A computationally designed TCR-like antibody recognizes fragments of the KRAS G12D mutation displayed on cancer cell surfaces, a target long considered out of reach.

Antibodies have become some of medicine’s most precise weapons against cancer. They latch onto proteins on the outside of tumor cells and either block their signals or flag the cells for destruction. But they have a fundamental limitation: they cannot get inside cells. Many of the mutations that actually drive cancer sit in proteins locked within the cell, invisible to conventional antibodies.

Researchers at the Korea Advanced Institute of Science and Technology (KAIST) say they have found a way around that blind spot. Working with TheraZyne, a protein-design startup founded by KAIST faculty, Professor Byung-Ha Oh’s team in the Department of Biological Sciences has designed an antibody that selectively recognizes cancer cells carrying one of the most common cancer-driving mutations, KRAS G12D. The work was published in Molecular Therapy.

The ‘undruggable’ protein

KRAS is a switch-like protein that relays growth signals inside cells, telling them when to grow and divide. Certain mutations jam the switch in the ‘on’ position. The G12D mutation, a single amino acid change at position 12 of the protein, is among the most frequent drivers of pancreatic, colorectal and lung cancers.

For decades KRAS was considered undruggable. Its smooth surface offered few pockets where small-molecule drugs could bind, and because it lives inside the cell, antibodies could not reach it at all. That picture has begun to change in recent years with small-molecule inhibitors aimed at specific KRAS mutations, but KRAS-driven cancers, especially pancreatic cancer, remain among the hardest to treat.

Reading the cell’s discarded fragments

The KAIST approach does not try to get inside the cell. Instead, it takes advantage of the cell’s own housekeeping. Cells routinely chop up old and damaged proteins, including mutant KRAS, and display some of the resulting fragments on their surface on molecules of the major histocompatibility complex, a kind of molecular display case that lets the immune system inspect what is happening inside. Fragments derived from mutations are known as neoantigens, because they are new to the body and can mark a cell as abnormal.

This is how T cells, the immune system’s roving inspectors, recognize cancerous or virus-infected cells: their T cell receptors read the fragments presented on the cell surface. The KAIST team set out to build an antibody that works the same way, a so-called TCR-like antibody that recognizes not the whole KRAS protein but the specific fragment produced by the G12D mutation, as it sits in its display case.

To design it, the researchers combined computational protein design with experimental screening, an approach the paper describes as an in silico-in vitro workflow. Computer models proposed antibody candidates that should fit the mutant fragment, and laboratory tests then winnowed them to the ones that actually worked.

Selective recognition

According to the release, the resulting antibody recognized cells carrying KRAS G12D while showing little or no reaction to normal cells or to other proteins. When the team tested it in an immunotherapy setting, it eliminated only the mutation-carrying cancer cells.

“The antibody developed in this study can selectively identify only cancer cells carrying the KRAS(G12D) mutation,” Oh said, “demonstrating the potential for precision antibody therapeutics that minimize damage to normal cells.”

That selectivity is the crux. Because the target fragment exists only in cells carrying the mutation, a therapy built on such an antibody could, in principle, attack tumors while leaving healthy tissue alone, a long-standing goal of cancer treatment.

The general strategy has already proven itself in the clinic. In 2022, regulators approved tebentafusp, a drug built from an engineered T cell receptor, for a rare eye cancer called uveal melanoma. It recognizes a protein fragment displayed on the tumor’s surface, the same kind of target the KAIST antibody is designed to read. Antibodies with T cell receptor-like specificity are harder to engineer, because the differences between a mutant fragment and its normal counterpart can come down to a single amino acid, but they offer the stability and manufacturing advantages of antibodies.

A platform for other mutations

Oh sees the work as a template rather than a one-off. “The computational antibody design technology developed in this research is expected to be widely applicable to the development of next-generation antibody therapeutics targeting KRAS as well as various other cancer mutations,” he said.

The approach has a built-in constraint worth noting. Molecules that present protein fragments vary considerably from person to person, and a given fragment is displayed only by certain variants. That means therapies of this kind typically work for patients with particular immune genetics, not everyone who carries the mutation. The release does not specify which presenting molecule the antibody was designed for.

The research is also still preclinical. The release reports no numerical efficacy data, and testing of the antibody’s effectiveness in animal disease models is underway. Many promising candidates fail at that stage or in the human trials that would follow.

SangPhil Ahn, a researcher at TheraZyne, is the first author of the study. Oh and Bo-Seong Jeong, TheraZyne’s head of research, are co-corresponding authors. The work was carried out with the New Drug Development Center of the Osong Biomedical Innovation Foundation and funded by Korea’s Ministry of Science and ICT and the National Research Foundation of Korea.

If it holds up, giving antibodies a way to read what is happening inside cells could widen the range of cancer targets they can reach, turning some of the hardest-to-treat mutations into visible marks on a tumor’s surface.

---

Notes

1. Ahn, S., et al. “Discovery of TCR-like antibodies to the KRAS G12D neoantigen via in silico-in vitro workflow.” Molecular Therapy, 3 June 2026. DOI: 10.1016/j.ymthe.2026.05.032.
2. Korea Advanced Institute of Science and Technology (KAIST). “KAIST gives antibodies ‘eyes’ to detect cancer, targeting cancer mutations inside cells.” EurekAlert!, 23 July 2026. https://www.eurekalert.org/news-releases/1137304.