I n cancer, as in life, DNA doesn’t have the last word. Although some gene mutations can predict response to certain targeted therapies, single-gene testing has limits. “Most of biology and probably the large fraction of cancer biology doesn’t occur at a chromosomal level,” said NCI geneticist Ken Buetow, Ph.D., at the 2010 American Association for Cancer Research conference on molecular diagnostics. “It occurs at a network level.” Many alterations in many genes interact to create the cancer phenotype, Buetow said, with the frequency of any single alteration “very rarely above single-digit percentages.” What’s needed, according to Buetow and others, are diagnostic tests that can examine multiple points in these networks simultaneously and use the information to predict how likely a given treatment is to work. “Clearly, we have to move from looking at single molecules and even single pathways to looking at networks and how they interact if we’re going to move forward,” said Gordon Mills, M.D., Ph.D. , chair of the department of systems biology at the M. D. Anderson Cancer Center in Houston. Such pathway and network diagnostic tests may be on the way. Some of the most advanced are designed to detect alterations in proteins, since proteins — especially phosphorylated signaling proteins — more directly reveal the activation state of pathways and networks than do DNA sequence or gene expression. “We give you contextual understanding of the net result of all the physiological, of all the genetic change, and all the epigenetic change, and probably other control system changes,” said David Parkinson, M.D., CEO of South San Francisco, Calif., biotech company Nodality, which is developing some of the new proteomic diagnostic tests. “That’s the level at which drugs work or don’t work.” The success of these methods will depend on the ability of companies to validate the tests ’ effectiveness and reproducibility in clinical trials. For some, such trials are already under way.