During the three years since the US National Cancer Institute–Food and Drug Administration (NCI-FDA) proteomics group published their seminal1 (but flawed2–4) study using mass spectrometry to profile the serum proteome of ovarian cancer patients, more than 60 published studies have applied similar technology to a wide range of cancers and other diseases. Although some studies have looked directly at tumor tissue, most have targeted easily accessible fluids such as serum. A few studies used highresolution instruments with a matrix-assisted laser desorption and ionization (MALDI) ion source and a time-of-flight (TOF) ion detector. The majority, however, have relied on lower resolution instruments and surface-enhanced laser desorption and ionization (SELDI), a MALDI variant that uses commercially prepared chromatographic surfaces (ProteinChips; Ciphergen, Fremont, CA, USA) to separate proteins. These studies have a common goal: to identify biomarker patterns in the proteome that can be used for diagnosis, prognosis or monitoring of disease. The preponderance of evidence from these studies seems to suggest that proteomic profiling is able to detect anonymous protein peaks that are expressed differently in cancer patients than in healthy individuals5–7. The studies also provide evidence that we may be able to discover patterns that are useful for prediction of the presence or absence of disease8–11. For serum proteomics to realize its full potential, however, several problems in sensitivity and reproducibility remain unresolved. This article outlines some possible improvements in experimental design and data analysis that might help in this regard.