MetaGxData: Clinically Annotated Breast, Ovarian and Pancreatic Cancer Datasets and their Use in Generating a Multi-Cancer Gene Signature

Deena M.a. Gendoo, Michael Zon, Vandana Sandhu, Venkata Manem, Natchar Ratanasirigulchai, Gregory M. Chen, Levi Waldron, Benjamin Haibe‐KainsScientific 2019 · DOI

Abstract

A wealth of transcriptomic and clinical data on solid tumours are under-utilized due to unharmonized data storage and format. We have developed the MetaGxData package compendium, which includes manually-curated and standardized clinical, pathological, survival, and treatment metadata across breast, ovarian, and pancreatic cancer data. MetaGxData is the largest compendium of curated transcriptomic data for these cancer types to date, spanning 86 datasets and encompassing 15,249 samples. Open access to standardized metadata across cancer types promotes use of their transcriptomic and clinical data in a variety of cross-tumour analyses, including identification of common biomarkers, and assessing the validity of prognostic signatures. Here, we demonstrate that MetaGxData is a flexible framework that facilitates meta-analyses by using it to identify common prognostic genes in ovarian and breast cancer. Furthermore, we use the data compendium to create the first gene signature that is prognostic in a meta-analysis across 3 cancer types. These findings demonstrate the potential of MetaGxData to serve as an important resource in oncology research, and provide a foundation for future development of cancer-specific compendia.

Introduction: The Challenge of Data Silos in Oncology

In the current era of high-throughput biology, the scientific community has generated a vast amount of transcriptomic and clinical data. However, these resources remain significantly under-utilized. The primary barrier is the lack of harmonization; data is often stored in disparate formats with inconsistent naming conventions for clinical variables—for instance, one study might use "survival" while another uses "vital status." Furthermore, varying methods for mapping probes to genes complicate cross-study comparisons.

This fragmentation makes large-scale meta-analyses and pan-cancer studies—which are essential for identifying universal cancer biomarkers—manual, error-prone, and time-consuming. To address this, researchers developed MetaGxData, a unified framework designed to harmonize and standardize cancer datasets for robust discovery.

The MetaGxData Framework and Processing Pipeline

The MetaGxData ecosystem consists of three specialized R packages: MetaGxBreast, MetaGxOvarian, and MetaGxPancreas. The core of this framework is a standardized processing pipeline that transforms raw, heterogeneous data into a consistent format suitable for meta-analysis.

The MetaGxData processing pipeline for standardizing feature data (fData) and patient data (pData).

As shown in the pipeline, the process involves semi-automatic curation where clinical and pathological metadata (such as tumor grade, stage, and survival) are standardized using nomenclature from The Cancer Genome Atlas (TCGA). A critical step is the probe-to-gene mapping, which uses a unified repository to map microarray probes to standardized Entrez Gene IDs. To ensure consistency across platforms, the "best" probe for each gene is selected based on the highest variance. Finally, duplicate samples are identified and removed before the expression, feature, and patient data are merged into the final assembly.

A Massive Curated Compendium

MetaGxData represents the largest curated compendium of its kind, encompassing 86 datasets and 15,249 samples across three major cancer types. This scale provides unprecedented statistical power for oncology research.

  • Breast Cancer: 39 datasets (10,004 samples), including major cohorts like METABRIC and TCGA.
  • Ovarian Cancer: 26 datasets (3,526 samples).
  • Pancreatic Cancer: 21 datasets (1,719 samples).
Overview of the MetaGxBreast compendium, showing the availability of clinical annotations like Grade, Vital Status, and Age across various studies.

The standardization of these datasets allows researchers to move beyond small-scale analyses of 3–5 datasets to comprehensive studies involving dozens of cohorts simultaneously, drastically increasing the generalizability of findings.

Generating a Multi-Cancer Prognostic Signature

Leveraging this harmonized data, the authors sought to identify commonalities across breast, ovarian, and pancreatic cancers—types that share biological features such as high frequencies of TP53 mutations and BRCA1/2 inactivation. Using the survcomp R package, they calculated the D.index (a robust estimate of the hazard ratio) for genes across independent datasets.

They identified a 53-gene signature (including CLDN4, DDB2, and FAM189A2) that showed significant prognostic value in both breast and ovarian training sets. This signature was then combined into a "signed average" score to create a unified multi-cancer prognostic tool.

Key Results and Validation

The MetaGx signature demonstrated remarkable performance across all three cancer types, often outperforming established, disease-specific signatures:

  • Pancreatic Cancer: The signature was the most prognostic among five tested (HR 1.64, p < 0.001) and uniquely remained significant across both "basal" and "classical" molecular subtypes.
  • Breast Cancer: In the challenging triple-negative (ER–/HER2–) subtype, the MetaGx signature (HR 1.61) outperformed clinical benchmarks like Mammaprint and Oncotype DX.
  • Ovarian Cancer: It was the only signature that provided statistically significant prognostic capabilities across every molecular subtype (immunoreactive, proliferative, mesenchymal, and differentiated).

Conclusion and Future Directions

The MetaGxData framework proves that biological commonalities across distinct organs are strong enough to support universal prognostic tools. While challenges remain—such as missing clinical data in older studies and underrepresentation of certain subtypes—this compendium provides a foundation for future oncology research. Future work will focus on expanding the framework to other cancer types and utilizing the entire compendium for initial signature training to develop even more robust clinical tools.

Summary updated

Abstract

A wealth of transcriptomic and clinical data on solid tumours are under-utilized due to unharmonized data storage and format. We have developed the MetaGxData package compendium, which includes manually-curated and standardized clinical, pathological, survival, and treatment metadata across breast, ovarian, and pancreatic cancer data. MetaGxData is the largest compendium of curated transcriptomic data for these cancer types to date, spanning 86 datasets and encompassing 15,249 samples. Open access to standardized metadata across cancer types promotes use of their transcriptomic and clinical data in a variety of cross-tumour analyses, including identification of common biomarkers, and assessing the validity of prognostic signatures. Here, we demonstrate that MetaGxData is a flexible framework that facilitates meta-analyses by using it to identify common prognostic genes in ovarian and breast cancer. Furthermore, we use the data compendium to create the first gene signature that is prognostic in a meta-analysis across 3 cancer types. These findings demonstrate the potential of MetaGxData to serve as an important resource in oncology research, and provide a foundation for future development of cancer-specific compendia.
BibTeX
@inproceedings{gendoo2019metagxdata,
  author = {Deena M.A. Gendoo and Michael Zon and Vandana Sandhu and Venkata Manem and Natchar Ratanasirigulchai and Gregory M. Chen and Levi Waldron and Benjamin Haibe‐Kains},
  title = {{MetaGxData: Clinically Annotated Breast, Ovarian and Pancreatic Cancer Datasets and their Use in Generating a Multi-Cancer Gene Signature}},
  booktitle = {Scientific},
  year = {2019},
  doi = {10.1038/s41598-019-45165-4},
}
Figure