Prostate Carcinoma Tissue Proteomics for Biomarker Discovery Yaxin Zheng, M.D.1,2 Ye Xu, M.D.1,2 Bin Ye, Ph.D.3 Junyi Lei, M.D.4 Michael H. Weinstein, M.D., Ph.D.4 Michael P. O’Leary, M.D., M.P.H.2 Jerome P. Richie, M.D.1,2 Samuel C. Mok, Ph.D.3 Brian C.-S. Liu, Ph.D.1,2 1
Molecular Urology Laboratory, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts.
Division of Urology, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts.
Department of Obstetrics, Gynecology and Reproductive Biology, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts.
Department of Pathology, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts.
BACKGROUND. The advent of the prostate-speciﬁc antigen (PSA) test has had a profound impact on the diagnosis and treatment of prostate carcinoma. However, the use of PSA levels alone for screening for prostate carcinoma was compromised by the variations in the amount of PSA produced by the benign prostatic tissue specimens. Proteins were involved in various pathways that determine the behavior of a cell. Therefore, information regarding proteins may reveal drug targets and/or markers for early detection. METHODS. The authors used surface-enhanced laser desorption/ionization timeof-ﬂight mass spectrometry to determine the protein proﬁles from fresh tissues of the prostate. Laser capture microdissection was performed to isolate pure populations of cells. RESULTS. The authors identiﬁed a protein with an average m/Z of 24,782.56 ⫾ 107.27 that was correlated with the presence of prostate carcinoma. Furthermore, using laser capture microdissection, they demonstrated that the origin of this protein, which the authors designated PCa-24, was derived from the epithelial cells of the prostate. PCa-24 expression was detected in 16 of 17 (94%) prostate carcinoma specimens but not in paired normal cells. In addition, this protein was not expressed in any of the 12 benign prostatic hyperplasia specimens that were assayed. CONCLUSIONS. PCa-24 may be useful a marker for prostate carcinoma. Cancer 2003;98:2576 – 82. © 2003 American Cancer Society. KEYWORDS: prostate carcinoma, surface-enhanced laser desorption/ionization time-of-ﬂight, proteomics, mass spectrometry.
Supported in part by Grant U01DK63665 from the National Institutes of Health (B.C.-S.L.). Yaxin Zheng and Ye Xu contributed equally to this work. Address for reprints: Brian Liu, Ph.D., Molecular Urology Laboratory, Brigham and Women’s Hospital, 221 Longwood Avenue, BLI 139, Boston, MA 02115; E-mail: [email protected]
Received May 28, 2003; revision received September 8, 2003; accepted September 12, 2003. © 2003 American Cancer Society DOI 10.1002/cncr.11849
rostate carcinoma is the most commonly diagnosed malignancy in American men age ⬎ 50 years. The American Cancer Society estimated that in 2003, 220,900 men would be diagnosed with prostate carcinoma and 28,900 would die of the disease.1 Currently, the standard for detecting prostate carcinoma involves screening for elevated blood levels of prostate-speciﬁc antigen (PSA), digital rectal examination, and needle biopsy of the prostate.2 The advent of the PSA test has had a profound impact on the diagnosis and treatment of prostate carcinoma. However, the use of PSA levels alone for screening for prostate carcinoma is compromised by the varying amount of PSA produced by the benign prostatic tissue. Prostate biopsy should be considered in all patients who have nodules and/or PSA levels greater than 4 ng/mL. However, benign prostatic hyperplasia (BPH) can produce PSA at levels greater than 4 ng/mL. In addition, 35% of the men with organ-conﬁned prostate carcinoma present with PSA levels below 4 ng/mL.2 Therefore, in screening studies using PSA as a single parameter, a positive predictive value of only 47% was achieved.2 This underscores the ﬁnding
Prostate Carcinoma Tissue Proteomics/Zheng et al.
that total PSA level alone is not a perfect marker for prostate carcinoma. The emerging technology of cDNA microarrays provides the ability to perform comparative analyses of mRNA expression of thousands of genes in parallel. Several reports have suggested that the integration of cDNA microarray, linked with clinicopathologic data, may be a powerful approach to the molecular proﬁling of human prostate carcinoma.3,4 However, this kind of analysis measures only the relative abundance of an mRNA encoding a protein, not the actual abundance of the protein itself. It has been shown that there is not always a direct correlation between mRNA expression and protein levels and that proteins can be modiﬁed after translation of the mRNA.5 In the current article, we describe the use of a tissue proteomics approach to understand prostate carcinoma. Using mass spectrometry– based proteomics, we report the expression of a speciﬁc protein that is expressed in tumor epithelial cells, but not in benign or normal cells, from prostate tissue specimens.
