{
  "abstract": "Background Peptide-MHC (pMHC) targets significantly expand the landscape for both biologics and cell therapies by enabling recognition of intracellular cancer antigens. However, the design and discovery of binders—whether TCRs or TCR mimics—against specific pMHC complexes remains a challenge, especially for novel targets. Conventional approaches (e.g. primary cell screening) are slow, often yielding low hit rates and requiring labor-intensive screening. Here, we present a rapid, fully integrated platform to generate and validate de novo protein binders against user-defined pMHC targets, combining ultra-fast computational design with high-throughput, cell-free protein synthesis and functional validation within two weeks.Methods We applied the Vcreate de novo computational design platform to generate and score minibinders against the HLA-A2-presented WT1 peptide (RMFPNAPYL), a clinically relevant target for leukemia and ovarian cancer. Millions of candidate minibinders are designed and scored for target specificity using a proprietary sequence-based scoring system. Our scoring model is over 1,000 times faster than current deep-learning protein folding algorithms (e.g., AlphaFold2), enabling the large-scale exploration of candidates in the sequence space. Top candidates were synthesized using a cell-free protein synthesis system and screened for binding to target and control pMHC complexes using artificial antigen presenting cells. The same approach was applied to a novel target presented by HLA-E, a universal but hard-to-drug HLA allele.Results Of the top 30 AI-designed binders, 6 showed strong, peptide-specific binding with minimal cross-reactivity to control peptides. Further computational optimization of lead binder AI-TCRm17 yielded 5 improved variants. The final candidate (AI-TCRm17-6) demonstrated a 10 nM estimated Kd for the WT1 target peptide. AI-TCRm17-6 was fused to an antibody Fc domain to create a drug-like molecule and showed negligible binding to off-target peptides. The final molecule bound to HLA-A2 WT1 positive cancer cells measured by simple flow cytometry. Using the same platform, we also generated de novo binders against HLA-E presented cancer peptides, challenging targets for TCR or TCRm based therapeutics.Conclusions This platform enables rapid, target-specific generation and functional validation of pMHC binders, offering a scalable solution for tackling hard-to-drug intracellular cancer targets, including those generated by cryptic cancer peptides. Our approach opens the door to faster development of next-generation cancer therapeutics without library screening, and it can be broadly applied to other protein targets.",
  "authors": [
    {
      "affiliations": [
        "Vcreate Inc. Menlo Park, CA, USA"
      ],
      "name": "Ethan Fast"
    },
    {
      "affiliations": [
        "Vcreate Inc. Menlo Park, CA, USA"
      ],
      "name": "Manjima Dhar"
    },
    {
      "affiliations": [
        "Vcreate Inc. Menlo Park, CA, USA"
      ],
      "name": "Millicent Ku"
    },
    {
      "affiliations": [
        "Vcreate Inc. Menlo Park, CA, USA"
      ],
      "name": "Binbin Chen"
    }
  ],
  "title": "1002 Rapid design and validation of highly specific binders against MHC-peptide targets",
  "uid": "45e137b9-008f-5c9d-9cbb-3cd8adaac8d0"
}