tumor-gtex-plots

Expression Plots

This analysis is DEPRECATED and was last run with OpenPedCan data release v12.

Module authors: Komal Rathi (@komalsrathi)

Purpose

1. Create tumor-normal-gtex and pan-cancer expression plots, with TPM on the y-axis, as follows:

• cohort + cancer_group pan-cancer plots: For each cohort + cancer_group, plot all tumors.
• cancer_group pan-cancer plots: For each cancer_group, plot all tumors.
• cohort + cancer_group tumor-normal-gtex plots: For each cancer_group within a cohort, plot cohort + cancer_group vs GTEx subgroups.
• cancer_group tumor-normal-gtex plots: For each cancer_group, plot cancer_group vs GTEx subgroups.

2. Export two tables (.tsv) cancer_group_level.tsv and cohort_cancer_group_level.tsv per comparison type i.e. tumor_normal_gtex_plots_ and pan_cancer_plots_ with the following fields: Gene_symbol, Gene_Ensembl_ID, Dataset, Disease, x_labels, mean, median, sd, plot_api. Add MONDO, RMTL and EFO using long-format-table-utils/annotator/annotator-cli.R. Convert to JSONL and gzip.

3. Export metadata files with the following fields: gene (gene name), plot_type (either pan_cancer or tumor_normal_gtex), cohort (tumor cohort name for tumor-normal-gtex plots and all_cohorts for pan-cancer plots) for , cancer_group (only applicable for tumor-normal-gtex plots and NA for tumors only plot), analysis_type (either cohort_cancer_group_level or cancer_group_level), plot_fname (plot filename) and table_fname (table filename).

Methods

1. Filter cancer_subgroup or gtex_subgroup with number of samples < 5

Module structure

├── 01-tumor-gtex-plots.R # script to call all functions and run analysis
├── 02-chunkwise-annotate-and-zip.R # script to annotate files with MONDO, RMTL, EFO fields, convert to JSONL and gzip
├── README.md 
├── plots # png files of expression boxplots (possible examples)
│   ├── GPC2_GMKF_Neuroblastoma_vs_GTEx_cohort_cancer_group_level.png
│   ├── GPC2_PBTA_Medulloblastoma_vs_GTEx_cohort_cancer_group_level.png
│   ├── GPC2_Ependymoma_vs_GTEx_cancer_group_level.png
│   ├── GPC2_PBTA_GMKF_pan_cancer_cancer_group_level.png
│   ├── GPC2_PBTA_GMKF_pan_cancer_cohort_cancer_group_level.png
│   └── ...
├── results
│   ├── metadata.tsv # metadata file for all comparisons
│   ├── pan_cancer_plots_cancer_group_level.jsonl.gz
│   ├── pan_cancer_plots_cancer_group_level.tsv
│   ├── pan_cancer_plots_cohort_cancer_group_level.jsonl.gz
│   ├── pan_cancer_plots_cohort_cancer_group_level.tsv
│   ├── tumor_normal_gtex_plots_cancer_group_level.jsonl.gz
│   ├── tumor_normal_gtex_plots_cancer_group_level.tsv
│   ├── tumor_normal_gtex_plots_cohort_cancer_group_level.jsonl.gz
│   └── tumor_normal_gtex_plots_cohort_cancer_group_level.tsv 
├── run-pan-cancer-plots.sh # script for pan-cancer plots
├── run-tumor-gtex-plots.sh # full analysis script
├── run-tumor-normal-gtex-plots.sh # script for tumor vs gtex plots
├── run-annotator.sh # script to call 02-chunkwise-annotate-and-zip.R
└── util
    ├── pubTheme.R # publication quality ggplot2 theme
    ├── pan_cancer_plot.R # function for pan-cancer plot
    └── tumor_normal_gtex_plot.R # function for tumor-normal-gtex plot

Analysis scripts

01-tumor-gtex-plots.R

Input parameters:

Rscript 01-tumor-gtex-plots.R --help

Options:
	--expr_mat=EXPR_MAT
		collapsed TPM expression data: HUGO gene symbol x Sample identifiers (.rds)

	--hist_file=HIST_FILE
		histologies file (.tsv)

	--map_file=MAP_FILE
		gene symbol-ensembl id mapping file

	--cohort_list=COHORT_LIST
		comma separated list of cohorts

	--tumor_vs_normal=TUMOR_VS_NORMAL
		TRUE or FALSE

	--analysis_type=ANALYSIS_TYPE
		cohort_cancer_group_level or cancer_group_level

	--plot_width=PLOT_WIDTH
		width in pixels

	--plot_height=PLOT_HEIGHT
		height in pixels

	--meta_file=META_FILE
		filename for writing out file names and other info

Inputs from data download:

Using the current version:

Expession matrix: gene-expression-rsem-tpm-collapsed.rds Histologies file: histologies.tsv

Outputs:

1. plots/*.png: expression boxplots with TPM as y-axis

Tumors only filename format: {gene_name}_{underscore_separated_cohort_name}_{analysis_type}.png

Tumors vs GTEx filename format:

# cohort + cancer_group level
{gene_name}_{cohort_name}_{cancer_group_name}_vs_GTEx_{analysis_type}.png

# cancer_group level
{gene_name}_{cancer_group_name}_vs_GTEx_{analysis_type}.png

2. results/pan_cancer_plots_cancer_group_level.tsv and results/pan_cancer_plots_cohort_cancer_group_level.tsv: corresponding data for pan-cancer expression boxplots

3. results/tumor_normal_gtex_plots_cancer_group_level.tsv and results/tumor_normal_gtex_plots_cohort_cancer_group_level.tsv: corresponding data for tumor-normal-gtex expression boxplots

4. results/metadata.tsv: metadata file for all comparisons

02-chunkwise-annotate-and-zip.R

Input parameters:

Rscript 02-chunkwise-annotate-and-zip.R --help

Options:
	--input_file=INPUT_FILE
		input file to annotate, convert to jsonl and gzip

Inputs:

This script takes the output .tsv files from 01-tumor-gtex-plots.R, annotates using the long-format-table-utils/annotator/annotator-cli.R script, reads back annotated data (.tsv) into chunks of 1 million rows, converts to JSON and then JSONL, followed by combining all JSONL files and gzipping the final JSONL file.

Outputs:

1. results/pan_cancer_plots_cancer_group_level.jsonl.gz and results/pan_cancer_plots_cohort_cancer_group_level.jsonl.gz: corresponding data for pan-cancer expression boxplots

2. results/tumor_normal_gtex_plots_cancer_group_level.jsonl.gz and results/tumor_normal_gtex_plots_cohort_cancer_group_level.jsonl.gz: corresponding data for tumor-normal-gtex expression boxplots

Running the analysis

# pan-cancer plots
bash run-pan-cancer-plots.sh

# tumor-normal-gtex plots
bash run-tumor-normal-gtex-plots.sh

# annotator script
bash annotator.sh

# running both scripts
bash run-tumor-gtex-plots.sh

Final output

The final plots as .png and tables as .tsv and .jsonl.gz files have been uploaded to s3://kf-openaccess-us-east-1-prd-pbta/open-targets/analyses/tumor-gtex-plots/