-
Parse the data from the UCSC Xena website in PANCAN cohort: https://xenabrowser.net/datapages/?cohort=TCGA%20Pan-Cancer%20(PANCAN)&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443
-
Copy number - somatic mutation (SNP and INDEL) - Gene level non-silent mutation
-
DNA methylation (Methylation450K)
- Dataset: https://xenabrowser.net/datapages/?dataset=jhu-usc.edu_PANCAN_HumanMethylation450.betaValue_whitelisted.tsv.synapse_download_5096262.xena&host=https%3A%2F%2Fpancanatlas.xenahubs.net&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443
- ID Map: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL16304
-
Gene expression RNAseq - TOIL RSEM fpkm
-
Protein expression - RPPA
-
Phenotype - Curated clinical data
Check the jupyter nodebook 'UCSC_union_raw_data_process.ipynb' for details.
-
parse the multiomics data from the synapse website of the dataset ROSMAP https://www.synapse.org/#!Synapse:syn23446022
-
Genome Variants https://www.synapse.org/#!Synapse:syn26263118
-
Methylation https://www.synapse.org/#!Synapse:syn3168763
-
RNA sequence https://www.synapse.org/#!Synapse:syn3505720
-
parse the clinical data https://www.synapse.org/#!Synapse:syn3191087
Check the jupyter nodebook 'ROSMAP_union_raw_data_process.ipynb' for details.
change the dataset arguments if different dataset will be selected to run the experiments.
python load_data.py --dataset 'UCSC'
Select the fold_number to run individual split of the dataset. Numbers ranging from 1 to 5 here are allowed since the 5-fold cross validation is used.
python geo_UCSC_tmain_gcn.py --fold_number 1
python geo_UCSC_tmain_gat.py --fold_number 1
python geo_UCSC_tmain_gin.py --fold_number 1
python python geo_UCSC_tmain_gfomrer.py --fold_number 1
python geo_ROSMAP_tmain_gcn.py --fold_number 1
python geo_ROSMAP_tmain_gat.py --fold_number 1
python geo_ROSMAP_tmain_gin.py --fold_number 1
python geo_ROSMAP_tmain_gformer.py --fold_number 1
The R programing language will be used here to visualize the result in the file 'UCSC_gformer_vis.R', with the attention mechanism in model GraphFormer.
Following is an patient-specific biomaker discovery analysis exmaple. For patient TCGA-D7-6525-01