MATERIALS AND METHODS Clinical Specimens Seventeen fresh prostate carcinoma specimens with paired normal prostate tissue samples or paired BPH samples were obtained randomly from patients undergoing radical prostatectomy. In addition, ﬁve BPH specimens with an insufﬁcient amount of paired tumor samples for collection were obtained from patients undergoing radical prostatectomy. Two BPH specimens from patients undergoing transurethral resection of the prostate (TURP) for treatment of lower urinary tract obstruction also were collected. Only excess tissue specimens that otherwise would have been discarded were used. These specimens were obtained with approval from the Institutional Review BoardHuman Subject Research Committee at Brigham and Women’s Hospital (Boston, MA).
Sample Preparation Frozen section slides measuring 8 m in thickness, were prepared from each specimen. One section of each specimen was stained with hematoxylin and eosin to conﬁrm the presence and absence of tumor cells. Four sections of each specimen containing prostate tumor cells, paired normal prostate epithelial cells, or BPH samples were lysed directly with 20 –30 L of 20 mM HEPES (pH 7.0) containing 0.1% Nonidet P-40 (all from Sigma, St. Louis, MO). The sections were vortexed for 5 minutes and centrifuged at 14,000 ⫻ g for 1 minute.
FIGURE 1. (A) Specimen 9 (15), from a 57-year-old man who presented with a serum prostate-speciﬁc antigen (PSA) level of 5.4 ng/mL at the time of surgery, a Gleason sum score of 9, and Stage T2b disease. The tumor proﬁle (15T) shows the expression of a protein with an average mass-to-charge ratio (m/Z) of 24,790.5 (asterisk). This protein is not expressed in cells from benign prostatic hyperplasia (15BPH) or from paired normal prostate samples (15N). The epithelial content for this specimen is approximately 90% for the tumor section and 40% for the paired normal sections. (B) Specimen 13 (21), from a 64-year-old man who presented with a serum PSA level of 4.8 ng/mL at the time of surgery, a Gleason sum score of 7, and Stage T2b disease. The tumor proﬁle (21T) shows the expression of a protein with an average m/Z of 24,799.1 (asterisk). This protein is not expressed in cells from benign prostatic hyperplasia (21BPH; arrow) or from paired normal prostate samples (21N). The epithelial content for this specimen is approximately 10% for the tumor specimen and 40% for the paired normal specimen. Laser Capture Microdissection To determine the cellular origin of the protein of interest, four paired samples were microdissected by laser capture microdissection (LCM; PixCell II LCM system, Arcturus, Mountain View, CA) as previously described.6 – 8 For each paired sample, approximately 30,000 –50,000 cells (5000 –10,000 shorts) were procured within an approximate time of 1 hour. Micro-
CANCER December 15, 2003 / Volume 98 / Number 12 TABLE 1 Clinical Information on Samples from Radical Prostatectomya Sample no. (internal code)
Epithelial content (%) (T/N)
1 (1) 2 (4) 3 (5) 4 (6) 5 (10) 6 (12) 7 (13) 8 (14)b 9 (15) 10 (18) 11 (19) 12 (20) 13 (21) 14 (22) 15 (51) 16 (52) 17 (53)
53 54 50 60 61 53 47 44 57 62 60 61 64 50 51 69 48
5.9 8.8 ⬍ 0.1 5.1 10.4 6.3 10.0 7.8 5.4 0.7 4.5 10.3 4.8 8.4 3 5.9 9.3
3⫹3 4⫹5 3⫹3 3⫹4 3⫹4 3⫹3 3⫹4 3⫹3 4⫹5 4⫹3 3⫹3 3⫹3 4⫹3 4⫹5 3⫹4 3⫹3 3⫹4
T2b T2b T3a T2b T2b T2b T3a T3a T2b T2a T2b T2b T2b T3b T2b T2b T2b
24,748.9 24,930.8 24,714.1 24,757.2 25,057.0 24,724.0 24,650.3 None 24,790.5 24,684.3 24,753.2 24,794.9 24,799.1 24,782.5 24,836.6 24,685.4 24,930.8
⫹/⫺ ⫹/⫺ ⫹/⫺ ⫹/⫺ ⫹/⫺ ⫹/⫺ ⫹/⫺ ⫺/⫺ ⫹/⫺ ⫹/⫺ ⫺/⫺ ⫺/⫺ ⫹/⫺ ⫹/⫺ ⫹/⫺ ⫹/⫺ ⫹/⫺
80/40 85/50 50/40 90/60 90/70 90/80 60/10 70/70 90/40 5/20 50/40 30/40 10/40 50/60 90/15 80/40 10/30
T: tumor; N: normal; ⫹: positive; ⫺: negative; TRAP: telomeric repeat ampliﬁcation protocol; SELDI-TOF: surface-enhanced laser desorption/ionization time of ﬂight mass spectrometry; m/Z: mass-to-charge ratio. a The average protein mass-to-charge ratio in positive specimens was 24,782.56 ⫾ 107.27 daltons. b Sample 8 is the only specimen that did not exhibit the SELDI peak.
