| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Article |
From the Department of Surgery (SB, CL) and Departments of Microbiology and Molecular Cell Biology (LHC, OJS, RRD), Eastern Virginia Medical School, and Virginia Prostate Center (LHC, OJS, RRD), Norfolk, Virginia; and the Department of Preventative Medicine (PW, HL), Creighton University, Omaha, Nebraska.
ABSTRACT
Background: BRCA-1 mutations predispose women to early onset breast cancer, but ~20% never develop cancer. Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) profiling can differentiate protein signatures of cancer and normal subjects. Our objective was to distinguish women with BRCA-1 mutations who developed breast cancer (BRCA-1 Ca) from those who did not (Carrier), normal volunteers (NL), and women with sporadic breast cancer (SBC), using SELDI-TOF.
Methods: Baseline serum specimens were obtained from women with BRCA-1 mutations without cancer, SBC, and NL. BRCA-1 women were later divided into two cohorts, pending cancer development. The sera were spotted onto protein chips for SELDI-TOF analysis and analyzed with classification algorithm software.
Results: BRCA-1 Ca patients (n = 15) developed cancer within 3 years of baseline, while BRCA-1 carriers (n = 15) were cancer-free in 7 years of follow-up. SELDI-TOF analysis revealed differentially expressed proteins (P < .05) between BRCA-1 Ca, Carrier, and SBC patients (n = 16), such that 13/15 BRCA-1 Ca vs. Carrier women were correctly identified (sensitivity/specificity of 87%/87%) and 14/15 BRCA-1 Ca vs. SBC patients were correctly identified (sensitivity/specificity 94%/100%). Profiles of Carriers resembled NL profiles (n = 16).
Conclusions: SELDI-TOF protein profiles from this small pilot study distinguished between women with BRCA-1 Ca, Carriers, and women with SBC. Whether BRCA-1 Ca represents earlier detection of occult cancer or other risk factors is unknown. Follow-up studies with larger numbers and longer follow-up are required to validate these findings but may allow more timely prophylactic or therapeutic strategies.
Key Words: BRCA-1 • Breast cancer • Protein profiling • Serum
Breast cancer is the most common malignancy that affects women in the United States today. In fact, one in nine women can expect to develop the disease within their lifetime. Although most causes of breast cancer are sporadic, approximately 7% to 10% of them are attributed to hereditary syndromes, with BRCA-1 mutations accounting for most.1 The BRCA-1 gene itself was only recently discovered by Hall2 and is the focus of ongoing research to define its functional tumor suppressor role in the DNA damage response pathways. However, two clinical manifestations for carriers of this gene mutation are quite clear: early onset breast cancer and lack of 100% penetrance of disease occurrence. Thus, for BRCA-1 mutation carriers, the challenge is to identify which women will develop breast cancer and to do so at an age at which our "gold standard" for cancer detection, mammography, is found to be least accurate. Conversely, the need exists to identify BRCA-1 mutation carriers at low risk for cancer.
Proteomic profiling of clinical body fluids to differentiate protein expression patterns characteristic of a cancerous or noncancerous state is an evolving field of study that may offer a solution. These differentially expressed proteins could be from the cancer itself, i.e., CA 27.29, CEA, or Her2/neu, or from the host response. With use of surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS), protein profiles have been generated that can accurately discriminate between cancerous and noncancerous samples for many cancer types.3–7 To date, however, there have been limited studies of its utility in breast cancer, with the first report originating from Li et al.,8 who utilized SELDI-TOF to discriminate breast cancer from benign breast disease with a sensitivity and specificity of 93% and 91%, respectively. In a similarly small cohort, Vlahou et al.9 demonstrated the power of SELDI-TOF MS to discriminate breast cancer patients from normal healthy volunteers. These studies illustrated the potential of proteomic profiling to discriminate between cancerous and noncancerous states with reasonable sensitivity and specificity.
Given the fact that up to 20% of women with BRCA-1 mutations will not develop breast cancer10 and the recent success of SELDI-TOF MS differentiation in breast cancer, we hypothesized that a proteomic profile could be generated that would allow for accurate discrimination between BRCA-1 mutation carriers at high risk for breast cancer and carriers at lower risk, as well as differentiate BRCA-1 breast cancers from sporadic breast cancers and normal healthy volunteers.
METHODS
Specimens
Following institutional review board (IRB) approval, age-matched baseline serum specimens were identified from a collection of sera obtained during genetic testing at the Hereditary Cancer Institute at Creighton University from 1987 to 2001. All samples were retrospectively labeled from women found to have germ-line BRCA-1 mutations who were without evidence of breast cancer at the time of the baseline sample. The BRCA-1 women were followed until the development of breast cancer or for at least 7 cancer-free years and subsequently divided into two cohorts according to the presence or absence of breast cancer development (BRCA-1 Cancer and Carrier cohorts).
At Eastern Virginia Medical School, women requiring an operative biopsy for an abnormal mammogram or clinical breast examination and normal healthy volunteers were eligible to participate in a study for breast cancer protein profiling. They were enrolled through the Division of Surgical Oncology after signing an IRB-approved consent form. For those women having a diagnostic biopsy, the blood sample was retrospectively labeled as "benign" or "cancer" once the pathologic evaluation was complete. Serum samples were randomly collected from healthy females during the same time period in which the sporadic breast cancer specimens were obtained. All samples were collected by venipuncture into a 10-cc SST Vacutainer tube (Becton Dickinson) and allowed to clot at 4°C for 30 minutes. Coagulated blood was spun at 3000 rpm for 10 minutes, and the serum portion was immediately aliquoted and frozen for storage at –80°C.
