AMPERE: Aligned Multimodal Protein Encoder Representations

AMPERE is a framework for integrating and aligning multiple protein encoder representations. It provides a unified interface to work with various protein language models and multimodal encoders, facilitating downstream tasks in protein engineering and analysis.

Key Features

• Integration of Multiple Protein Encoders: Seamlessly work with various state-of-the-art protein encoders.
• Representation Alignment: Techniques for aligning different embedding spaces for better multimodal integration.
• Protein Sequence & Structure Support: Tools for processing both sequence and structural information.

Supported Encoders

The repository integrates several protein encoding models, including:

• ESM series
• SaProt
• ProTrek
• ProteinDT
• DiMA
• And more...

Installation

This project uses Pixi for dependency management.

pixi install

Usage

Refer to the scripts and auto-scripts directories for examples on how to run experiments and process data.