Paradoxes of Non-Trivial Gene Networks: How Cancer-Causing Mutations Can Appear to Be Cancer-Protective

Abstract

Abnormalities of gene structure or expression are commonly found in cancers, where they are used as prognostic markers, predicting the likely severity of disease or chances of response to therapy. An odds ratio (OR) of <1 indicates that a marker's presence is correlated with better outcome. An OR of <1 is also often taken to mean that the gene concerned has a protective effect in the mechanism of cancer. I show that this is not necessarily so. Modeling of the genes involved in the causation of cancer as a network of weak, failure-prone elements shows that "cancer-causing" genes (i.e., genes whose abnormality is causal in driving cancer) can nevertheless appear as "protective" markers in later stage cancers. This implies that results suggesting that well-known oncogenes have an OR of <1 are quite valid, and that predicting a "protective" role from an apparently protective prognostic value is not valid. I identify mdm-2 and bax as candidates for genes with an apparently protective role through this mechanism.