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

Biology

BioNeMo OpenFold3

NVIDIA NIM for OpenFold3 diffusion-based structure prediction.

Model details

Developed by
NVIDIA
Model family
BioNeMo
Use case
biology
Variant
NIM
Hardware
H100
License
NVIDIA AI Foundation Models Community License Agreement

OpenFold3 is an all-atom biomolecular complex structure prediction model from the OpenFold Consortium and the AlQuraishi Laboratory, a PyTorch implementation of the JAX-based AlphaFold3 model that extends structure prediction to complete biomolecular complexes including proteins, DNA, RNA, and small molecule ligands. It takes protein, DNA, RNA, and ligand sequences as input — plus optional MSAs and, for proteins, structural templates in CIF format — and outputs the all-atom 3D structure of the complex.

Call BioNeMo OpenFold3 on Baseten using the NIM structure prediction 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    "/biology/openfold/openfold3/predict"
9)
10
11sequence = "LSDEDFKAVFGMTRSAFANLPLWKQQNLKKEKGLF"
12minimal_msa = f">query\n{sequence}\n"
13
14payload = {
15    "inputs": [
16        {
17            "input_id": "villin_hp35_test",
18            "molecules": [
19                {
20                    "type": "protein",
21                    "sequence": sequence,
22                    "msa": {
23                        "main": {
24                            "a3m": {
25                                "alignment": minimal_msa,
26                                "format": "a3m",
27                            }
28                        }
29                    },
30                }
31            ],
32            "diffusion_samples": 1,
33            "output_format": "pdb",
34        }
35    ]
36}
37
38response = requests.post(
39    BASETEN_URL,
40    headers={
41        "Authorization": f"Bearer {BASETEN_API_KEY}",
42        "Content-Type": "application/json",
43    },
44    json=payload,
45    timeout=300,
46)
47response.raise_for_status()
48result = response.json()
49print(result)
50

JSON output

1{
2    "outputs": [
3        {
4            "input_id": "villin_hp35_test",
5            "structures_with_scores": [
6                {
7                    "structure": "<PDB text>",
8                    "format": "pdb",
9                    "confidence_score": 0.85
10                }
11            ]
12        }
13    ]
14}