dissected cells were resuspended with 10 L lysis buffer as described earlier (see “Sample Preparation”), vortexed for 5 minutes, and centrifuged at 14,000 ⫻ g for 1 minute.
Surface-Enhanced Laser Desorption/Ionization Time-ofFlight Analysis Surface-enhanced laser desorption/ionization timeof-ﬂight (SELDI-TOF) analysis was performed with the immobilized metal afﬁnity capture (IMAC) chip (Ciphergen Biosystems, Fremont, CA). The bait surfaces on the IMAC chip were pretreated with 10 L of 50 mM CuSO4 as described by the manufacture. Five microliters of each lysate then was applied to the bait surface. The analytes were allowed to incubate for 1 hour and then washed twice for 5 minutes each in binding buffer (0.1 M sodium phosphate, 0.5 M sodium chloride, 10 mM of imidazol; Sigma) to remove weakly bound proteins. This step was followed by the addition of 0.5 L of energy-absorbing molecules containing saturated sinapinic acid in 0.5% triﬂuoroacetic acid/50% acetonitrile (all from Sigma) to the washed surface. After allowing it to dry, the procedure was repeated once. The solution then was allowed to crystallize, causing the retained proteins to become embedded within the matrix. Next, SELDI-TOF mass spectrometry was performed on the ProteinChip sys-
tem as described by the manufacturer (Ciphergen). Mass resolution and accuracy were assessed by routine calibration with 5733.58 and 12,230.92-dalton (D) polypeptides. The protein chips were read and analyzed using the following settings: a) a laser intensity of 250; b) a detection sensitivity of 10; c) 50 shots per sample; and d) auto-identify peaks 3000 –50,000 D. We achieved a mass accuracy of 0.1% for proteins and polypeptides with mass-to-charge ratios (m/Z) of 3000 –30,000 in this system.
Sample Conﬁrmation with Telomerase Activity To conﬁrm that the captured cells and/or tissue samples contained viable proteins, as well as to conﬁrm the presence and absence of tumor cells, we used telomerase activity as a marker. The telomeric repeat ampliﬁcation protocol (TRAP) assay was used to determine telomerase activity.9,10 A detailed protocol for the TRAP assay is described elsewhere.9,10
RESULTS Using SELDI-TOF mass spectrometry, we report the expression of a speciﬁc protein that is correlated with the presence and absence of prostate carcinoma. Figure 1 shows the position of a protein peak to be approximately m/Z ⫽ 24,749 –24,790 when proﬁled by SELDI-TOF. In 16 of 17 (94%) prostate carcinoma
Prostate Carcinoma Tissue Proteomics/Zheng et al.
