DIDL: A deep learning approach to predict inter-omics interactions in multi-layer networks

This repository contains the Python code and dataset for the paper titled “A deep learning approach to predict inter-omics interactions in multi-layer networks” by Niloofar Borhani, Jafar Ghaisari, Maryam Abedi, Marzieh Kamali and Yousof Gheisari. Data Integration with Deep Learning (DIDL) is a nonlinear deep learning framework designed to predict inter-omics interactions. It combines automatic feature extraction and interaction prediction, achieving state-of-the-art performance across diverse biological networks such as drug–target, TF–DNA, and miRNA–mRNA interactions. The paper is available at https://doi.org/10.1186/s12859-022-04569-2.

Training and Evaluation with k-Fold Cross Validation

To train the DIDL model and validate it using k-fold cross-validation, run the following command:

python main.py \
  --data_name dataset/miRNAmRNA.csv \
  --negative_sampling 1.0 \
  --mir_layer [64,32,20] \
  --prot_layer [64,32,20] \
  --dropout 0.5 \
  --reg_L2 0.08 \
  --batch_size 32 \
  --n_fold 10 \
  --learning_rate 1e-5 \
  --epochs 20 \
  --n_epochs_stop 5

Contact Information

For further inquiries, please contact.

Niloofar Borhani
Ph.D. Student, Control Engineering
Isfahan University of Technology
Email: n.borhani@ec.iut.ac.ir
CV: Google Scholar