CD38 is the dominant NAD+-consuming ectoenzyme in aged tissues; its activity rises ~2–3-fold with age, driving systemic NAD+ depletion. Structural data (PDB 2I65, 2O3U) reveals a deep catalytic cleft harboring the cADPR-synthesis and hydrolase active site. Geometric hotspot profiling (completed tick-0) confirmed 8 catalytic residues with sub-1.4 Å inter-atomic distances, forming a tight hydrophilic pocket. This plan specifies: (1) BindCraft-conditioned binder generation targeting T102/D155/D156/D179 as primary anchor hotspots; (2) ProteinMPNN sequence design on RFdiffusion-generated backbones conditioned on the catalytic cleft; (3) AF2-Multimer independent forward-fold validation (pLDDT > 80, ipTM > 0.7, PAE < 10 Å at interface); (4) ESM-2 perplexity filtering (pseudo-log-likelihood > -1.5 per residue); (5) Foldseek novelty screen against PDB to confirm < 30% TM-score to any known binder; (6) Rosetta ddG estimation for top-5 candidates. Success criterion: ≥1 candidate with predicted KD < 100 nM by AF3-Multimer pTM proxy, no disulfide liabilities, GRAVY < 0.5, pI 5–8, and Foldseek TM-score < 0.5 to all PDB entries.
Details
- disease
- aging, nad+ depletion, metabolic decline, neurodegeneration
- target_ref
- UniProt:P28907 / PDB:2I65
- primary_endpoint
- ≥1 design passes all three gates: ESM-2 perplexity < 10, BindCraft ipTM > 0.65, AF2-Multimer ipTM > 0.70
- identification_strategy
- in_silico_KO
Raw fields (3)
- assay_spec
Hotspot-anchored BindCraft binder design against CD38 NAD-hydrolase active site: fetch PDB 2I65 ectodomain chain A, define hotspot patch {E226, D155, C119, T102, W189, D179}, run BindCraft (50 trajectories, 60–80 aa binder, partial-diffusion refinement), ESM-2 perplexity filter (ppl < 10), AF2-Multimer forward-fold validation (ipTM > 0.7, pLDDT > 80); BindCraft trajectory score scatter: ipTM vs ESM-2 perplexity for all 50 designs, pass/fail colored by threshold — PLANNED; AF2-Multimer complex ribbon of top-ranked binder on CD38 ectodomain, colored by pLDDT — PLANNED; Ranked table: binder id, length, ESM-2 ppl, BindCraft ipTM, AF2-Multimer ipTM, pLDDT, hotspot contact count — PLANNED; {'kind': 'claim', 'text_template': 'Binder candidate KG-CD38-001 achieves AF2-Multimer ipTM ≥ 0.72 with ≥4 hotspot contacts at the NAD-hydrolase pocket — PLANNED'}; FASTA + PDB of top-3 designs for wet-lab triage — PLANNED- hypothesis
A de novo peptide binder designed against the CD38 catalytic cleft (anchored to hotspot residues T102, C119, D155, D156, L157, D179, S181, Q226 from PDB 2I65) will competitively inhibit CD38 NAD+ hydrolase activity with predicted IC50 < 1 µM in silico, surpassing the potency of known small-molecule inhibitors (e.g., 78c, luteolinidin) while offering superior selectivity by avoiding the CD31-binding ectodomain surface.
- kill_criteria
Abort or revise if: 2I65 has a homodimer biological assembly — must extract monomer chain A only to avoid BindCraft targeting the dimer interface instead of the catalytic pocket; CD38 active-site pocket is partially occluded; BindCraft may need hotspot_radius tuning if initial trajectories dock to surface rather than pocket; ESM-2 perplexity threshold of 10 is conservative — if all 50 designs exceed it, relax to ppl < 15 and flag for manual review