ProFOLD2

ProFOLD2 - A protein 3D structure prediction application

Requirements

• Python3.9+
• Dependencies

Running ProFOLD2

1. Clone this repository and cd into it.

$git clone https://github.com/bigict/ProFOLD2.git
$cd ProFOLD2
$git submodule update --init  # required if use FusedEvoformer, recommended.

2. Create a virtual enviroment and install dependencies

$conda create -n profold2 python=3.11
$conda activate profold2
$bash install_env.sh

3. Train a model

$python main.py train --prefix=OUTPUT_DIR

There are a lot of parameters, you can run

$python main.py train -h

for further help.

ProFOLD2 logs it's metrics to TensorBoard. You can run

$tensorboard --logdir=OUTPUT_DIR

Then open http://localhost:6006 in you browser.

4. Inference

$python main.py predict --models [MODEL_NAME1:]MODEL_FILE1 [MODEL_NAME2:]MODEL_FILE2

Just like train, you can run

$python main.py predict -h