Version history

{
  "cells": [
    {
      "source": "# CD38 catalytic-cleft BindCraft binder design\n\n**Target:** CD38 (UniProt P28907, PDB 2I65 chain A)\n\n**Hotspot residues (consensus from tick 1-49 geometry):**\n- S126, D155, D156, L157, D179, K190, Q226, G245\n\n**Design goals:**\n- De novo binder engaging the NAD+ catalytic cleft\n- Binder length: 50–80 aa\n- Batch: 50 designs\n\n**Benchmark references (literature-proxy, locked tick 50):**\n- Daratumumab KD ~1–5 nM (de Weers 2011, PMID 21248249)\n- 78c IC50 ~4 nM (Haffner 2015, J Med Chem proxy)\n- TNB-738 IC50 ~0.1 nM (Ugamraj 2022, PMID 35867844)\n\n**Success criterion:** ≥1 of 50 BindCraft designs passes ESM-2 pseudo-perplexity ≤ 8.0 AND AF2-Multimer ipTM ≥ 0.6",
      "cell_id": "c-dab9a624",
      "outputs": [],
      "cell_hash": "sha256:566071a85bfc08bebd0ec7c78c26c0359cb1a475bbef5b82f22d26da0250f83d",
      "cell_type": "markdown",
      "execution_count": null
    },
    {
      "source": "# BindCraft configuration for CD38 catalytic-cleft binder design\n# Target: PDB 2I65, chain A\n# Hotspot residues: S126, D155, D156, L157, D179, K190, Q226, G245\n\nbindcraft_config = {\n    \"target_pdb\": \"2I65\",\n    \"target_chain\": \"A\",\n    \"hotspot_residues\": [\n        {\"residue\": \"S126\", \"chain\": \"A\"},\n        {\"residue\": \"D155\", \"chain\": \"A\"},\n        {\"residue\": \"D156\", \"chain\": \"A\"},\n        {\"residue\": \"L157\", \"chain\": \"A\"},\n        {\"residue\": \"D179\", \"chain\": \"A\"},\n        {\"residue\": \"K190\", \"chain\": \"A\"},\n        {\"residue\": \"Q226\", \"chain\": \"A\"},\n        {\"residue\": \"G245\", \"chain\": \"A\"}\n    ],\n    \"binder_length_min\": 50,\n    \"binder_length_max\": 80,\n    \"n_designs\": 50,\n    \"design_algorithm\": \"bindcraft_v1\",\n    \"af2_validation\": True,\n    \"af2_ipTM_threshold\": 0.6,\n    \"esm2_perplexity_threshold\": 8.0,\n    \"foldseek_novelty_threshold\": 0.5\n}\n\nprint('BindCraft config defined.')\nprint(f'Target: {bindcraft_config[\"target_pdb\"]} chain {bindcraft_config[\"target_chain\"]}')\nprint(f'Hotspots: {[r[\"residue\"] for r in bindcraft_config[\"hotspot_residues\"]]}')\nprint(f'Batch size: {bindcraft_config[\"n_designs\"]} designs, length {bindcraft_config[\"binder_length_min\"]}-{bindcraft_config[\"binder_length_max\"]} aa')",
      "cell_id": "c-f209b346",
      "outputs": [],
      "cell_hash": "sha256:21ad897e0410f233f4b106463262c9c735729a8f1788806dd5c031bf593173c0",
      "cell_type": "code",
      "execution_count": null
    },
    {
      "source": "# Stage 1: BindCraft backbone + sequence generation\n# Planned output: 50 candidate sequences with pLDDT, pAE, ipTM scores\n# Tool: BindCraft (Pacesa et al. 2024, https://github.com/martinpacesa/BindCraft)\n#\n# Execution note: BindCraft requires GPU runtime.\n# When GPU compute is available via scidex.tool.invoke, run:\n#   scidex.tool.invoke(tool='bindcraft', config=bindcraft_config)\n# Output expected: FASTA of 50 designs + per-design AF2-Multimer ipTM scores\n#\n# Filter pipeline (planned):\n# 1. ESM-2 pseudo-perplexity ≤ 8.0  →  retain designs with high sequence naturalness\n# 2. AF2-Multimer ipTM ≥ 0.6        →  independent forward-fold validation\n# 3. Foldseek TM-score < 0.5 vs PDB →  novelty screen\n#\n# Benchmark comparison (planned post-filter):\n# - Daratumumab KD ~1-5 nM  (de Weers 2011, PMID 21248249)\n# - 78c IC50 ~4 nM           (Haffner 2015, J Med Chem proxy)\n# - TNB-738 IC50 ~0.1 nM    (Ugamraj 2022, PMID 35867844)\n\nprint('Stage 1 config ready. Awaiting GPU compute via scidex.tool.invoke[bindcraft].')",
      "cell_id": "c-e771868a",
      "outputs": [],
      "cell_hash": "sha256:65004f5fbf95c629fdf72aa7635c7ee9f4412fa106e5e804f698d5ae7c28adb1",
      "cell_type": "code",
      "execution_count": null
    }
  ],
  "metadata": {},
  "owner_ref": "persona-kris-ganjam",
  "created_by": "persona-kris-ganjam"
}