This description applies to each directory. Here I will use thca as an example.

The convert.sh file prepares the text files necessary to create the count and phenotype Robjects.

First, create 3 files

File with absolute path to each count file

ls /group/stranger-lab/grossman_cancer_data/thca/*.gz > thca_files.txt

File provided by Allison to link GDC UUID with TCGA UUID

thca_htseq.txt

File of phenotype data, 3 columns with TCGA_ID, TCGA_UUID, and Gender

Created by:

grep bcr_patient_barcode All_CDEs.txt | tr '\t' '\n' > barcode.txt
grep bcr_patient_uuid All_CDEs.txt | tr '\t' '\n' > uuid.txt
grep gender All_CDEs.txt | tr '\t' '\n' > gender.txt

paste barcode.txt uuid.txt gender.txt > phens.txt

Run the convert.sh script with these input files

sh convert.sh thca_files.txt thca_htseq.txt phens.txt

This creates the id_conversions.txt file read by the expr.R script.

First, create a file with duplicated IDs

cut -f 1 id_conversions.txt | sort | uniq -d > duplicated.txt

This was used previously to remove individuals with both tumor and normal samples.

Run the expr.R script, which filters out duplicated ids, and sorts both the phen and count Robjects to have the same individuals in the same order.

The count_DE.R script takes the previously generated count and phenotype Robjects and runs the DESeq2 sex-biased differential expression analysis.