Breast cancer molecular profiling with single sample predictors: a retrospective analysis.

Area of Science:

• Oncology
• Genomics
• Biostatistics

Background:

• Breast cancer is classified into five molecular subtypes using microarray expression profiling.
• Single Sample Predictors (SSPs) are utilized for individual sample classification.
• Variability exists in the agreement between different microarray-based SSPs.

Purpose of the Study:

• To assess the agreement between three established microarray-based Single Sample Predictors (SSPs).
• To evaluate the consistency of molecular subtype identification across different SSPs in breast cancer datasets.

Main Methods:

• Applied three previously described microarray-based SSPs to one in-house and three public breast cancer datasets.
• Analyzed agreement between SSPs for overall classification and individual molecular subtypes within each cohort.

Main Results:

• Fair-to-substantial agreement (kappa=0.238-0.740) was observed between SSP pairs.
• Only basal-like breast cancer subtypes showed consistent, almost-perfect agreement (kappa>0.812) across SSPs.
• Classification of luminal A, luminal B, HER2, and normal breast-like subtypes varied significantly depending on the SSP used.

Conclusions:

• While SSPs identify molecular subtypes with similar survival outcomes, they do not reliably assign individual patients to the same subtype.
• Standardization of methodologies and definitions for breast cancer molecular subtype identification is essential for clinical integration.