FIGURE 2. (A) Specimen 1 (1), from a 53-year-old man who presented with a serum prostate-speciﬁc antigen (PSA) level of 5.9 ng/mL at the time of surgery, a Gleason sum score of 6, and Stage T2b disease. The top two proﬁles are from undissected tissue specimens (tumor and paired normal tissue specimens, respectively), and the bottom two proﬁles are from prostate epithelial cells (tumor and paired normal cells, respectively) captured by laser capture microdissection (LCM). The expression of a protein with an average mass-to-charge ratio (m/Z) of 24,967.6 in the captured prostate epithelial cells (asterisk) is observed in the tumor proﬁle but is absent from the paired normal proﬁle. The undissected tumor proﬁle shows the expression of a protein with an average m/Z of 24,748.9 (asterisk). (B) Specimen 4 (6), from a 60-year-old man who presented with a serum PSA level of 5.1 ng/mL at the time of surgery, a Gleason sum score of 7, and Stage T2b disease. The top two proﬁles are from undissected tissue specimens (tumor and paired normal tissue specimens, respectively), and the bottom two proﬁles are from prostate epithelial cells (tumor and paired normal cells, respectively) captured by LCM. The expression of a protein with an average m/Z of 24,862.4 in the captured prostate epithelial cells (arrow) is observed in the tumor proﬁle but is absent from the paired normal proﬁle. The undissected tumor proﬁle shows the expression of a protein with an average m/Z of 24,757.2 (arrow).
specimens, only the tumor cells express this protein (Table 1). The average m/Z of the protein in all tumor samples is 24,782.56 ⫾ 107.27 (Table 1). To determine the origin of this protein, we performed LCM to capture pure populations of cells from the tissue specimens. Figure 2 demonstrates that this protein originated in malignant prostate epithelial cells. Figure 3 shows a composite proﬁle of three paired tumor and normal prostate specimens. The expression of a protein peak at m/Z ⫽ 24,730.1 ⫹ H–24,761.2 ⫹ H is observed in the tumor proﬁles but not in the paired normal proﬁles (Fig. 3). Using the ProteinChip software, this peak was found to be signiﬁcant in discriminating between tumor and normal prostate specimens (Fig. 3). To determine whether BPH expresses this protein, we proﬁled 12 BPH specimens. Of the 12 BPH specimens, 3 (Samples 5, 16, and 17 in Table 1) have paired prostate tumors and 2 (Samples 9 and 13 in Table 1) have paired prostate tumor as well as paired normal cells. Another ﬁve BPH specimens (Samples B1–B5), also obtained from patients who underwent radical prostatectomy, did not contain a sufﬁcient amount of paired tumor cells for sample collection (Table 2). Finally, two BPH specimens were obtained from patients who underwent TURP for treatment of lower urinary tract obstruction (Table 2). Our results demonstrate that none of the 12 BPH specimens express this protein (Fig. 1). To conﬁrm that the tissue sections contain tumor cells, as well as to demonstrate the integrity of the proteins in the cancer specimens, we used the TRAP assay to determine telomerase activity. Telomerase activity was found in 14 of 17 (82%) tumor specimens but was absent in the paired normal cells (Table 1 and Fig. 4). Furthermore, no telomerase activity was found in the 12 BPH specimens.
DISCUSSION The goal of proteomic analysis of biologic samples is to determine the overall set of proteins that are important in normal cellular physiology or altered by disease process. SELDI-TOF mass spectrometry provides a major advance in protein proﬁling. This approach uses ProteinChip technology to selectively capture subsets of proteins from extracts using various chromatographic surfaces coupled with SELDI-TOF mass spectrometry to generate a mass spectrum of the sample being analyzed.11,12 Using a combination of different ProteinChip surfaces and tissue extraction buffers, a proﬁle of several hundred proteins can be rapidly resolved for a particular sample. SELDI is sensitive to the femtomole/attomole level and uses much
CANCER December 15, 2003 / Volume 98 / Number 12
FIGURE 3. Composite proﬁle of three paired tumor and normal prostate specimens. Note that the expression of a protein peak at mass-to-charge ratio (m/Z) ⫽ 24,730.1⫹ H–24,761.2 ⫹ H (arrows) is observed in the tumor proﬁles but not in the paired normal proﬁles. Using the ProteinChip software from Ciphergen, this peak was found to be signiﬁcant in discriminating between tumor and normal prostate specimens. Open circles: prostate carcinoma; open squares: normal tissue specimen. TABLE 2 Clinical Information on Benign Prostatic Hyperplasia Sample no.
