AutoEdge-CCP

A novel cancer-centric multi-association prediction model based on multi-source heterogeneous network integrating node intrinsic attribute information and link information.

Please see our manuscript for more details.

Requirements

• python>=3.7
  
• pytorch>=1.11
  
• pyg>=2.0
  
• numpy>=1.20
  
• networkx>=2.6
  
• pandas>=1.3
  
• scikit-learn>=1.0
  
• Java environment

Code and Data

• data_preprocessing.py:Obtaining node features and edge indices of the multi-source heterogeneous network.
  
• main.py:Extracting edge features from the multi-source heterogeneous network.
  
• circRNA_set.csv:A complete mapping from circRNA index IDs to circRNA names.
  
• drug_set.csv:A complete mapping from drug index IDs to drug meshID, smile and pubchemID.
  
• cancer_set.csv:A complete mapping from cancer index IDs to cancer meshID, cancer names.
  
• circ2cancer_assoMatrix.csv:circRNA-cancer association matrix.
  
• circ2drug_assoMatrix.csv:circRNA-drug association matrix.
  
• cancer2drug_assoMatrix.csv:cancer-drug association matrix.
  
• circRNA_functional_similarity.csv:circRNA functional similarity scores.
  
• circRNA_GIP_similarity.csv:circRNA Gaussian Interaction Profile kernel similarity scores.
  
• circRNA_simfusion.csv:circRNA fusion similarity scores.
  
• drug_structure_similarity.csv:drug chemical structure similarity scores.
  
• drug_GIP_similarity.csv:drug Gaussian Interaction Profile kernel similarity scores.
  
• drug_simfusion.csv:drug fusion similarity scores.
  
• cancer_semantic_similarity.csv:cancer semantic similarity scores.
  
• cancer_GIP_similarity.csv:cancer Gaussian Interaction Profile kernel similarity scores.
  
• cancer_simfusion.csv:cancer fusion similarity scores.
  
• circRNA_cancer_LabEdgEmbs4LTR.txt:Input data of learning to rank for circRNA-cancer prediction.
  
• drug_cancer_LabEdgEmbs4LTR.txt:Input data of learning to rank for drug-cancer prediction.
  Note：1) The header IDs “0-476” in the mentioned dataset can be mapped to specific information about circRNAs, drugs, and cancers by the three mapping files circRNA_set.csv, drug_set.csv, and cancer_set.csv, respectively. 2) See manuscript Section 2.5 for data format details of circRNA_cancer_LabEdgEmbs4LTR.txt and drug_cancer_LabEdgEmbs4LTR.txt.

Usage

• Here, we provide a demo for the first application scenario:"associated cancer ranking for novel queries." Depending on the specific prediction task, data for ${train.txt} and ${test.txt} can be obtained from either data/circRNA-cancer or data/drug-cancer.
  train:
  

  >> java -jar bin/RankLib.jar -train ${train.txt} -ranker 6  -metric2t NDCG@10  -tree 1000 -shrinkage 0.1 -tc 256 -mls 1 -save model.txt
  

  test:
  

  >> java -jar bin/RankLib.jar -load model.txt -rank ${test.txt} -score score.txt