Drug Discovery Pipeline

药物发现流程

Screen drug candidates end-to-end using three BioNeMo NIMs in sequence:

Step 1: GenMol    →  Step 2: DiffDock  →  Step 3: Boltz2
(Generate mols)      (Dock to target)      (Predict affinity)

依次使用三个BioNeMo NIMs端到端筛选候选药物：

Step 1: GenMol    →  Step 2: DiffDock  →  Step 3: Boltz2
(Generate mols)      (Dock to target)      (Predict affinity)

Overview

概述

This pipeline is used for:

De novo hit discovery: generate drug-like molecules and screen them against a target
Lead optimization: start from a known scaffold and generate improved analogs, then dock and score
Virtual screening: dock a library of candidates and filter by docking confidence + affinity

本流程适用于：

全新命中发现：生成类药物分子并针对靶点进行筛选
先导化合物优化：从已知骨架出发生成改良类似物，然后进行对接和评分
虚拟筛选：对接候选化合物库，并根据对接置信度和亲和力进行筛选

Before you start

开始之前

Confirm with the user:

Target protein: PDB file or sequence of the binding target
Starting point: de novo (no scaffold) or scaffold decoration (known core)?
Scoring: drug-likeness (QED) or lipophilicity (LogP)?
API mode: hosted or local Docker?

For local Docker, do not assume all NIMs are running on

localhost:8000

at the same time. Either run one container at a time and hand files/results between steps, or start each NIM on a distinct host port and set the per-step URLs.

请与用户确认以下信息：

靶点蛋白质：结合靶点的PDB文件或序列
起始方式：全新生成（无骨架）还是骨架修饰（已知核心）？
评分标准：类药性（QED）还是亲脂性（LogP）？
API模式：托管式还是本地Docker？

对于本地Docker模式，请勿假设所有NIMs同时运行在

localhost:8000

。可以每次运行一个容器并在步骤间传递文件/结果，或者在不同的主机端口启动每个NIM并设置各步骤的URL。

Step 1: Generate molecules with GenMol

步骤1：使用GenMol生成分子

GenMol requires SAFE notation input (not raw SMILES). Use the

safe-mol

package.

python

import requests, json, os
import safe as sf                          # pip install safe-mol
from pathlib import Path

NGC_API_KEY = os.getenv("NGC_API_KEY")
HOSTED = True

if HOSTED:
    genmol_url = "https://health.api.nvidia.com/v1/biology/nvidia/genmol/generate"
    headers = {"Content-Type": "application/json",
               "Authorization": f"Bearer {NGC_API_KEY}"}
else:
    genmol_url = "http://localhost:8000/generate"
    headers = {"Content-Type": "application/json"}

GenMol需要SAFE符号输入（而非原始SMILES）。请使用

safe-mol

包。

python

import requests, json, os
import safe as sf                          # pip install safe-mol
from pathlib import Path

NGC_API_KEY = os.getenv("NGC_API_KEY")
HOSTED = True

if HOSTED:
    genmol_url = "https://health.api.nvidia.com/v1/biology/nvidia/genmol/generate"
    headers = {"Content-Type": "application/json",
               "Authorization": f"Bearer {NGC_API_KEY}"}
else:
    genmol_url = "http://localhost:8000/generate"
    headers = {"Content-Type": "application/json"}

De novo generation (no scaffold):

safe_input = "[*{20-30}]"

Scaffold decoration (known core):

scaffold_smiles = "c1ccccc1"

safe_input = sf.encode(scaffold_smiles) + ".[*{5-10}]"

payload = { "smiles": safe_input, # field is named 'smiles' but takes SAFE notation "num_molecules": 30, # request more to compensate for post-generation filtering "scoring": "QED", # QED or LogP "unique": True, "temperature": "1.0", # NOTE: must be string, not float "noise": "1.0", # NOTE: must be string, not float }

r = requests.post(genmol_url, headers=headers, json=payload) r.raise_for_status() molecules = r.json()["molecules"] molecules_sorted = sorted(molecules, key=lambda x: x["score"], reverse=True) top_20 = molecules_sorted[:20]

print(f"Generated {len(molecules)} valid molecules (requested 30)") print("Top 5 by QED score:") for m in top_20[:5]: print(f" {m['smiles'][:50]} score={m['score']:.4f}")

