Oncogenic pathway signatures in human cancers as a guide to targeted therapies in this month's Nature by Bild et al. describes a fairly comprehensive report on how the signature of expression of a group of genes (rather than genes acting individually as has been done in the past) can predict outcomes and subtypes of various cancers. Really here 'pathways' seem to be groups of genes whose expression is most correlated with a cancer outcome and then they are reduced to their dominant principal components. The principal components are used to classify and predict cancer outcomes. To test this approach they move from human to mouse models for these types of cancer and find that the same group of genes in mouse is diagnostic of cancer subtypes. Moving forward they use this approach for identifying metagenes (pathways) involved in cancer survival time, and drug sensitivity.
Overall the paper was very interesting and really showed a great in-depth validation of a bioinformatic approach that was previously published.
The mouse work was a very interesting method of validation that was particularly convincing.