Transcriptomic subtypes in high-grade serous ovarian cancer are driven by tumor cellular composition

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• High-grade serous ovarian carcinoma (HGSC) is aggressive, with established transcriptomic subtypes (mesenchymal, immunoreactive, proliferative, differentiated) lacking therapeutic translation.
• The biological basis of these HGSC subtypes remains unclear, hindering clinical application.

Purpose of the Study:

• To investigate whether HGSC transcriptomic subtypes reflect intrinsic malignant programs or tumor cellular composition.
• To re-evaluate the biological underpinnings of HGSC subtypes and their implications for research.

Main Methods:

• Integrated single-cell-derived pseudobulk simulations with deconvolution analysis of 1,834 primary HGSC tumors (RNA-seq and microarray).
• Assessed the predictive power of cellular composition on subtype classification.
• Quantified the contribution of composition to subtype-associated transcriptomic variation.

Main Results:

• HGSC transcriptomic subtypes are primarily determined by tumor cellular composition, particularly stromal and immune variations.
• Cellular composition accurately predicted subtype labels (ROC-AUC = 0.81-0.95) and explained significant transcriptomic variation.
• The mesenchymal (C1.MES) subtype strongly reflects composition-driven signals, while intrinsic signals are secondary.

Conclusions:

• HGSC transcriptomic subtypes represent features of the tumor ecosystem, not distinct malignant states.
• This reinterpretation necessitates revised approaches for inferring tumor biology from subtype labels.
• Provides a framework for disentangling composition-driven and intrinsic signals in bulk tumor data.