A new artificial intelligence-driven pipeline developed in a collaborative research combines protein structure prediction, sequence design, and live-cell screening together to enable rapid conversion of antibody sequences into functional intracellular antibodies (intrabodies) that are stable within living cells. By preserving antigen-binding regions and improving structural stability, the approach overcomes major barriers encountered in intrabody development—emerging as a simpler, more cost-effective tool for diagnostics, imaging, and biomedical research.
Antibodies are proteins that are widely used in biology and medicine. These are target-selective and can precisely recognize and bind to their targets. Due to their selectivity, they are often used to detect and study specific molecules. However, most conventional antibodies only work outside living cells, which limits their ability to probe important biological processes that occur inside cells. Intracellular antibodies (also known as intrabodies) offer a way to overcome this limitation by functioning within living cells. However, developing intrabodies has proven to be quite challenging as antibodies tend to misfold or lose activity within the cells.
To tackle this issue, a team of researchers led by Professor Hiroshi Kimura from the Institute of Integrated Research, Institute of Science Tokyo (Science Tokyo), Japan, along with Mr. Daiki Maejima, a third-year doctoral student from Science Tokyo, Associate Professor Timothy J. Stasevich of Colorado State University, USA, and Professor Yasuyuki Ohkawa of Kyushu University, Japan, developed a new design strategy that uses artificial intelligence to design functional intrabodies. Their findings were published in Volume 12, Issue 1 of the journal Science Advances on January 02, 2026.
“We created a pipeline that combines artificial intelligence (AI)-guided protein structure prediction with sequence redesign and live-cell screening,” says Kimura.
The method keeps the target-binding regions intact and redesigns the surrounding framework regions. This ensures that the antibodies fold appropriately and stay stable within the cells without compromising their ability to bind to specific molecules.
The team tested around 26 existing antibody sequences. Among these, 19 were successfully converted into functional intrabodies. Interestingly, 18 of these had previously failed to work as functional intrabodies when researchers used conventional approaches.
“Many antibodies that were not functional when expressed as intrabodies were ultimately found to be converted into functional ones in our study,” comments Kimura. “This confirms that AI allowed us to redesign structures that were compatible within the cellular environment.”
The major focus of the study was intrabodies that specifically target modifications in histone proteins (proteins that play a key role in packaging DNA and regulating gene activity). These modifications can serve as stable markers of gene activity or can change rapidly and are difficult to study using traditional labeling techniques. The newly designed intrabodies were able to detect these changes within living cells and responded as modification levels increased or decreased based on fluorescence, demonstrating high potential for studying gene regulation.
Further experiments confirmed that the redesigned molecules were stable, functional, soluble, and retained high target specificity within the living cell. Moreover, they also showed consistent and predictable behavior across varying cellular conditions, highlighting the reliability of the research tool.
“By combining AI-based design with live-cell testing, we can now accelerate intrabody development with far greater confidence,” explains Kimura.
Beyond basic research, this approach has the potential to streamline intrabody development, making it faster, cheaper, and more accessible. As antibody sequence databases continue to expand, the ability to convert existing antibodies into functional intracellular probes could enable a broad range of applications, from diagnostics and fluorescence imaging to the development of therapeutic strategies. Overall, the study shows how AI can help unlock the full potential of antibodies, opening new paths for exploring complex biological processes.
***
About Institute of Science Tokyo (Science Tokyo)
Institute of Science Tokyo (Science Tokyo) was established on October 1, 2024, following the merger between Tokyo Medical and Dental University (TMDU) and Tokyo Institute of Technology (Tokyo Tech), with the mission of “Advancing science and human wellbeing to create value for and with society.”
Science Advances
Experimental study
Cells
AI-assisted protein design to rapidly convert antibody sequences to intrabodies targeting diverse peptides and histone modifications
2-Jan-2026
The authors declare that they have no competing interests