Try the new DeepSeek V4 Flash today. Frontier intelligence at a fraction of the cost. Here

Biology

BioNeMo GenMol

NVIDIA NIM for GenMol molecule generation and scaffold decoration.

Model details

Developed by
NVIDIA
Model family
BioNeMo
Use case
biology
Variant
NIM
Hardware
A10G
License
NVIDIA AI Foundation Models Community License Agreement

GenMol is a masked diffusion model trained on Sequential Attachment-based Fragment Embedding (SAFE) representations for fragment-based molecule generation, serving as a generalist tool across drug discovery tasks. It takes a SAFE-format molecular template as input — letting users specify fixed fragments to keep unchanged, attachment positions for new fragments, and desired fragment lengths — and generates novel drug-like molecules tailored to specific chemical properties through an iterative masking-unmasking process.

Call BioNeMo GenMol on Baseten using the NIM generation route.

Input

1import os
2import requests
3
4MODEL_ID = os.environ["BASETEN_MODEL_ID"]
5BASETEN_API_KEY = os.environ["BASETEN_API_KEY"]
6BASETEN_URL = (
7    f"https://model-{MODEL_ID}.api.baseten.co/environments/production/sync"
8    "/generate"
9)
10
11payload = {
12    "smiles": "[C@H]1O[C@@H](CO)[C@H](O)[C@@H]1O.[*{15-15}]",
13    "num_molecules": 5,
14    "temperature": 1.0,
15    "noise": 1.0,
16    "scoring": "QED",
17}
18
19response = requests.post(
20    BASETEN_URL,
21    headers={
22        "Authorization": f"Bearer {BASETEN_API_KEY}",
23        "Content-Type": "application/json",
24    },
25    json=payload,
26    timeout=300,
27)
28response.raise_for_status()
29result = response.json()
30print(result)
31

JSON output

1{
2    "status": "success",
3    "molecules": [
4        {
5            "smiles": "OC[C@@H]1O[CH][C@@H](O)[C@H]1O",
6            "score": 0.397
7        }
8    ]
9}