Cancer Research · asia
Making Intracellular Mutations Visible to Antibodies: AI-Designed Molecule Targets KRAS G12D Cancer Marker
A research team combined computational design with experimental screening to create a TCR-like antibody that recognizes the KRAS G12D peptide–HLA complex. Cell experiments showed that it can direct immune attacks, but eligibility is limited by HLA type, and human treatment remains some way off.
Antibodies excel at recognizing markers on cell surfaces, but cancer-causing mutations hidden inside cells are often beyond their reach. KAIST and startup TheraZyne are now attempting to bypass this obstacle: rather than directly pursuing the intracellular KRAS G12D protein, they recognize a molecular marker formed after cancer cells display a fragment of it on their surface.
The study, published in *Molecular Therapy*, targets a 10-amino-acid peptide produced by KRAS G12D and the HLA-C*08:02 molecule responsible for presenting it. These “T-cell receptor-like antibodies” mimic the way T cells recognize peptide–MHC complexes, potentially extending the reach of conventional antibodies to tumor neoantigens derived from intracellular proteins.
The team first used computational methods to generate multiple variable-region structures derived from human antibodies, simulated their docking with the target complex, and then conducted limited sequence design on the complementarity-determining regions directly involved in recognition. The researchers subsequently created a yeast surface-display antibody library and, after multiple rounds of experimental screening and optimization, obtained candidate antibodies with high affinity and selectivity. The key to this process was not to replace experiments entirely with AI, but to narrow the search space and then use physical screening to eliminate ineffective designs.
In specificity testing, the researchers used a phage-display database to test human peptides that might cause cross-reactivity. Within the range of sequences tested, they found no evident off-target reactivity against human peptides; computational analysis also predicted low immunogenicity. After the candidate antibody was converted into a chimeric antigen receptor or a bispecific T-cell engager, it selectively killed cells bearing the target complex in cell experiments. TheraZyne refers to the related candidate asset as TZ-Ab101, which is currently listed as being in the preclinical research stage.
The findings also have clear limits to their applicability. The antibody recognizes not only the KRAS G12D mutation but also the specific HLA-C*08:02 presentation context. Therefore, only tumors that simultaneously carry the matching mutation and HLA type and can adequately display the peptide may be eligible for treatment. Antigen density on tumor surfaces, cross-reactivity in different tissues, and the potential immune toxicity caused by CAR-T cells or bispecific molecules all require more comprehensive animal and human studies.
KAIST said the team is validating the effects in animal disease models, but the core evidence provided by the published paper remains at the cellular and preclinical levels and cannot be regarded as proof of efficacy in patients. At this stage, the more important significance is the establishment of a workflow spanning computational design, antibody-library screening, and immune-killing function tests. Whether it can become a viable therapy will still depend on animal safety, manufacturing quality, patient-selection strategies, and subsequent clinical trials.