B1 B2 B3 B4 B5 B6 (TURP) B7 (TURP)
58 70 58 60 61 62 88
6.2 8.6 3.4 7.2 5.6 3.8 12.8
3⫹3 4⫹5 3⫹3 3⫹3 3⫹3 N/A N/A
T2a T2a T2a T2a T2b N/A N/A
PSA: prostate-speciﬁc antigen; TURP: transurethral resection of the prostate.
smaller amounts of material than other techniques, such as two-dimensional (2-D) gel electrophoresis.11,12 The underlying principle in SELDI is its surfaceenhanced afﬁnity capture through the use of protein chips consisting of chemical or biologic surfaces that bind proteins. The use of these protein chips enables both biomarker discovery and protein proﬁling directly from sample source without preprocessing. This technology has been used to detect known markers of prostate carcinoma, i.e., PSA, prostataic acid phosphatase (PAP), prostate-speciﬁc membrane antigen, and
for discovery of potential markers that are either overexpressed or underexpressed in prostate carcinoma cells and body ﬂuids.13,14 PSA is an approximately 33 kD serine protease produced by the epithelium of the prostatic gland.2 Serum levels of PSA are elevated in men with an enlarged prostate and in men with prostate carcinoma. In men with enlarged prostates, serum PSA levels have been assumed to be elevated because of increased epithelial mass, local tissue destruction, and concomitant leakage of PSA into the systemic circulation. Unfortunately, BPH can produce PSA at a signiﬁcant level in the serum, i.e., 4 –10 ng/mL, which prevents the distinction between BPH and early prostate carcinoma.2 In the current report, we have described the presence of what appears to be a unique prostate carcinoma–associated protein, which we will now call PCa-24. This protein, with an m/Z of 24,782.56 ⫾ 107.27, is expressed in the cells of 16 of 17 (94%) prostate carcinoma specimens (Table 1) but was not expressed in any specimens containing BPH or normal prostate cells (Fig. 1). Furthermore, by using LCM, we determined that PCa-24 originated from prostate epithelial cells (Fig. 2). The reason for the absence of this protein in one prostate carcinoma specimen (Sample 8) currently is unknown. To conﬁrm the integrity of the proteins in the tumor specimens, we used telomerase activity
Prostate Carcinoma Tissue Proteomics/Zheng et al.
FIGURE 4. Gel illustrating the telomeric repeat ampliﬁcation protocol assay for telomerase activity. The prostate carcinoma cell line LNCaP was used as the positive control. Note the laddering pattern of telomeric repeats that demonstrates a functional telomerase enzyme. Specimens 9 and 13 are shown on the gel. Note that telomere extension by the telomerase enzyme is detected in the tumor specimens (T) but not in the normal tissue specimens (N) or in areas with benign prostatic hyperplasia (BPH).
for characterization studies. However, our attempts to identify this protein based on its mass, using the TagIdent tool18 from the ExPASy molecular biology server, yielded no deﬁnitive matches. One possible application for our ﬁndings involves determining the presence or absence of this protein in the serum of patients with prostate carcinoma versus those with other urologic diseases, including BPH. Work along this line is underway. Another potential application for our ﬁndings is the development of a minimal invasive assay for the detection of prostate carcinoma in patients with elevated PSA levels. Currently, prostate biopsy should be considered in all patients who have nodules and/or PSA levels greater than 4 ng/mL. However, as many as 50% of prostate nodules are not malignant as determined by biopsy results, but biopsy is the only means to differentiate cancer from other causes of these nodules. Perhaps the expression of the PCa-24 protein, coupled with recent ﬁndings of protein patterns in serum,19,20 will allow us to distinguish between these patients and spare them from unnecessary biopsies.
