DDT December 2005.qxp

1697 www.drugdiscoverytoday.com Although early detection of pathology can greatly improve patient outcomes, there are relatively few good noninvasive molecular tools for diagnosis of early-stage disease, such as premalignant cancer [1,2], or for prognosis, such as identifying cardiac patients at higher risk of dying (e.g. the Framingham Heart Study), or of non-responsiveness to therapy [3]. The development of satisfactory therapeutics is also hampered by the lack of suitable bioassays that evaluate drug efficacy or toxicity [4]. Therefore, there is currently a pressing need to develop methodology that allows for routine and reliable identification and stringent validation of molecular indicators – biomarkers – in readily accessible patient samples, such as blood, urine or sputum. Such biomarkers could provide a method for clinical monitoring of broad classes of common illnesses (e.g. cancer, heart disease and infection), for assessing the effectiveness of therapeutic intervention regimes and for improving the reliability and accuracy of clinical trials. However, because of considerable disease heterogeneity, interpatient variation and other irrelevant sources of biological variability, biomarker discovery is extremely challenging. Moreover, to a certain extent, biomarker discovery presupposes knowledge of which variables, such as irrelevant epiphenomena, account for confounding factors. Because biochemical responses to disease or drug action are likely to be reflected in the patterns of protein expression and turnover in affected cells, Jennifer Listgarten Department of Computer Science, University of Toronto, Toronto, Ontario, M5S 3G4, Canada Andrew Emili* Program in Proteomics and Bioinformatics, Banting and Best Department of Medical Research, University of Toronto, Toronto, Ontario, M5G 1L6, Canada *e-mail: andrew.emili@ utoronto.ca Practical proteomic biomarker discovery: taking a step back to leap forward