---

payload = { "smiles": safe_input, # field is named 'smiles' but takes SAFE notation "num_molecules": 30, # request more to compensate for post-generation filtering "scoring": "QED", # QED or LogP "unique": True, "temperature": "1.0", # NOTE: must be string, not float "noise": "1.0", # NOTE: must be string, not float }

r = requests.post(genmol_url, headers=headers, json=payload) r.raise_for_status() molecules = r.json()["molecules"] molecules_sorted = sorted(molecules, key=lambda x: x["score"], reverse=True) top_20 = molecules_sorted[:20]

print(f"Generated {len(molecules)} valid molecules (requested 30)") print("Top 5 by QED score:") for m in top_20[:5]: print(f" {m['smiles'][:50]} score={m['score']:.4f}")

---

Step 2: Dock molecules with DiffDock

步骤2：使用DiffDock进行分子对接

Prepare the protein and dock each candidate:

python

undefined

准备蛋白质并对接每个候选分子：

python

undefined

Load protein (ATOM records only)

receptor_pdb_raw = Path("target.pdb").read_text() receptor_pdb = "\n".join(line for line in receptor_pdb_raw.splitlines() if line.startswith("ATOM"))

if HOSTED: diffdock_url = "https://health.api.nvidia.com/v1/biology/mit/diffdock" else: diffdock_url = "http://localhost:8000/molecular-docking/diffdock/generate"

docking_results = []

for i, mol in enumerate(top_20): payload = { "protein": receptor_pdb, "ligand": mol["smiles"], "ligand_file_type": "txt", # "txt" for SMILES input "num_poses": 5, "time_divisions": 20, "steps": 18, "save_trajectory": False, }

r = requests.post(diffdock_url, headers=headers, json=payload)
r.raise_for_status()
result = r.json()

best_conf = result["position_confidence"][0]  # rank 1 pose
best_pose = result["ligand_positions"][0]

docking_results.append({
    "smiles": mol["smiles"],
    "qed_score": mol["score"],
    "docking_confidence": best_conf,
    "best_pose_sdf": best_pose,
})
print(f"  Mol {i+1:2d}: QED={mol['score']:.3f}  docking_conf={best_conf:.4f}")

receptor_pdb_raw = Path("target.pdb").read_text() receptor_pdb = "\n".join(line for line in receptor_pdb_raw.splitlines() if line.startswith("ATOM"))

if HOSTED: diffdock_url = "https://health.api.nvidia.com/v1/biology/mit/diffdock" else: diffdock_url = "http://localhost:8000/molecular-docking/diffdock/generate"

docking_results = []

for i, mol in enumerate(top_20): payload = { "protein": receptor_pdb, "ligand": mol["smiles"], "ligand_file_type": "txt", # "txt" for SMILES input "num_poses": 5, "time_divisions": 20, "steps": 18, "save_trajectory": False, }

r = requests.post(diffdock_url, headers=headers, json=payload)
r.raise_for_status()
result = r.json()

best_conf = result["position_confidence"][0]  # rank 1 pose
best_pose = result["ligand_positions"][0]

docking_results.append({
    "smiles": mol["smiles"],
    "qed_score": mol["score"],
    "docking_confidence": best_conf,
    "best_pose_sdf": best_pose,
})
print(f"  Mol {i+1:2d}: QED={mol['score']:.3f}  docking_conf={best_conf:.4f}")

Rank by docking confidence

docking_results.sort(key=lambda x: x["docking_confidence"], reverse=True) print(f"\nTop 3 by docking confidence:") for d in docking_results[:3]: print(f" {d['smiles'][:50]} conf={d['docking_confidence']:.4f}")

---

docking_results.sort(key=lambda x: x["docking_confidence"], reverse=True) print(f"\nTop 3 by docking confidence:") for d in docking_results[:3]: print(f" {d['smiles'][:50]} conf={d['docking_confidence']:.4f}")

---

Step 3: Predict binding affinity with Boltz2

步骤3：使用Boltz2预测结合亲和力

For the top docking candidates, predict structure-based binding affinity:

python

if HOSTED:
    boltz_url = "https://health.api.nvidia.com/v1/biology/mit/boltz2/predict"
else:
    boltz_url = "http://localhost:8000/biology/mit/boltz2/predict"

针对排名靠前的对接候选分子，基于结构预测结合亲和力：

python

if HOSTED:
    boltz_url = "https://health.api.nvidia.com/v1/biology/mit/boltz2/predict"
else:
    boltz_url = "http://localhost:8000/biology/mit/boltz2/predict"