as a marker. Several studies have demonstrated that telomerase is expressed in cells from human prostate carcinoma but not in cells from normal or BPH specimens.15–17 The detection of telomerase in these three studies ranged from 80% to 92%. The TRAP assay for telomerase activity requires not only that the enzyme be present but also that its function be maintained.9,10 Therefore, the TRAP assay for telomerase activity is an ideal method for conﬁrming the integrity of proteins in our tumor specimens. In 14 of 17 specimens, we were able to detect telomerase activity with the TRAP assay (Table 1, Fig. 4). In Sample 8, however, we were not able to detect telomerase activity by the TRAP assay, despite the presence of tumor cells (Table 1). One possibility regarding the presumably false-negative expression of PCa-24 in this sample is that protein degradation and/or modiﬁcation occurred. With our proﬁles, even with as little as 5% epithelial content in a tumor specimen (Sample 10 in Table 1), we were able to detect the presence of PCa-24. Figure 1B further illustrates our ability to detect the presence of PCa-24 in a sample that contains approximately 10% epithelial content in the tumor specimen. The exact nature of the PCa-24 protein currently is not known. We are in the process of purifying this protein
Jemal A, Murray T, Samuels A, Ghafoar A, Ward E, Thun MJ. Cancer statistics, 2003. CA Cancer J Clin. 2003;53:5–26. 2. Brawer MK. Prostate-speciﬁc antigen: current status. CA Cancer J Clin. 1999;49:264 –282. 3. Dhanasekaran SM, Barrette TR, Ghosh D, et al. Delineation of prognostic biomarkers in prostate cancer. Nature. 2001; 412:822– 826. 4. Singh D, Febbo PG, Ross K, et al. Gene expression correlates of clinical prostate cancer behavior. Cancer Cell. 2002;1:203– 209. 5. Anderson L, Seilhamer J. A comparison of selected mRNA and protein abundances in human liver. Electrophoresis. 1997;18:533–537. 6. Emmert-Buck MR, Bonner RF, Smith PD, et al. Laser capture microdissection. Science. 1996;274:998 –1001. 7. Bonner RF, Emmert-Buck MR, Cole KA, et al. Laser capture microdissection: molecular analysis of tissue. Science. 1997; 278:1481–1483. 8. Liu BC, LaRose I, Weinstein LJ, Ahn M, Weinstein MH, Richie JP. Expression of telomerase subunits in normal and neoplastic prostate epithelial cells isolated by laser capture microdissection. Cancer. 2001;92:1943–1948. 9. Kim NW, Piatyszek MA, Prowse KR, et al. Speciﬁc association of human telomerase activity with immortal cells and cancer. Science. 1994;266:2011–2015. 10. Kavaler E, Landman J, Chang Y, Droller MJ, Liu BC. Detecting human bladder carcinoma cells in voided urine samples by assaying for the presence of telomerase activity. Cancer. 1998;82:708 –714. 11. Weinberger SR, Dalmasso EA, Fung ET. Current achievements using ProteinChip Array technology. Curr Opin Chem Biol. 2001;6:86 –91.
CANCER December 15, 2003 / Volume 98 / Number 12
12. Paweletz CP, Gillespie JW, Ornstein DK, et al. Rapid protein display proﬁling of cancer progression directly from human tissue using a protein biochip. Drug Dev Res. 2000;49:34 – 42. 13. Adam BL, Vlahou A, Semmes OJ, Wright GL Jr. Proteomic approaches to biomarker discovery in prostate and bladder cancers. Proteomics. 2001;1:1264 –1270. 14. Wang S, Diamond DL, Hass GM, Sokoloff R, Vessella RL. Identiﬁcation of prostate speciﬁc membrane antigen (PSMA) as the target of monoclonal antibody 107-1A4 by ProteinChip Array, surface-enhanced laser desorption/ionization (SELDI). Int J Cancer. 2001;92:871– 876. 15. Lin Y, Uemura H, Fujinami K, Hosaka M, Harada M, Kubota Y. Telomerase activity in primary prostate cancer. J Urol. 1997;157:1161–1165. 16. Takahashi C, Miyagawa I, Kumano S, Oshimura M. Detec-
tion of telomerase activity in prostate cancer by needle biopsy. Eur Urol. 1997;32:494 – 498. Orlando C, Gelmini S, Selli C, Pazzagli M. Telomerase in urological malignancy. J Urol. 2001;166:666 – 673. Swiss Institute of Bioinformatics. TagIdent tool [database online]. Available from URL: http://expasy.ch/tools/ tagident.html [accessed 10 May 2003]. Petricoin EF III, Ornstein DK, Paweletz CP, et al. Serum proteomic patterns for detection of prostate cancer. J Natl Cancer Inst. 2002;94:1576 –1578. Qu Y, Adam BL, Yasui Y, et al. Boosted decision tree analysis of surface-enhanced laser desorption/ionization mass spectral serum proﬁles discriminates prostate cancer from noncancer patients. Clin Chem. 2002;48:1835– 1843.