CD38 is the dominant NAD-consuming enzyme in aged tissues and its activity rises ~2-3 fold with age. Existing small-molecule inhibitors (78c, daratumumab epitope) lack tissue-selective delivery. This plan designs a de novo protein binder against the catalytic cleft of human CD38 (UniProt P28907, PDB 2I65 / 4F45) using the RFdiffusion → ProteinMPNN → AF2-Multimer validation stack. Hotspot conditioning targets the catalytic triad-adjacent residues C119, D155, D156, D179, D219, Q226. Success criteria: (1) BindCraft or RFdiff-MPNN generates ≥5 designs with AF2-Multimer ipTM > 0.75 and pAE_interaction < 10 Å; (2) at least one design is predicted to occupy the NAD-binding pocket by geometric clash with NAD substrate; (3) Foldseek TM-score < 0.5 vs all PDB entries confirms novelty; (4) ESM-2 log-likelihood perplexity < 10 per residue. Planned wet-lab route: gene synthesis → E. coli periplasmic expression → SPR binding assay against ectodomain CD38 (residues 43–300) → NAD-glycohydrolase inhibition IC50 assay (fluorometric εNAD substrate). Open question: selectivity vs SARM1 (TIR-domain NAD-consumer in axonal degeneration) and tissue-specific CD38 isoforms.
Details
- disease
- aging, age-related metabolic decline, nad+ depletion
- target_ref
- UniProt:P28907 / PDB:2I65
- primary_endpoint
- Hotspot JSON contains ≥5 residues with Cα-Cα distance < 5 Å to at least one of E226 or C119
- identification_strategy
- in_silico_KO
Raw fields (3)
- assay_spec
Extract CD38 catalytic-site hotspot geometry from PDB 2GVJ (NAMPT proxy run already done; pivot to CD38 PDB 1YH3), compute inter-residue contact map for residues E226/C119/D155/D156/K218/Q226, deposit hotspot JSON, then run BindCraft binder design against the NAD+-binding pocket as a 65-85 aa helical binder; cd38_hotspot_geometry.json — per-residue centroid coordinates, Cα distances, and neighbor list for E226/C119/D155/D156/K218/Q226 from 1YH3 chain A; gates BindCraft conditioning; cd38_bindcraft_recipe.json — BindCraft input config: target PDB path, hotspot residue list, binder length range 65-85 aa, n_designs=20, secondary_structure_bias=helical; {'kind': 'claim', 'text_template': 'Hotspot pocket volume at CD38 NAD+-binding site (1YH3) supports a 65-85 aa helical binder with ≥5 direct contact residues to catalytic dyad E226/C119'}- hypothesis
A de novo protein binder designed to occupy the CD38 NAD-glycohydrolase catalytic cleft (hotspots C119, D155, D156, D179, D219, Q226 from PDB 2I65/4F45) will competitively inhibit CD38-mediated NAD+ consumption in aged tissues, raising intracellular NAD+ levels and improving mitochondrial function, with selectivity over structurally-distinct SARM1 and PARP1 NAD-consumers.
- kill_criteria
Abort or revise if: 1YH3 has two chains (A/B homodimer); must isolate chain A and strip ligand atoms before hotspot extraction; E226 is the catalytic glutamate — insertion code or numbering offset in 1YH3 vs UniProt P28907 must be verified before residue selection; BindCraft job itself is compute-heavy (>30 min); this plan gates only the recipe + hotspot artifact, not the full BindCraft run