Translatomer

predicting ribosome profiling reveals translational regulation and interprets disease variants

Translatomer

This is our implementation for the paper:

Jialin He, Lei Xiong#, Shaohui Shi, Chengyu Li, Kexuan Chen, Qianchen Fang, Jiuhong Nan, Ke Ding, Jingyun Li, Yuanhui Mao, Carles A. Boix, Xinyang Hu, Manolis Kellis, Jingyun Li and Xushen Xiong#. Deep learning prediction of ribosome profiling with Translatomer reveals translational regulation and interprets disease variants. (Nature Machine Intelligence)

Introduction

Translatomer is a transformer-based multi-modal deep learning framework that predicts ribosome profiling track using genomic sequence and cell-type-specific RNA-seq as input. Overview

Citation

If you want to use our codes and datasets in your research, please cite:


Prerequisites

To run this project, you need the following prerequisites:

You can install all the required packages using the following command:

conda create -n pytorch python=3.9.16
conda activate pytorch
pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117
pip install -r requirements.txt 

Data Preparation

Example data for model training can be downloaded from Zenodo

[options]:


Example to run the codes:

find data/ -type d -name ‘output_features’ -exec mkdir -p ‘{}/tmp’ \; find data/ -type d -name ‘input_features’ -exec mkdir -p ‘{}/tmp’ \; nohup python generate_features_4rv.py –assembly hg38 –celltype HepG2 –study GSE174419 –region_len 65536 –nBins 1024 & nohup python generate_features_4rv.py –assembly hg38 –celltype K562 –study GSE153597 –region_len 65536 –nBins 1024 &


## Model Training
To train the Translatomer model, use the following command:

python train_all_11fold.py [options]

[options]:

Tutorial

License

This project is licensed under MIT License.

Contact

For any questions or inquiries, please contact xiongxs@zju.edu.cn.