Abstract
Removal of the proliferation component of gene expression by proliferating cell nuclearantigen (PCNA) adjustment via statistical methods has been addressed in numerous survivalprediction studies for breast cancer and all cancers in the Cancer Genome Atlas (TCGA). Thesestudies indicate that the removal of proliferation in gene expression by PCNA adjustment removesthe statistical significance for predicting overall survival (OS) when gene selection is performed ona genome-wide basis. Since cancers become addicted to DNA repair as a result of forced cellularreplication, increased oxidation, and repair deficiencies from oncogenic loss or geneticpolymorphisms, we hypothesized that PCNA adjustment of DNA repair gene expression does notremove statistical significance for OS prediction. The rationale and importance of this translationalhypothesis is that new lists of repair genes which are predictive of OS can be identified to establishnew targets for inhibition therapy. A candidate gene approach was employed using TCGARNA-Seq data for 121 DNA repair genes in 8 molecular pathways to predict OS for 18 cancers.Statistical randomization test results indicate that after PCNA adjustment, OS could be predictedsignificantly by sets of DNA repair genes for 61% (11/18) of the cancers. These findings suggest thatremoval of the proliferation signal in expression by PCNA adjustment does not remove statisticalsignificance for predicting OS. In conclusion, it is likely that previous studies on PCNA adjustmentand survival were biased because genes identified through a genome-wide approach are stronglyco-regulated by proliferation.