NVIDIA BioNeMo pretrained model weights: ESM-2, Evo-2, Geneformer

Necessary Disclaimers and Legal

The user is responsible for reviewing and complying with the license requirements of the models, notebooks, and data referenced in this documentation.

Users are responsible for the costs associated with analyzing notebooks and its storage in their project spaces.

Instance type availability and pricing are subject to the contract between the user or the user’s organization and DNAnexus.

Citations for foundation models

If you use any of the following models in your work, please cite the corresponding publications:

• ESM-2 model:

  • Hugging face: https://huggingface.co/facebook/esm2_t33_650M_UR50D

  • Paper: https://www.biorxiv.org/content/10.1101/2022.07.20.500902v2

• Evo-2 model:

  • Hugging face: https://huggingface.co/arcinstitute/evo2_7b

  • Paper: https://www.biorxiv.org/content/10.1101/2025.02.18.638918v1

• Geneformer:

  • Hugging face: https://huggingface.co/ctheodoris/Geneformer

  • Paper: https://www.nature.com/articles/s41586-023-06139-9

Overview

Foundation models are transforming drug discovery, but deploying them requires specialized infrastructure that most pharma IT organizations cannot provide: large GPU instances, pre-downloaded model weights, and compatible software environments. This accelerator makes foundation models accessible within DNAnexus's secure, compliant environment. On DNAnexus, you can access pre-download model weights of three models: ESM-2, Evo-2 and Geneformer.

Summary of Models to Access

Model Family	Domain Focus	Checkpoint
ESM-2	Proteins / Amino Acids	esm2_t33_650M_UR50D
Evo 2	Genomes / DNA	evo-2-7b
Geneformer	Transcriptomics (scRNA-seq)	Geneformer

These models are packaged into a unified notebook snapshot image "NVIDIA BioNeMo Lite", built on the existing "ML" feature container image with HuggingFace packages and foundation model architectures pre-installed. Three notebooks are provided to showcase and verify the snapshot for each model respectively.

Using foundation models on DNAnexus

Copying demo notebooks and model weights into a project

To explore the demo analyses, copy the notebooks and model weights from the public project into your own project space:

1. Create a project for your analyses, billed to your own organization. Tutorials on project setup can be found here.

2. Navigate to the Resources tab, locate the project titled "Public Datasets AWS US (East)," and select the "Foundation_models" folder.

1. Select the three notebooks (.ipynb) and models folder.

1. Click "Copy" in the top-right menu and select the project created in Step 1.

2. Return to your project space to begin exploring the notebooks.

To run the notebooks, refer to the JupyterLab section of the Academy documentation, which includes instructions for the standard environment. We recommend using the prepared snapshot-NVIDIA_BioNeMo_Lite-jupyterlab.tar.gz, available in the Notebook_snapshot folder of the public project.

Running the notebooks

Three notebooks are developed for JupyterLab with Python, R, Stata, ML, Image Processing (version 2.11). When launching JupyterLab, select the ML feature and the snapshot snapshot-NVIDIA_BioNeMo_Lite-jupyterlab.tar.gz.

Notebook	Instance
ESM2_verification_test_base-env.ipynb	mem2_ssd1_gpu_x16
Evo2_verification_test_base-env.ipynb	mem2_ssd2_gpu1_v2_x8
Geneformer_verification_test_base-env.ipynb	mem2_ssd2_gpu1_v2_x8

Steps for launching notebooks

1. Choose a snapshot from Notebook Snapshot folder. Note: the path for this snapshot should be in your own project.

1. Choose an instance for the notebook.

1. Set the job name and output destination, then launch the analysis. Note: please change the output location to your project.