Integrated Multi-Omics Maps of Lower-Grade Gliomas

Area of Science:

• Computational biology
• Genomics
• Machine learning in oncology

Background:

• Multi-omics data from patient-matched tumor specimens offer fragmented insights.
• Integrative analysis is crucial for a holistic understanding of complex biological systems.
• Existing methods often struggle to combine diverse omics data effectively.

Purpose of the Study:

• To present an intuitive machine learning method for combined analysis of multi-omics data.
• To enable visual comparison of expression, methylation, and copy number variation (CNV) landscapes.
• To apply this method to classify lower-grade gliomas and identify molecular subtypes.

Main Methods:

• Utilized self-organizing maps (SOMs) for machine learning-based integrative analysis.
• Developed a gene-centered coordinate system to portray multi-omics data (expression, methylation, CNV).
• Applied the method to patient-matched tumor specimens from lower-grade gliomas.

Main Results:

• The method visually portrays and compares different omics layers on a personalized tumor basis.
• Identified molecular subtypes of lower-grade gliomas defined by genetic key lesions.
• These subtypes associate with distinct DNA methylation and gene expression patterns, influencing cell fate and prognosis.

Conclusions:

• The integrative multi-omics approach provides a holistic view of tumor biology.
• The method facilitates the classification of gliomas into distinct molecular subtypes with varying prognoses.
• This approach can be extended to integrate additional omics data and identify prognostic markers.