bionemo 0.3.1.post1

Python client for the BioNeMo Service

BioNeMo Service Python Client

The BioNeMo Service Python Client provides a python API into the BioNeMo inference service. The BioNeMo service uses NVIDIA Triton inference infrastructure to deploy ML models relevant to biology and chemistry applications. This client exposes easy-to-use python functions that call directly to the inference service, allowing users to use state-of-the-art models with little setup.

Learn more and apply for access to BioNeMo here.

Example: Generate novel protein sequences and perform folding

from bionemo.api import BionemoClient
from time import sleep

# Create a client instance
api = BionemoClient("APIKEY")

# Generate novel proteins
novel_proteins = api.protgpt2_sync(max_length=200, num_return_sequences=10)

# Request folding of novel proteins in parallel
submitted_requests = []
for protein in novel_proteins["generated_sequences"]:
    request_id = api.openfold_async(protein)
    submitted_requests.append(request_id)

# Wait for results, write to disk
while len(submitted_requests):
    sleep(10)
    for request_id in submitted_requests:
        if api.fetch_task_status(request_id) == "DONE":
            folded_protein = api.fetch_result(request_id)
            with open(str(request_id) + ".pdb", "w") as f:
                f.write(folded_protein)
            submitted_requests.remove(request_id)

Currently, the following models and workflows are available for inference:

Protein Structure Prediction:

• AlphaFold-2
• OpenFold
• ESMFold

Protein Embedding:

• ESM-2
• ESM-1nv

Docking:

• Diffdock

Protein Sequence Generation:

• ProtGPT2

Small Molecule Embedding and Generation:

• MegaMolBart
• MoFlow
• MolMIM

Multiple Sequence Alignment