For any questions about the source code and datasets, please contact Jingchao Ni (jingchao.ni@case.edu), thanks.

Publication

• Jingchao Ni, Mehmet Koyuturk, Hanghang Tong, Jonathan Haines, Rong Xu and Xiang Zhang. "Disease gene prioritization by integrating tissue-specific molecular networks using a robust multi-network model". BMC Bioinformatics, 2016.

Requirements

• All codes are tested using MATLAB R2013a.

Datasets

• P_G_NoSN_PPICenter.mat: the NoSN with tissue-specific PPI networks as the center networks.
• P_G_NoSN_GCNCenter.mat: the NoSN with either tissue-specific gene co-expression networks or tissue-specific PPI networks as the center networks.
• P_G_NoSN_PPICenter_2GCNs.mat: the NoSN with tissue-specific PPI networks as the center networks, and two sets of tissue-specific gene co-expression networks as the auxiliary networks.

Details

• PhenotypeSimNet: the disease similarity network.
• PhenotypeID: the MIM IDs of diseases in the disease similarity network.
• TSGeneNets: the tissue-specific molecular networks corresponding to the diseases in the disease similarity network. Each disease corresponds to at least one tissue-specific molecular network.
• TSGeneNetsID: the corresponding Entrez IDs of genes in TSGeneNets.
• AllGeneID: the Entrez IDs of all genes in all tissue-specific molecular networks.
• Seeds: the known causal genes of diseases in each tissue-specific molecular network, corresponding to TSGeneNets.
• TissueDict: the dictionary of the names of the most associated tissues of the diseases in the disease similarity network.

Algorithm CR

• CR_CrossValidation.m: leave-one-out cross validation of CR.
• CR_Precomputation.m: the precomputation step of CR.
• CR: CR power method.
• J_CR.m: the objective function value of CrossRank.
• AUCEvaluation.m and AUCValue.m: AUC value evaluation with up to 50, 100, 300, 500, 700 and 1000 false positives.

Run CR_CrossValidation.m to see the evaluation results of the leave-one-out cross validation of CR algorithm. CR only uses the center networks in the NoSN datasets.

Algorithm CRstar

• CRstar_CrossValidation.m: leave-one-out cross validation of CRstar.
• CRstar_Precomputation.m: the precomputation step of CRstar.
• CRstar: CRstar power method.
• J_CRstar.m: the objective function value of CrossRankStar.
• AUCEvaluation.m and AUCValue.m: AUC value evaluation with up to 50, 100, 300, 500, 700 and 1000 false positives.

Run CRstar_CrossValidation.m to see the evaluation results of the leave-one-out cross validation of CRstar algorithm.

Algorithm WCRstar

• WCRstar_CrossValidation.m: leave-one-out cross validation of WCRstar.
• WCRstar_Precomputation.m: the precomputation step of WCRstar.
• WCRstar: WCRstar alternating minimization approach.
• J_WCRstar.m: the objective function value of Weighted CrossRankStar.
• J_aux.m: the center-auxiliary network inconsistency value of Weighted CrossRankStar.
• AUCEvaluation.m and AUCValue.m: AUC value evaluation with up to 50, 100, 300, 500, 700 and 1000 false positives.

Run WCRstar_CrossValidation.m to see the evaluation results of the leave-one-out cross validation of WCRstar algorithm. WCRstar uses P_G_NoSN_PPICenter_2GCNs.mat dataset as default.

Note

• CR, CRstar and WCRstar have precomputation steps, which may take some time. The precomputation steps only need to be computed once for a dataset. If a precomputation file exists for a dataset, CR, CRstar and WCRstar will detect it and start ranking directly.