Ovarian Cancer Subtyping Analysis

The manifest of TCGA-OV samples is located in:

data/tcga-ov-metadata/metadata.cohort.2025-06-23.json

That JSON is used to define the TCGA-OV case cohort. The pipeline then queries GDC for current mRNA, miRNA, and clinical data.

Requirements

• gdc client

Pipeline Steps

Here are the steps of the full workflow:

1. scripts/pull_gdc_tcga_ov_omics.py

   • queries GDC for current TCGA-OV mRNA STAR-count files
   • queries GDC for current TCGA-OV mature miRNA quantification files
   • pulls clinical records for the union of mRNA/miRNA cases
   • writes manifests and file-to-sample link tables

2. Characterize mRNA and miRNA clusters

   • run the cluster assignment workflows
   • write cluster/subtype assignments before survival is calculated

3. scripts/harmonize_gdc_omics_labels.py

   • joins existing mRNA subtype labels onto current GDC mRNA rows
   • joins miRNA NMF subtype labels onto current GDC miRNA rows
   • writes harmonized label tables and subtype agreement summaries

4. scripts/link_gdc_omics_samples_to_survival.py

   • computes case-level overall survival fields
   • joins survival fields onto mRNA and miRNA sample/file rows
   • appends survival fields to the harmonized label tables

Downloading and Harmonizing Data and Metadata

To identify the mRNA and miRNA data associated with the TCGA-OV cases and download the corresponding mRNA and miRNA data, run:

python3 scripts/run_gdc_omics_pipeline.py --download

This uses gdc-client with the generated manifests and writes outputs under:

data/gdc_tcga_ov_omics/

Key Outputs

data/gdc_tcga_ov_omics/gdc_manifest_mrna_star_counts.tsv
data/gdc_tcga_ov_omics/gdc_manifest_mirna_mirbase21_mature.tsv
data/gdc_tcga_ov_omics/mrna_file_sample_link.tsv
data/gdc_tcga_ov_omics/mirna_file_sample_link.tsv
data/gdc_tcga_ov_omics/harmonized_labels/mrna_samples_labels.tsv
data/gdc_tcga_ov_omics/harmonized_labels/mirna_samples_labels.tsv
data/gdc_tcga_ov_omics/harmonized_labels/omics_sample_assay_availability_labels.tsv
data/gdc_tcga_ov_omics/harmonized_labels/omics_samples_survival_labels_long.tsv

Identify mRNA Clusters

The first analysis generates mRNA clusters from the TCGA-OV data using the package ConsensusOV. This analysis runs out of the mRNA_clusters directory. Run the AssignClusters.Rmd notebook or AssignClusters.R script to generate cluster assignments and scores for each sample. These outputs will be written to the output directory within mRNA_clusters

Identify miRNA Clusters

The miRNA clustering workflow uses the miRNA files downloaded by the shared GDC pipeline:

data/gdc_tcga_ov_omics/downloads/miRNA/
data/gdc_tcga_ov_omics/mirna_file_sample_link.tsv

Run the miRNA notebooks in order:

quarto render miRNA_clusters/01_mirna_preprocess.qmd
quarto render miRNA_clusters/02a_mirna_pca.qmd
quarto render miRNA_clusters/02b_mirna_kmeans.qmd
quarto render miRNA_clusters/02c_mirna_nmf.qmd

The preprocessing notebook writes the normalized handoff matrix and GDC sample metadata to:

data/mirna_data/mirna_lcpm_matrix.rds
data/mirna_data/mirna_sample_metadata.csv

The NMF notebook writes the miRNA subtype assignments to:

miRNA_clusters/mirna_nmf_subtypes_k4.csv

That CSV is the default miRNA subtype input for harmonization. It should contain one row per clustered miRNA file/sample and a cluster column; the current pipeline also carries through GDC identifiers such as file_name, case_submitter_id, sample_submitter_id, and aliquot_barcode.

Older copies of this CSV may have a sample_id column that actually contains miRNA quantification file names. Those values are not expected to match the mRNA subtype sample column or mRNA file_id values, because those are assay-specific GDC files. Harmonization resolves miRNA and mRNA labels through the shared GDC metadata, using sample_submitter_id and case_submitter_id as the cross-assay identifiers.

Harmonize Labels With Clinical Data

After mRNA and miRNA cluster assignments have been generated, run:

python3 scripts/run_gdc_omics_pipeline.py --skip-pull --harmonize-labels --survival

This harmonizes mRNA subtype labels and miRNA NMF subtype labels, then appends clinical/survival fields. By default, miRNA labels are read from:

miRNA_clusters/mirna_nmf_subtypes_k4.csv

Use the harmonized outputs, not raw mRNA/miRNA file names, to compare assays:

data/gdc_tcga_ov_omics/harmonized_labels/omics_sample_assay_availability_labels.tsv
data/gdc_tcga_ov_omics/harmonized_labels/omics_samples_survival_labels_long.tsv

To use a different miRNA subtype file, pass it explicitly:

python3 scripts/run_gdc_omics_pipeline.py --skip-pull --harmonize-labels --survival --mirna-subtype-file path/to/mirna_nmf_subtypes_k4.csv

For survival analyses, filter to:

has_os_time == TRUE