Integrative omics data analysis uncovers biomarker genes and potential candidate drugs for G3 medulloblastoma

Loreina Guo, Kendall Cornick, Vincent Xu, Tianyao Hao, Ma. Xenia G. Ilagan, William Buchser, Joshua B. Rubin, Fuhai LibioRxiv 2020 · DOI

Abstract

Medulloblastoma (MB) is the most common malignant brain tumor in infants and children, categorized into four molecular subtypes: WNT, SHH, Group 3 (G3), and Group 4 (G4). Compared with WNT and SHH subtypes, G3 MBs exhibit significantly worse outcomes and higher metastatic rates, yet effective treatments remain elusive. Furthermore, G3 and G4 MBs show a distinct sex bias, occurring more frequently in males. This study aims to identify potential biomarkers associated with poor survival, high metastasis, and sex bias using an integrative analysis of gene expression, methylation, and copy number variation datasets. We identified a 128-gene network signature that effectively stratifies G3 MB patients and prioritized several FDA-approved drugs as potential therapeutic candidates.

Introduction

Medulloblastoma is not a monolithic disease. Among its subtypes, Group 3 (G3) represents the most significant clinical challenge, often referred to as the "black box" of pediatric oncology. It carries the worst prognosis—a 5-year survival rate of approximately 50%—and is frequently metastatic at the time of diagnosis. Despite its severity, the molecular mechanisms driving G3 remain poorly understood. Current treatments rely on aggressive surgery and toxic chemoradiotherapy, highlighting an urgent need for targeted therapies. This research explores the molecular drivers of G3’s clinical phenotypes, including its notable prevalence in male patients.

Methods and Data Integration

The researchers employed an integrative multi-omics approach, combining three distinct layers of biological data to identify robust biomarkers. The workflow utilized publicly available datasets from the Gene Expression Omnibus (GEO), including GSE85217 (expression), GSE85212 (methylation), and GSE37384 (SNP/Copy number data).

  1. Statistical Filtering: Genes were identified based on significant up-regulation, hypo-methylation, or genomic amplification in G3 patients.
  2. Network Construction: Using the BioGRID protein-protein interaction database and shortest-path algorithms, the authors linked these biomarkers into a connected signaling network.
  3. Drug Repurposing: The network signature was used to query the Connectivity Map (CMAP) and DrugBank to identify drugs capable of reversing the G3 genetic profile.
Figure 1: The network signature of biomarker genes associated with G3 medulloblastoma phenotypes, integrating expression, methylation, and SNP data.

Key Results

Superior Subtyping and Prognosis

The most significant finding is the efficacy of the 128-gene "network signature" compared to whole-genome analysis. The selected biomarkers successfully divided G3 patients into two subgroups with a statistically significant difference in survival (p = 0.012). In contrast, a signature utilizing all genes failed to provide significant separation (p = 0.14).

Metastasis and Sex Bias

The network signature also demonstrated superior performance in predicting metastasis. It identified a high-risk group with a 50% metastasis rate versus a lower-risk group at 33%. Furthermore, the study identified specific biomarkers associated with the higher prevalence of G3 in males, reinforcing that sex bias is a critical biological variable in cancer progression.

Figure 2: Comparison of survival probability, metastasis rates, and gender distribution between the picked network genes and the all-gene signature.

Therapeutic Candidates

The analysis prioritized 12 FDA-approved drugs that target the G3 signaling network. Notable candidates include Valproic acid (targeting PPARG) and Nicardipine. The study also suggested potential synergy between PI3K and HDAC inhibitors, which may be effective in inhibiting MYC-driven tumors, a hallmark of the G3 subtype.

Discussion and Future Directions

This exploratory study demonstrates that G3 medulloblastoma is driven by specific, interconnected signaling circuits rather than isolated mutations. By focusing on a network of 128 genes, the researchers achieved better clinical stratification than whole-genome approaches.

While the computational identification of known drugs like Valproic acid validates the methodology, the findings require "wet lab" validation. Future work will focus on testing these prioritized drug combinations in G3 cell lines and patient-derived xenograft (PDX) models to confirm their therapeutic potential in a clinical context.

Summary updated

Abstract

Abstract Medulloblastoma (MB) is the most common malignant brain tumor in infants and children. Four molecular subtypes of MB are recognized: WNT, SHH, Group 3 (G3), and Group 4 (G4). Compared with WNT and SHH subtypes, G3 MBs exhibit significantly worse outcomes and higher metastatic rates, and there is no effective treatment yet. Moreover, G3 and G4 MBs are much more common in boys than girls, i.e., sex bias, which also plays important roles in cancer prognosis and drug response. However, the molecular mechanism of G3 remains unclear, and there are no well-identified biomarker genes associated with these phenotypes, i.e., worse survival rate, higher metastasis rate, and sex bias. In this exploratory study, we aim to identify potential biomarkers associated with the three phenotypes using integrative analysis of gene expression, methylation and copy number variation datasets. In the results, we identified a set of biomarker genes and linked them into a network signature. The network signature showed better performance in the separation of G3 MB patients into subtypes with a significant difference in terms of the three phenotypes. To identify potentially effective drugs for G3 MBs, a set of drugs with diverse targets were prioritized, which can potentially inhibit the network signature. These drugs or combinations thereof might be effective for G3 treatment.
BibTeX
@misc{guo2020integrative,
  author = {Loreina Guo and Kendall Cornick and Vincent Xu and Tianyao Hao and Ma. Xenia G. Ilagan and William Buchser and Joshua B. Rubin and Fuhai Li},
  title = {{Integrative omics data analysis uncovers biomarker genes and potential candidate drugs for G3 medulloblastoma}},
  year = {2020},
  doi = {10.1101/2020.06.19.122515},
  howpublished = {bioRxiv},
}
Figure