12967_2024_Article_5401

Abstract

  • Background:

    • IDH1-wildtype glioblastoma multiforme (IDHwt-GBM) is characterized by heterogeneity and poor prognosis.

    • The aim was to develop multiomics molecular subtypes for improving diagnosis and treatment.

  • Methods:

    • Analyzed data from 184 IDHwt-GBM patients using consensus clustering and machine learning algorithms combining mRNA and MRI data.

    • Validated models across five datasets and created an online interactive system.

    • Assessed clinical impact and molecular associations of identified subtypes.

  • Results:

    • Two molecular subtypes (class 1 & class 2) were identified with distinct survival rates (HR=1.68 for class 1).

    • Class 2 showed better treatment sensitivity to radiotherapy and temozolomide.

    • Notable differences in mutation patterns, pathways related to PCD, and the immune microenvironment between classes.

  • Conclusion:

    • Multiomics clustering effectively stratified IDHwt-GBM patients, providing insight for personalized treatment approaches.

Introduction

  • Glioblastoma Overview:

    • Most lethal primary brain tumor; 15-month median survival; only 5% survive beyond 5 years.

    • Aggressive nature complicates treatment due to genetic diversity and the blood-brain barrier.

  • Historical Context:

    • Classification of glioblastoma evolved from histology-based to molecular characteristics post-2021 WHO classification.

    • Introduces molecular GBM (molGBM) based on molecular abnormalities compared to histologically classified GBM (histGBM).

  • Research Gap:

    • Previous studies had inconsistent survival findings between histGBM and molGBM.

    • This study aims to analyze multiomics data for a thorough understanding of IDHwt-GBM classifications using ten integration strategies.

Materials and Methods

  • Data Collection:

    • IDH1wt-GBM data sourced from TCGA, CGGA, GLASS, CPTAC, UPenn, and UCSF-PDGM datasets.

  • Methods of Analysis:

    • Used consensus clustering to apply advanced algorithms to multiomics data.

    • Employed partition around medoids (PAM) for subtype prediction using external validation cohorts.

    • Bioinformatic analyses measured differences in gene expression, somatic mutations, and therapeutic responses.

Results

Patient Demographics

  • Discovery set:

    • 184 IDH1wt-GBM patients (62% male, median age: 60 years).

    • Supporting cohorts had similar demographics with minor age discrepancies.

Molecular Subtype Discovery

  • Identification of Subtypes:

    • Two subtypes classified (class 1 & class 2) revealed significant differences in overall survival rates.

    • Class 1 showed better prognosis; class 2 more responsive to treatments.

Mutation and Pathway Analysis

  • Somatic Mutations:

    • Class 2 patients often had TP53 mutations and increased amplification at 4q12, correlating with worse survival.

  • Mutational Signatures:

    • Detected various mutational signatures linked to treatment responses, particularly relating to temozolomide efficiency.

  • Pathways:

    • Enrichment in specific pathways like PD-1 checkpoint and ferroptosis highlighted distinctions between subtypes.

Clinical Implications

  • Predictive Modeling:

    • Established survival prediction models based on molecular subtypes integrating age and sex, allowing tailored treatment strategies.

    • Online platform developed for real-time predictions of IDHwt-GBM subtypes.

Discussion

  • Clinical Relevance:

    • Findings contribute to the understanding of GBM tumors in the context of molecular subtyping, influencing future treatment approaches.

    • Emphasizes the need for continuing observation of treatment responses and survival rates in different GBM subtypes.

  • Limitations:

    • Acknowledged the retrospective nature, variation in detection equipment, and sample size limitations across studies.

Conclusion

  • Successfully identified two distinct molecular subtypes of IDHwt-GBM using a comprehensive multiomics approach that directs more personalized treatment options.

  • Online tools for real-time patient profiling could greatly enhance clinical decision-making efficiency.