chenxofhit / loc-pca-cmi

a novel method of gene regulatory network structure inference

Loc-PCA-CMI

Loc-PCA-CMI is a novel method of gene regulatory network structure inference on gene knock-out expression data,which first identify Local Overlapped gene Clusters, and then in conjunction with PCA-CMI method each local cluster structure is refined.

System Environment

Ubuntu/Linux Bash, Matlab 2015b, R 3.3.1 (with package "readr", "R.matlab", "Matrix" , "RLowPC", "minet", "igraph","gtools").

Running Step

Step 1, To run the complete experiment with only one bash command in the current folder (sudo privillege perhaps needed according to your running environment):

cd ./ && bash ./main_locpcacmi_DREAM3_batch.sh

The command will generate four sub folders as result_loc_pcacmi, result_loc_pcapmi, result_pca_cmi and result_pca_pmi and the result including AUPR and AUROC will be output to the folders. For a more visualized summary you may use the below command to collect the results:

cd ./ && bash ./result_merge.sh

and the summary will be output to result_merge_final.res .

Step 2, For comparison with more methods including ARACNE, MRNET, PCA-PMI, PCA-CMI and loc-PCA-PMI as in the paper stated, it is encouraged to run the following bash command in the folder comparison:

cd ./comparison && bash ./loc-PCA-CMI_comparison.batch.sh 

Above command will generate both txt files and image files with AUPR and AUROC details in the same folder.

Step 3 (Addtional), In order to to validate the parameter order0 in the PCA alogrithm, main_locpcacmi_DREAM3_Ecoli_tpl_k.sh and main_locpcacmi_DREAM3_Yeast_tpl_k.sh are applied and result is output to folder result_*_k.

Any question, please do not hesitate to contact me with following address with bash command for decryption:

echo "Y2hlbnhvZmhpdEBnbWFpbC5jb20K"|base64 -d

or submit issue in the repository directly.