Cerebra

A computationally efficient framework for accurate protein structure prediction

Install software on Linux

1. download Cerebra

git clone https://github.com/Gonglab-THU/Cerebra.git
cd Cerebra

2. install Anaconda / Miniconda software

3. install Python packages

conda create -n cerebra python=3.11
conda activate cerebra

conda install pytorch 
pip install click
pip install numpy
pip install ml_collections
pip install scipy
pip install einops
pip install dm-tree
pip install biopython

Usage

• You should modify the MSA file
• You should download the model1-6 parameters and move it into the model folder

wget https://zenodo.org/records/10608346/files/model.zip?download=1
tar -zcvf model.zip ./
bash run.sh -i example/test.a3m -o example

• If you need to analyze the pearson correlation between path metrics and PSA attention weights, please use model6.pth.
• As searching MSA can be time-consuming, please use search_MSA.py to search MSA if you want to obtain results consistent with those in Cerebra.

Reference

Cerebra: a computationally efficient framework for accurate protein structure prediction