CancerSig: This work proposes an evolutionary learning method called CancerSig to identify stage-specific miRNA signatures for cancer stage prediction. CancerSig obtains a small set of miRNA biomarkers as a signature and establishes a panel of miRNAs to predict cancer stage across 15 cancer types.
miRNA expression (RPM log2) with CSV format. User can put their in-house miRNA expression into csv file.
Dataset | Abbreviation | miRNA panel |
---|---|---|
Bladder urothelial carcinoma | BLCA | BLCA_input.csv |
Breast invasive carcinoma | BRCA | BRCA_input.csv |
Colon adenocarcinoma | COAD | COAD_input.csv |
Esophageal Carcinoma | ESCA | ESCA_input.csv |
Head and neck squamous cell carcinoma | HNSC | HNSC_input.csv |
Kidney renal clear cell carcinoma | KIRC | KIRC_input.csv |
Kidney renal papillary cell carcinoma | KIRP | KIRP_input.csv |
Liver hepatocellular carcinoma | LIHC | LIHC_input.csv |
Lung adenocarcinoma | LUAD | LUAD_input.csv |
Lung squamous cell carcinoma | LUSC | LUSC_input.csv |
Rectum adenocarcinoma | READ | READ_input.csv |
Skin cutaneous melanoma | SKCM | SKCM_input.csv |
Stomach adenocarcinoma | STAD | STAD_input.csv |
Thyroid carcinoma | THCA | THCA_input.csv |
Uveal melanoma | UVM | UVM_input.csv |
git clone https://github.com/mingjutsai/CancerSig.git
cd CancerSig
build LIBSVM
cd libsvm
make
python cancersig_main.py -h
usage: cancersig_main.py [-h] -t T -i I
CancerSig obtains a small set of miRNA biomarkers as a signature and establishes a panel of miRNAs to predict cancer stage across 15 cancer
types.
optional arguments:
-h, --help show this help message and exit
-t T choose one of the cancer type(BLCA, BRCA, COAD, ESCA, HNSC, KIRC, KIRP, LIHC, LUAD, LUSC, READ, SKCM, STAD, THCA, UVM)
-i I the file of miRNA gene expression for specific cancer type
python cancersig_main.py -t BLCA -i models/BLCA/BLCA_input.csv
svm-predict exist:libsvm/svm-predict
svm-scale exist:libsvm/svm-scale
cancer type:BLCA
expression file:models/BLCA/BLCA_input.csv
model exist:models/BLCA/BLCA.model
prediction score:0.320973
Normalized probabilities between 0 to 1, with higher scores more likely to be advanced stages and lower scores more likely to be early stages.
- Srinivasulu Yerukala Sathipati: sathipathi.srinivasulu@marshfieldclinic.org
- Ming-Ju Tsai: mingjutsai@hsl.harvard.edu