DeepCellEss

An interpretable deep learning-based cell line-specific essential protein prediction model.

The DeepCellEss web server for prediction and visualization available at http://bioinformatics.csu.edu.cn/DeepCellEss

Requirements

• python=3.7.0
• numpy=1.19.2
• pandas=1.1.5
• scikit-learn=0.24.2
• scipy=1.7.1
• pytorch=1.9.0
• gensim=3.8.3

Usage

An demo to train DeepEssCell on the dataset of HCT-116 cell line using linux-64 platform.

1. Clone the repo

$ git clone https://github.com/lynn-1998/DeepCellEss.git
$ cd DeepCellEss

2. Create and activate the environment

$ cd DeepCellEss
$ conda create --name deepcelless --file requirments.txt
$ conda activate deepcelless

3. Train model

The trained models will be saved at file folder '../protein/saved_model/HCT-116/'.

$ cd code
$ python main.py protein --cell_line HCT-116 --gpu 0

4. Specify model hyperparameters

--batch_size is the size of each batch while training. --kernel_size is the kernel number of the CNN layer.
--head_num is the number of attention heads.
--hidden_dim is the dimention of the hidden state vector.
--layer_num is the number of lstm layers.
--gpu is the gpu number you used to build and train the model. The defalt value of 0 means "cuda:0". No gpu will default to cpu.

License

This project is licensed under the MIT License - see the LICENSE.txt file for details

Concat

Please feel free to contact us for any further questions.

• Yiming Li [email protected]
• Min Li [email protected]