Prerequisites and Support Matrix for MSA Search NIM | NVIDIA BioNeMo NIM Documentation

Sequence Databases and Search Types

The MSA Search NIM includes pre-indexed databases optimized for GPU Server operation:

Database Name	Use Case	Paired Search	Template Search	Source
uniref30_2302	MSA and profile generation	Yes	Required	ColabFold Databases
colabfold_envdb_202108	MSA expansion	No	No	ColabFold Databases
pdb70_220313	Template search (70% cluster)	No	Yes	ColabFold Databases
pdb100_230517	Template search (100% cluster)	No	Yes	ColabFold Databases
pdb_20251028	mmCIF structure retrieval	No	Yes	RCSB PDB

Total size: 1660 GB

Paired MSA search requires databases with taxonomy information. Only uniref30_2302 supports paired search among the default databases.

Template search requires uniref30_2302 for profile generation plus at least one template database (PDB70 or PDB100). The profile generated from Uniref30 is used to search the template databases for structural homologs.

To reduce storage requirements, you can use task-specific database profiles that load only the databases needed for your use case. Refer to Task-Specific Profiles and Custom Databases for details.

Supported Hardware

Minimum System Hardware Requirements

CPU Cores	RAM (GB)	NVMe SSD Storage (GB)	GPU Count
24	64	1660	One or more NVIDIA GPU from the Supported NVIDIA GPUs

Supported NVIDIA GPUs

The MSA Search NIM supports NVIDIA GPUs with at least 48 GB of GPU Memory. It has been tested on the following recommended configurations:

GPU	GPU Memory (GB)	GPU Count
A100	80	1-8
H100	80	1-8
H200	141	1-8
B200	180	1-8
GB200	186	1-8
GB300	288	1-8
GH200	144	1-8
L40S	48	2-8
RTX 6000 Ada Generation	48	2-4

Starting with version 2.0.0, GPUs with 48GB memory (L40S, RTX 6000 Ada) require a minimum of 2 GPUs when using GPU Server (enabled by default). This is because GPU Server holds databases in memory for optimal performance.

Starting with version 2.4.0, the MSA Search NIM adds support for NVIDIA H200, GH200, GB200, and GB300. GH200, GB200, and GB300 are Grace-based platforms (Hopper or Blackwell GPU with NVIDIA Grace CPU). Use container images and host OS builds intended for your CPU architecture (aarch64/arm64 on Grace systems, x86_64/amd64 on typical x86 servers).

Software Prerequisites

Begin with Linux-based Docker-supported operating system
Install Docker - minimum version: 23.0.1
Install NVIDIA Drivers - minimum version: 580
Install the NVIDIA Container Toolkit - minimum version: 1.13.5
Verify your container runtime supports NVIDIA GPUs by running

docker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi

Example output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 580.82.07    Driver Version: 580.82.07    CUDA Version: 13.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA H100 ...     On  | 00000000:1B:00.0 Off |                    0 |
| N/A   33C    P0             75W /  700W |       1MiB /  81559MiB |      0%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

For more information on enumerating multi-GPU systems, refer to the NVIDIA Container Toolkit’s GPU Enumeration Docs

Supported CPU architectures: x86_64 / amd64 for typical discrete-GPU servers, and aarch64 / arm64 for NVIDIA Grace-based platforms (GH200, GB200, GB300) when using the corresponding container images.

NGC (NVIDIA GPU Cloud) Account

To download and run the MSA Search NIM, you need an NGC account and API key:

Create an account on NGC if you don’t already have one
Generate an API Key - ensure that NGC Catalog is selected from the “Services Included” dropdown
Export the API key in your terminal:

export NGC_API_KEY=<your-api-key>

echo "$NGC_API_KEY" | docker login nvcr.io --username '$oauthtoken' --password-stdin

Personal keys allow you to configure an expiration date, revoke, or delete the key using an action button, and rotate the key as needed. For more information about key types, refer to the NGC User Guide.