rare tools for probing intracellular events like protein interactions or epigenetic changes. 67

Histone modifications—chemical tags on DNA-packaging proteins that regulate gene expression—are particularly hard to study dynamically. Existing methods rely on fixation or indirect labeling, missing real-time insights. Science Tokyo's team targeted these pain points, focusing on single-chain variable fragments (scFvs), the smallest functional antibody units. 69

The AI Pipeline: Step-by-Step Innovation

The core of this breakthrough is an accessible AI-driven workflow that anyone with computational resources can use. Here's how it works:

  • Annotation: Use ANARCI software to identify complementarity-determining regions (CDRs)—the antigen-binding parts—and framework regions.
  • Structure Prediction: Employ LocalColabFold, based on AlphaFold2, to model the scFv bound to its target peptide or histone tail.
  • Sequence Redesign: Apply ProteinMPNN to redesign framework regions, fixing CDRs and nearby residues (within 3-5 Å) for stability. Generate 15 designs per antibody, ranked by predicted Local Distance Difference Test (pLDDT) scores.
  • Live-Cell Screening: Fuse top designs to green fluorescent protein (GFP), transfect into cells like HEK293T or HeLa, and validate via imaging, FRAP (fluorescence recovery after photobleaching), and binding assays.

This process, coded openly on GitHub, reduces development time from months to days. 67 69

Impressive Results: 73% Success Rate

Testing 26 diverse antibodies—from anti-FLAG (M2) to SARS-CoV-2 nucleocapsid epitope and 19 histone-targeting ones—the pipeline succeeded with 19 functional intrabodies (73%). Notably, 18 had failed conventional methods. For histones, 17/24 worked, tripling probes like H3K27ac, H3S10ph, and H4K16ac. 68

Key metrics included nuclear-to-cytoplasmic (N/C) ratios confirming localization, FRAP half-times of 2-15 seconds indicating reversible binding, and solubility under reducing conditions. Successful designs showed biases like serine-to-lysine mutations for stability.Read the full paper for detailed figures. 69

Schematic of AI pipeline using AlphaFold2 and ProteinMPNN for intrabody design

Applications in Live-Cell Imaging of Gene Regulation

The redesigned mintbodies (minimal intrabodies for nuclear targets) enabled real-time tracking of histone dynamics. For instance, H3K27ac mintbodies responded to trichostatin A (TSA), an HDAC inhibitor, by increasing nuclear enrichment. During mitosis, they visualized phosphorylation changes like H3S10ph. These tools reveal S-phase replication foci and transcription sites, offering unprecedented views of epigenetics. 67

Such precision aids studies on cancer, neurodegeneration, and development, where histone marks dictate cell fate.

Broader Biomedical and Therapeutic Potential

Beyond imaging, stable intrabodies hold therapeutic promise. They could inhibit pathogenic proteins inside cells, like viral components or oncogenes, without delivery issues of small molecules. Diagnostics benefit from specific intracellular detection, and the pipeline scales with growing antibody databases from phage display or immunized animals.

Cost savings are significant: no need for animal immunization or extensive screening, just sequence-to-function conversion.

International Collaboration Fuels Success

This work unites Science Tokyo with Colorado State University (led by Assoc. Prof. Timothy J. Stasevich) and Kyushu University (Prof. Yasuyuki Ohkawa). PhD student Daiki Maejima bridged computational and experimental arms. Such partnerships highlight Japan's role in global biotech, attracting talent via programs like faculty positions in Japan.

Implications for Japanese Higher Education and Research Careers

Science Tokyo's rapid output post-merger underscores Japan's push for world-class universities. Government initiatives like the ¥10 billion funding boost for research exemplify support. For students and postdocs, labs like Kimura's offer training in AI, cell biology, and imaging—ideal for careers in biotech. Explore Japan university jobs or academic CV tips.

Future Directions and Challenges

Next steps include integrating advanced tools like AbMPNN, training on failure data for better predictions, and expanding to nanobodies or full IgGs. Challenges remain, like unpredictable cases (e.g., H3K9me3 failure), but open-source code democratizes access.

Kimura notes: "AI allows redesign of structures compatible with the cellular environment." This could revolutionize protein engineering globally.

Why This Matters for Researchers and Students

For higher ed professionals, this exemplifies AI's role in accelerating discovery. Institutions like Science Tokyo prioritize interdisciplinary PhDs and postdocs. Actionable advice: Master Python for tools like ProteinMPNN; contribute to GitHub repos; network via conferences. Visit Rate My Professor for insights on mentors like Prof. Kimura.

Conclusion: Pioneering the Future of Cellular Research

The Institute of Science Tokyo's AI-driven intrabodies mark a milestone in protein design, empowering precise intracellular studies. As Japan invests in higher ed innovation, opportunities flourish for researchers worldwide. Stay ahead with higher ed jobs, university jobs, career advice, and professor reviews at AcademicJobs.com. For Japan-specific roles, see /jp.

Learn more at the Science Tokyo press release.

crowded pedestrian lane in urban area at night

Photo by Fabrizio Chiagano on Unsplash

Discussion

0 comments from the academic community

Sort by:
You

Please keep comments respectful and on-topic.

Frequently Asked Questions

🧬What are intrabodies?

Intrabodies are engineered antibodies (scFvs) that function inside living cells, overcoming stability issues in the reducing cytoplasm for imaging and therapy.

🤖How does the AI pipeline work?

It uses ANARCI for annotation, AlphaFold2 for structure prediction, ProteinMPNN for framework redesign, and live-cell screening for validation. GitHub code.

📈What is the success rate?

73% (19/26 antibodies), tripling histone mintbodies like H3K27ac.

👨‍🔬Who led this research?

Prof. Hiroshi Kimura at Institute of Science Tokyo, with Daiki Maejima, Timothy J. Stasevich (Colorado State), and Yasuyuki Ohkawa (Kyushu Univ).

🔬What are mintbodies?

Fluorescent intrabodies for monitoring histone modifications in live cells, tracking epigenetics dynamically.

🩺Applications of these intrabodies?

Live imaging of gene regulation, viral proteins, potential therapeutics targeting intracellular pathogens.

🏫What is Institute of Science Tokyo?

Merged Tokyo Tech and TMDU in 2024, focusing on interdisciplinary life sciences and AI in Yokohama.

Challenges overcome by AI?

Misfolding and aggregation in cytoplasm; redesigned frameworks ensure solubility and specificity.

🚀Future of AI in protein design?

Scalable to nanobodies, therapeutics; open-source accelerates global adoption.

💼Career opportunities at Science Tokyo?

Postdocs, faculty in AI-biotech. See postdoc jobs and Japan roles.

💻How to use the pipeline?

Download from GitHub, run on ColabFold; validate in your cell lines.