We proposed a two-step computational framework called CNA_origin to predict the tissue-of-origin of a tumor from its gene CNA levels. CNA origin set up an intellectual deep-learning network mainly composed of autoencoder and convolution neural network (CNN).

If you want to use CNA_origin, you must have gene-level CNA file and label file.

The use of CNA_origin: CNA_origin.py -T PATH_GENE_CNV:  File of the gene CNV
                   -G PATH_LABEL:  File of the sample label
                   [-d DIM_NUMBER]:The Number of Features after Dimension Reduction, default:100
                   [-k K_CROSS_VALIDATION]: k fold cross validation, default:10
                   [-s TRAINING_PART_SCALE]: Split scale for train/test,default:0.1
                   [-o OUTPUT_FILE]:  The result output path

The merge-group file contains sample label information. The merge-sample file contains the gene-level CNA information of 50 samples. The complete datasets were from primary solid tumor samples released by MSKCC in 2013, which could be downloaded from http://cbio.mskcc.org/cancergenomics/pancan_tcga/ or http://gdac.broadinstitute.org/. We recommend using dataset with sample size greater than 400.

for example:  python CNA_origin.py  -T merge-sample   -G merge-group

CNA origin was implemented in python 3.7.3 using keras (2.24) with the backend of tensorflow (1.14.0)

The program now has a bug that can only be run using CPU (not GPU). We are trying to fix it.

If you have any question,please send email to [email protected].  We will continue to improve the code of CNA_origin.