DeepBind

• This repository is for DeepBind reproduction using Pytorch.
  

• Please note that this reproduction is designed for only Chip-seq datasets.
  

• You can check the DeepBind Papaer on here, the corresponding supplementary notes on here, and the original code with tensorflow 1.x on here.
  

Preparations

1. Dependencies
   Create virtual environment DeepBind using following commands
   conda env create --file environment.yml

2. Download datasets
   You can get the required datsets on
   https://github.com/jisraeli/DeepBind/tree/master/data/

How to run the code

1. TF_Binding_Predcition.ipynb
   *no longer supported

2. TF_Binding_Prediction.py
   This code is the same as ipynb format file, but you can experiment multiple datasets using the following commands
   python TF_Binding_Prdiction.py --TF ARID3A
   You can choose datasets among
   [ARID3A / CTCFL / ELK1 / FOXA1 / GABPA / MYC / REST / SP1 / USF1 / ZBTB7A] *no longer supported

3. Logo/seq_logo_from_model.ipynb using this code, you can create sequence logos for specific TF model you trained

4. TF_Binding_Prediction_hyperparameter_experiments.py
   This code is designed for hyperparameter tuning experiments.
   You can execute this code using the command shwon below
   python TF_Binding_Prediction_hyperparameter_experiments.py –-TF {TF Name} –-id {experiments id}
   

TF Binding Prediction AUC Results

• You can check the trainig and testing results on here.
  

Sequence Logo

• You can check the sequence logos created by using the trained models on here.
  

---

• Reference Code : https://github.com/MedChaabane/DeepBind-with-PyTorch