ProtChat: An AI Multi-Agent for Automated Protein Analysis Leveraging GPT-4 and Protein Multimodal Model

Website

we have developed a website that allows users from any computer background to use it without the need to pay for a GPT-4 API Key. Future maintenance and updates follow this standard.

Setup

Clone the repository and navigate into the directory:

git clone [email protected]:SIAT-code/ProtChat.git
cd ProtChat

Installation

To use this project, do as the following to install the dependencies.

1. Create a new environment using: conda create --name protchat python=3.9.
2. Activate the environment you have just created: conda activate protchat.
3. Finally, run the following command to install the required packages inside the conda environment:

pip install -r requirements.txt

Dataset

You can download the datasets from this link, which provides some samples. The directory of data files like the following:

| -- downstream_task
|    |-- stability
|    |-- fluorescence
|    |-- remote_homology
|    |-- secondary_structure
|    |-- antigen_binding
|    |-- pdbbind
|    |-- kinase
|    |-- skempi

All the datasets that we used in the project can be found in the datasets folder. They are preprocessed and ready to use.You can use them directly.

Checkpoints

We have prepared model checkpoints for direct inference. Please refer to this link. The directory of the checkpoint files is as follows:

| -- checkpoints
|    |-- stability
|    |-- fluorescence
|    |-- remote_homology
|    |-- secondary_structure
|    |-- antigen_binding
|    |-- pdbbind
|    |-- kinase
|    |-- skempi
|        |-- fold1
|        |-- fold2
          ...
|        |-- fold10

Inference

Please refer to AutoGen-MASSA.ipynb for the inference code, the dialogue template for the agent is as follows:

user_proxy.initiate_chat(manager, message="""Can you analyze on kinase task which belongs to protein-drug interaction prediction benchmark, where the data file and format are as follows, and evaluate the predictions based on the task, finally visualize the evaluation results?

downstream_task/kinase/samples_seq_mole_go.txt

Start the work now.
""")

Fine-tune

If you wish to perform fine-tuning, please refer to the AutoGen-MASSA.ipynb as well, the dialogue template for the agent is as follows:

user_proxy.initiate_chat(manager, message="""Can you fine-tune and analyze on antigen binding task which belongs to protein property prediction benchmark, where the data file and format are as follows, and evaluate the predictions based on the task, finally visualize the evaluation results?

downstream_task/antigen_binding/sequence_go.txt

Start the work now.
""")