Predicting the clinical status of human breast cancer by using gene expression profiles

We have developed Bayesian regression models that provide predictive capability based on gene expression data derived from DNA microarray analysis of a series of primary breast cancer samples

Mike West; Carrie Blanchette; Holly Dressman; Erich Huang; Seiichi Ishida; Rainer Spang; Harry Zuzan; John A. Olson; Jeffrey R. Marks; Joseph R. Nevins


Scholarcy highlights

  • Recent studies demonstrate that gene expression information generated by DNA microarray analysis of human tumors can provide molecular phenotyping that identifies distinct tumor classifications not evident by traditional histopathological methods
  • DNA microarrays, together with the absence of bias in assumptions as to what type of pathway might be affected in a particular tumor, the analysis of gene expression profiles offers the potential to impact clinical decision-making based on more precise determinations of tumor cell phenotypes
  • The analyses presented here further demonstrate that clinically relevant phenotypes can be determined for primary breast tumor samples through the analysis of gene expression
  • Similar studies have used gene expression profiles in out-of-sample crossvalidation studies, and most approaches use some form of initial gene screening to select discriminatory subsets; we have stressed and illustrated the practical importance of repeating such gene screening exercises within each crossvalidation to adequately assess realistic uncertainties about predictions and avoid misleading confidence in predictive accuracy and validity
  • The analysis of breast cancer phenotypes likely represents a context of considerably more biological heterogeneity, reflecting subtle aspects of tumor phenotype
  • It is critical to develop methods, as we report here, that validate classifications with out-of-sample crossvalidation methods, but that provide appropriate and adequate assessments of the inherent uncertainties found with such predictions
  • The finding that the group of genes that contribute most weight to the discrimination includes estrogen receptor and ER pathway genes and genes that encode proteins that synergize with ER, such as HNF3␣ and androgen receptor, points to the potential power of the analysis in identifying functionally significant relationships

Need more features? Save interactive summary cards to your Scholarcy Library.