MOINER: a multi-omics integration framework with information enhancement and image representation learning

MOINER Introduction

Install

conda create -n IE-MOIF python=3.7
conda activate IE-MOIF
pip install -r requirements.txt
pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 –f https://download.pytorch.org/whl/torch_stable.html --user
pip install ./utils_map/ lapjv-1.3.1.tar.gz

Usage

Integration Step

python main_IE-MOIF.py --data omics1_file omics2_file omics3_file --label label_file --type omics_1_name omics_2_name omics_3_name --fs_num 1000 1000 500

Classification Step

python VIT.py --task num_class --patch num_patch --mark dataset

Example

python main_IE-MOIF.py --data ./example/mRNA.csv ./example/meth.csv ./example/miRNA.csv --label ./example/label.csv --type mRNA meth miRNA --drm fs --fs_num 1000 1000 500 --fem tsne
python main_En-VIT.py --n_class 2 --patch 25 --Example

NOTE

In the MOINER framework, ViT is used as the default classification model, and En-ViT based on ensemble learning needs to be manually set to open in parameters because it takes a long time to train.