Proteomic Patterns in Biological Fluids: Do They Represent the Future of Cancer Diagnostics?

Abstract

Writing on the future of cancer diagnostics, this author has predicted that multiparametric biomarker analysis, in combination with artificial neural networks and pattern recognition, will likely represent one of the most promising methodologies for diagnosing and monitoring cancer (1)(2). Over the last few years, we have witnessed publication of many reports dealing with proteomic patterns in biological fluids, and especially serum, by using the so-called “SELDI-TOF” technique (surface-enhanced laser desorption/ionization time-of-flight mass spectrometry), in combination with artificial intelligence (3)(4)(5)(6)(7). The reported sensitivities and specificities of this method for ovarian, prostate, and breast cancer diagnosis are clearly impressive, and they are superior to the sensitivities and specificities obtained with current serologic cancer biomarkers (8)(9)(10)(11)(12). In particular, these techniques appear to detect early as well as advanced disease with similar efficiency, making them candidate tools for cancer screening, an application that is not currently recommended, by utilizing the classical cancer biomarkers, e.g., CA125, carcinoembryonic antigen (CEA), and α-fetoprotein (AFP) (1). In addition to scientific journals, these reports have also been presented in international news media and have attracted public attention. Despite of some important shortcomings of these methodologies, criticism has been minimal (13)(14). It seems that the impressive bottom line (very high diagnostic sensitivity and specificity) overshadows potential problems. The recent publication of three reports, from two different research groups, on the use of this technology in the diagnosis of prostate cancer allows for comparison of the data and the methodology and for the presentation of some important questions that have not been adequately addressed. In the following paragraphs, I will focus on some critical questions and provide discussion that could form the basis for further investigations. I will concentrate only on prostate cancer, but the same questions are likely valid for ovarian and other cancers. Adam et al. (10) report 83% sensitivity at 97% specificity for prostate cancer detection, whereas Petricoin et al. (9) report 95% sensitivity at 78–83% specificity. Qu et al. (12) reported 97–100% sensitivity at 97–100% specificity. I consider these data roughly comparable, impressive, and clearly superior to the specificity obtained by prostate-specific antigen (PSA) testing (∼ 25%) at the same sensitivities (15). However, it is surprising that the two groups in three studies obtained these results using different methodologies and distinguishing peaks (Table 1 ). Adam et al. (10) and Qu et al.(12) (the same research group) used an IMAC-Cu metal-binding chip for serum adsorption after evaluating other types of chips, including hydrophobic, ionic, cationic, and metal binding. Petricoin et al. (9) found that a hydrophobic C-16 chip was superior. Furthermore, Adam et al. (10) used nine peaks at m/z ratios of 4475, 5074, 5382, 7024, 7820, 8141, 9149, 9507, and 9656, whereas Petricoin et al. (9) selected different peaks at m/z ratios of 2092, 2367, 2582, 3080, 4819, 5439, and 18220. Qu et al. (12) identified 12 major peaks at m/z ratios of 9656, 9720, 6542, 6797, 6949, 7024, 8067, 8356, 3963, 4080, 7885, and 6991 for differentiating noncancer from cancer and 9 peaks at m/z ratios of 7820, 4580, 7844, 4071, 7054, 5298, 3486, 6099, and 8943 for differentiating healthy individuals from patients with benign prostatic hyperplasia. It should be surprising that none of the peaks identified by Petricoin et al. (9) were identified by either Adam et al.(10) or Qu et al. (12)). Even more surprising is the fact that although Adam et al. (10) and Qu et al.(12) used the same chip for serum extraction and the same instrument for peak identification, their distinguishing peaks are very different. Notably, only two peaks are the same, at m/z ratios of 7024 and 9656 (Table 1 ). Another peak, at a m/z ratio of 7820, was identified by Adam et al. (10) as distinguishing for cancer vs noncancer patients and by Qu et al. (12) as distinguishing between healthy individuals and patients with benign prostatic hyperplasia, but not between noncancer and cancer patients. One (albeit very unlikely) explanation to this finding is that there must be thousands of potential distinguishing peaks in serum and that the chance of two groups finding the same peaks would be very low. Another, more likely explanation is that the methods for extracting these potential molecules from serum are very sensitive to the experimental details or to serum storage conditions, even if the same extraction devices are used. Comparison of three reports for prostate cancer diagnosis based on SELDI-TOF technology. m/z ratios were rounded to whole numbers for simplicity. m/z ratios in bold font represent those identified by Adam et al. (10) and Qu et al. (12) for differentiati