ABSTRACT
The publication of our successes1-4 in using mass spectral patterns to identify patients with ovarian and prostate cancers and animals with premalignant pancreatic cancer has led to increased interest in using these methods to characterize a variety of disease and biological states. Adoption of a pattern discovery and recognition
approach to biomarker identification requires the acceptance of two basic premisesthat there is a need and that a pattern of features in a signal constitute a biomarkeras well as choosing an adequate set of tools. This chapter will present a case for acceptance of both premises, describe some of the methods available, and present the pitfalls associated with the process.