Use the target protein sequence (not PDB)

target_sequence = "<YOUR_TARGET_PROTEIN_SEQUENCE>"

affinity_results = [] for d in docking_results[:5]: # score top 5 docking hits payload = { "polymers": [ {"id": "A", "molecule_type": "protein", "sequence": target_sequence} ], "ligands": [ {"id": "L1", "smiles": d["smiles"], "predict_affinity": True} ], "recycling_steps": 3, "sampling_steps": 50, "diffusion_samples": 1, "output_format": "mmcif", }

r = requests.post(boltz_url, headers=headers, json=payload)
r.raise_for_status()
result = r.json()

aff = result["affinities"]["L1"]
pic50 = aff["affinity_pic50"][0]
prob_binding = aff["affinity_probability_binary"][0]

affinity_results.append({
    **d,
    "pic50": pic50,
    "probability_binding": prob_binding,
})
print(f"  {d['smiles'][:40]}  pIC50={pic50:.2f}  P(bind)={prob_binding:.3f}")

target_sequence = "<YOUR_TARGET_PROTEIN_SEQUENCE>"

affinity_results = [] for d in docking_results[:5]: # score top 5 docking hits payload = { "polymers": [ {"id": "A", "molecule_type": "protein", "sequence": target_sequence} ], "ligands": [ {"id": "L1", "smiles": d["smiles"], "predict_affinity": True} ], "recycling_steps": 3, "sampling_steps": 50, "diffusion_samples": 1, "output_format": "mmcif", }

r = requests.post(boltz_url, headers=headers, json=payload)
r.raise_for_status()
result = r.json()

aff = result["affinities"]["L1"]
pic50 = aff["affinity_pic50"][0]
prob_binding = aff["affinity_probability_binary"][0]

affinity_results.append({
    **d,
    "pic50": pic50,
    "probability_binding": prob_binding,
})
print(f"  {d['smiles'][:40]}  pIC50={pic50:.2f}  P(bind)={prob_binding:.3f}")

Final ranking by pIC50

affinity_results.sort(key=lambda x: x["pic50"], reverse=True)

---

affinity_results.sort(key=lambda x: x["pic50"], reverse=True)

---

Interpreting results

结果解读

GenMol QED score: 0–1; >0.5 is drug-like
DiffDock confidence: higher = more reliable binding pose prediction
Boltz2 pIC50: predicted -log10(IC50); >6 = sub-micromolar, >8 = very potent
P(bind): probability of binary binding; >0.7 = likely binder

GenMol QED评分：0–1；大于0.5表示具有类药性
DiffDock置信度：数值越高，结合构象预测越可靠
Boltz2 pIC50：预测的-log10(IC50)；大于6表示亚微摩尔级活性，大于8表示活性极强
P(bind)：二元结合概率；大于0.7表示很可能是结合剂

Quick reference — skill dependencies

快速参考 — 技能依赖

Step	Skill	Key endpoint
Molecule generation	`genmol-nim`	`/biology/nvidia/genmol/generate`
Docking	`diffdock-nim`	`/molecular-docking/diffdock/generate`
Affinity prediction	`boltz2-nim`	`/biology/mit/boltz2/predict`

步骤	技能	关键端点
分子生成	`genmol-nim`	`/biology/nvidia/genmol/generate`
分子对接	`diffdock-nim`	`/molecular-docking/diffdock/generate`
亲和力预测	`boltz2-nim`	`/biology/mit/boltz2/predict`

drug-discovery-pipeline

Original

Translation

Drug Discovery Pipeline

药物发现流程

Overview

概述

Before you start

开始之前

Step 1: Generate molecules with GenMol

步骤1：使用GenMol生成分子

De novo generation (no scaffold):

De novo generation (no scaffold):

Scaffold decoration (known core):

Scaffold decoration (known core):

scaffold_smiles = "c1ccccc1"

scaffold_smiles = "c1ccccc1"

safe_input = sf.encode(scaffold_smiles) + ".[*{5-10}]"

safe_input = sf.encode(scaffold_smiles) + ".[*{5-10}]"

Step 2: Dock molecules with DiffDock

步骤2：使用DiffDock进行分子对接

Load protein (ATOM records only)

Load protein (ATOM records only)

Rank by docking confidence

Rank by docking confidence

Step 3: Predict binding affinity with Boltz2

步骤3：使用Boltz2预测结合亲和力

Use the target protein sequence (not PDB)

Use the target protein sequence (not PDB)

Final ranking by pIC50

Final ranking by pIC50

Interpreting results

结果解读

Quick reference — skill dependencies

快速参考 — 技能依赖