Installing AIDE

AIDE is designed with a modular architecture to minimize dependency conflicts while providing access to a wide range of protein prediction tools. The base package has minimal dependencies and provides core functionality, while additional components can be installed based on your specific needs.

Quick Install

The package is not currently available on PyPI, please clone the repo:

git clone https://github.com/beckham-lab/aide_predict

For basic functionality, simply install AIDE using:

# Create and activate a new conda environment
conda env create -f environment.yaml

# Install AIDE
pip install .

Supported Tools by Installation Level

AIDE provides bespoke embedders and predictors as additional modules that can be installed. These fall into three categories, with environment weight in mind: those available in the base package, those that can be installed with minimal additional pip dependencies, and those that should be built as an independant environment.

Base Installation

The base installation provides:

  • Core data structures for protein sequences and structures

  • Sequence alignment utilities

  • One-hot encoding embeddings

  • K-mer based embeddings

  • Basic Hidden Markov Model support

  • mmseqs2 MSA generation pipeline

Minor Pip Dependencies

Pure transformers models

ESM2 and SaProt can be defined with the transformers library. To install these models:

pip install -r requirements-transformers.txt

This enables:

  • ESM2 embeddings and likelihood scoring

  • SaProt structure-aware embeddings and scoring

MSA Transformer

MSA transformer requires bespoke components from fair-esm:

pip install -r requirements-fair-esm.txt

This enables:

  • MSA transformer embeddings and likelihood scoring

EVmutation

For evolutionary coupling analysis:

pip install -r requirements-evmutation.txt

This enables:

  • EVMutation for protein mutation effect prediction

VESPA Integration

For conservation-based variant effect prediction:

pip install -r requirements-vespa.txt

Independent Environment

EVE Integration

EVE requires special handling due to its complex environment requirements:

  1. Clone the EVE repository outside your AIDE directory:

git clone https://github.com/OATML/EVE.git

  1. Set required environment variables:

export EVE_REPO=/path/to/eve/repo

  1. Create a dedicated conda environment for EVE following their installation instructions.

  2. Set the EVE environment name:

export EVE_CONDA_ENV=eve_env

Verifying Your Installation

You can check which components are available in your installation:

from aide_predict.utils.checks import get_supported_tools
print(get_supported_tools())

Common Installation Issues

CUDA Compatibility

If you’re using GPU-accelerated components (ESMFold, transformers), ensure your CUDA drivers are compatible:

  • Check CUDA version: nvidia-smi

  • Match PyTorch installation with CUDA version

  • For Apple Silicon users: Some components may require alternative installations