DL-mo

A benchmark study of deep learning based multi-omics data fusion methods for cancer

---

We here compare the performances of 10 deep learning methods in three contexts:

1. Simulated datasets
2. Cancer datasets
3. Single-cell datasets

We use python and R to code the programs. The python scripts are in ./python-scripts/ folder .The R scripts are in ./R-scripts/ folder .

---

16 deep learning methods

• lfAE
• efAE
• lfDAE
• efDAE
• lfVAE
• efVAE
• lfSVAE
• efSVAE
• lfmmdVAE
• efmmdVAE
• lfNN
• efNN
• lfCNN
• efCNN
• moGCN
• moGAT

---

Input data

The data for python scripts is in ./python-scripts/data/ folder .The data for R scripts is in ./R-scripts/data/ folder .
For python-scripts,Simulated datasets are in ./python-scripts/data/simulations,Cancer datasets are in ./python-scripts/data/cancer ,Single-cell datasets are in ./python-scripts/data/single-cell.

---

python scripts

Each of the three datasets above corresponds to a differnet python scripts in this repositiory:

1. runSimulations*.py
2. runCancer*.py
3. runSingle*.py

---

R scripts

Each of the three datasets above corresponds to a differnet Jupyter notebook in this repositiory:

1. simulated*.ipynb
2. cancer*.ipynb
3. single-cell*.ipynb

---

Install the R software environment

Use conda to create a new environment: conda create -n momix -c conda-forge -c bioconda -c lcantini momix r-irkernel

---

Install the python software environment

You need to build a virtual environment for python.
You need to install the following main libraries: Python==3.7.0,Tensorflow==1.15.0, scikit-learn==0.20.0, Jupyter==1.0.0.