README.md

geo2drug

E Flynn

This repository annotates GEO datasets using DrugBank and Cellosaurus labels. Then, we download and sex label all data.

The code directory is set up as follows:

00_annot/

1. GEO metadata is downloaded using 00_get_list_all_gse.R

• output: (in data/01_sample_lists/) * <>_gse_gsm.csv * gse_<>.csv * gse_gsm_all_geo_dedup.csv * gse_all_geo_info.csv

2. 01_cell_labeling/

• 00_parse_cellosaurus.py
  • input: cellosaurus XML, located in data/00_db_data/cellosaurus.xml
  • output: data/00_db_data/cellosaurus.json
• 01_process_cell_df.R
  • output: data/00_db_data/cellosaurus_df.txt
• 02_cell_line_labeling.R
  • uses data/01_sample_lists/gse_all_geo_info.csv
  • output: data/02_labeled_data/cell_line_mapped_gse.txt

3. 02_drug_labeling/

• 00_drugbank_synonyms.py
  • input: drugbank.xml
  • output: data/00_db_data/drugbank_info.json
• 01_process_drugbank.R
  • output: data/00_db_data/drugbank_parsed.txt
• 02_drug_gse_labeling.R
  • uses data/01_sample_lists/gse_all_geo_info.csv
  • output: data/02_labeled_data/drugbank_mapped_gse.txt

01_download/

This uses the NCBI aspera client and relies on wrenlab software packages and is slightly complicated to install. sbatch 00_download_gse_wrapper.sh ${ID} ${GSE_LIST}

This runs download_geo_chunk.sh which runs downloadGEO.py for each individual GSE. We do this for ID=["mouse", "rat", and "human"], and use the files "data/sample_lists/gse_${ID}.csv".

// TODO: download GPL scripts

02_process/

This directory processes downloaded GSEs using exprsex based on a reference list of files.

• 00_convert_to_mat.sh
• 01_label_mat.sh
• 02_combine_mat.sh
• 03_train_test_divide.R
• 04_run_meta.sh
• 05_train_test_mat.sh

Required files:

• GEOMetadb (update GEOMetadb in utils path to this)
• drugbank and cellosaurus XML in the 00_db_data directory
• jake_stopwords file