Adjustment of breast cancer expression data for a cell proliferationproliferation metagene (meta-PCNA), derived from genes most correlated with the proliferation marker PCNA, almost entirely abrogated the outcome association of both published and random gene signatures. Additionally, hazard ratio of outcome association strongly correlated with meta-PCNA expression (R2 = 0.9), and more than 50% of the breast cancer transcriptome was correlated with meta-PCNA. [@venet_most_2011]

Definitions

Synthesis

Multiple studies establish that proliferation acts as a pervasive confounding factor in the statistical association between gene expression signatures and clinical outcomes, particularly in cancer contexts such as breast cancer and glioblastoma. The mechanistic basis for this confounding extends beyond classical cell cycle gene expression, as evidenced by findings that removing annotated cell cycle genes from signatures fails to eliminate the proliferation-related confounding effect on outcome associations measured by hazard ratios and other statistical metrics. This suggests that proliferation programs influence broader transcriptome patterns that pervade many gene signatures, including those derived from random gene sets, calling into question whether observed outcome associations genuinely reflect the biological relevance of the specific mechanisms those signatures purportedly represent. The relationship between single-cell heterogeneity in proliferation programs and bulk-tissue signature performance remains unresolved, as does the question of which analytical approaches can adequately control for proliferation’s confounding influence when evaluating the mechanistic significance of gene expression markers.

Bibliography