{
  "version": 3,
  "created_at": "2026-07-09T15:37:28.541Z",
  "task_summary": "Sequence optimization and nanoparticle scaffold fusion for pMHC therapeutics",
  "agents": [],
  "phases": [
    {
      "name": "Plan",
      "delegations": [
        {
          "steps": [
            {
              "title": "Optimize pMHC backbones with LigandMPNN",
              "description": "Use LigandMPNN inverse-folding with the co-folded DR7 groove and seated peptides as context to redesign the MHC-binding interface for improved presentation kinetics and stability. Input: 3 pMHC-II:peptide structures (dairy/wheat/soy from the prior co-folds). Output: 5 LigandMPNN designs per complex, ranked by pAE/pLDDT and burial of critical DR-interaction residues. Save optimized sequences and designs table."
            },
            {
              "title": "Validate optimized sequences with ESMFold2",
              "description": "Co-fold the top LigandMPNN designs back against their MHC context on GPU to verify groove-peptide geometry is preserved and ipTM remains >0.85. Output: PDB structures and metrics for lead designs. Flags if redesign disrupts the groove."
            },
            {
              "title": "Design nanoparticle scaffold fusion architecture",
              "description": "Design the linker and attachment geometry for the nanoparticle format: (a) the peptide-linker-DRβ tether (currently GGGGSGGGGSGGS, 13 aa); (b) the C-terminal His6–GGGSC free-thiol handle for maleimide-iron-oxide NP conjugation. Specify the surface-attachment chemistry, stoichiometry (multivalent display model), and solvation profile. Output: fusion-architecture spec and a draft linker-optimization design if flexibility gains are needed."
            },
            {
              "title": "Model the NP-bound conformation",
              "description": "Build a minimal coarse-grained model of the NP-pMHC-II assembly: iron-oxide NP core (10–30 nm, modeled as a sphere), maleimide-bound pMHC copies (3–5 copies per NP, density-optimized for TCR accessibility without steric clash). Output: a visualization/spec of the multivalent display geometry and the effective epitope density on the NP surface."
            },
            {
              "title": "Write optimization report and updated spec sheets",
              "description": "Summarize sequence designs, fold validation, and NP fusion architecture. Produce updated construct FASTAs (with optimized sequences), a nanoparticle fusion spec with attachment details, and side-by-side comparison of original vs. optimized backbone fold metrics."
            }
          ]
        }
      ],
      "id": "phase-0"
    }
  ],
  "desired_outputs": [
    "Optimized pMHC sequences (FASTA)",
    "LigandMPNN design metrics table",
    "Nanoparticle fusion-architecture spec",
    "Updated construct specification",
    "NP assembly model and visualization"
  ],
  "feasibility": {
    "confidence": "high",
    "rationale": "LigandMPNN is a standard inverse-folding tool that works well with structural context (PDB templates). The co-folded complexes already provide the MHC groove context. NP scaffold design is straightforward geometric/chemical engineering (linker length, stoichiometry). ESMFold2 revalidation is routine. All pieces are well-defined."
  }
}