SELDI Processing Of Serum Samples
Serum samples were processed robotically on a Biomek 2000 liquid handling system in a 96-well format for SELDI analysis (Biomek 2000; Beckman Coulter, Fullerton, CA) in the following manner. In brief, 20 µL of serum was pretreated with 8 M urea and 1% CHAPS and was vortexed for 10 minutes at 4°C. A further dilution was made in 1 M urea, 0.125% CHAPS, and PBS. Each sample position was randomized and spotted in duplicate onto copper-coated immobilized metal affinity (IMAC-Cu). ProteinChips were used for SELDI-TOF analysis with the aid of a bioprocessor. The protein chips were then incubated at room temperature for 30 minutes, followed by washes of PBS and water. The IMAC-Cu chip arrays were allowed to air dry, and a saturated solution of sinapinic acid in 50% (v/v) acetonitrile and 0.5% (v/v) trifluoroacetic acid was added to each spot. The protein chip arrays were analyzed with the SELDI ProteinChip System (PBS-II, Ciphergen Biosystems, Fremont, CA). The spectra were generated by the accumulation of 192 shots at laser intensity 220 in a positive mode. The protein masses were calibrated externally with use of purified peptide standards.
Afterward, each protein peak was labeled and its intensity was normalized for total ion current to account for variation in ionization efficiencies. Peak clustering was performed with Biomarker Wizard Software (Ciphergen Biosystems) at settings that provide a 5% minimum peak threshold, 0.2% mass window, and 1 to 3 signal/noise determination. The peak intensities from duplicate samples were then averaged.
SELDI Data Analysis
Pattern recognition and sample classification were performed with the Biomarker Pattern Software (Ciphergen Biosystems), in which multiple decision trees are initially generated with use of all the peaks as variables. During the analysis, a pruning step occurs in which branches are removed and the cost of the removal determined to establish a minimal tree size. This is referred to as a learning set. Second, the decision tree was subjected to cross-validation. In this step the data were partitioned such that randomly selected samples were categorized with the decision tree being tested, to ensure that the decision tree was valid. The number of samples used to test the tree during cross-validation was increased from 10 to 20. While choosing a smaller number of samples (e.g., 5) to test the tree can result in overfitting the data, increasing the number of samples used to test the tree is an acceptable means of ensuring the validity of the splitters in a small sample population.
The peaks that formed the main splitters of the tree(s) with the highest prediction rates in the cross-validation analysis were then selected and used to make a final decision tree with the greatest possible predictive power.
RESULTS
Thirty age-matched baseline serum specimens (± 3 years; except the oldest patient with cancer, who was age-matched within 6 years) were collected from women determined to have a BRCA-1 mutation during genetic testing at the Hereditary Cancer Institute at Creighton University from 1987 to 2001. No demographic information, in particular any relation to Ashkenazi Jewish descent, is available on these samples. They were subsequently followed with 15 women developing breast cancer within 3 years and the other 15 women remaining cancer-free for the 7 years of follow-up. The mean time to diagnosis of breast cancer from the baseline serum sample was 1.3 years (range = 1 to 36 months). The mean age of the women who developed cancer at the time of baseline serum sampling was 44.2 years, and the mean age was 44.6 years among those BRCA-1 Carriers who did not develop breast cancer (age range for cancer = 29–74 years; age range for noncancer = 29–68 years). All the cancers identified were found to be an invasive ductal histology with the following breakdown of stages: stage I (n = 8), stage II (n = 3), stage I or II (n = 3), and stage III (n = 1). The three patients who were classified as stage I or II had incomplete staging information: one woman had a primary tumor <2 cm but the lymph node status was unknown, and two women were node-negative but the primary tumor size was <5 cm.
At Eastern Virginia Medical School, 16 sera samples were obtained from normal healthy volunteers and 16 samples were obtained from patients with sporadic breast cancer diagnosed during the same time frame, from 2001 to 2003. The mean age at sample procurement for the healthy volunteers was 45 years (range = 18 to 81 years), and for the women with sporadic breast cancer it was 52.2 years (range: 31 to 67 years). No attempt was made to age-match-control the samples between the two institutions. In the sporadic breast cancer group, all samples were from women found to have invasive ductal carcinoma of stage I or II at diagnosis. Thirteen of the 16 were sentinel lymph node–negative, while three were sentinel lymph node–positive but had a small primary lesion.
The first sera groups analyzed were from the 15 BRCA-1 noncancer subjects versus the 15 individuals who developed BRCA-1 breast cancers during follow-up. Aliquots of diluted serum were applied to copper-coated affinity ProteinChips. The resulting SELDI-TOF MS data were analyzed with Biomarker Wizard and generated a total of 107 peaks, representing differentially expressed proteins in the 1500-Da to 100,000-Da range. Of these, 23 protein peaks were significantly (P < .05) overexpressed in the BRCA-1 cancer cohort, 19 of which were in the 6500-Da to 9200-Da molecular weight range, as listed in Table 1. Differentially expressed lower-mass peptides (<6200 Da) tended to be present at higher levels in the BRCA-1 noncancer samples (Table 1). Examples of the peak intensities in all samples for two of the most differentially expressed peaks are shown in the scatter plots in Figure 1. The protein peak at 8138-Da was overexpressed in the BRCA-1 cancer cohort (Fig. 1A), and a protein at 5909 was underexpressed relative to the BRCA-1 noncancers (Fig. 1B).
|
|
The peak intensity values of the 107 differentially expressed peaks were collectively applied to the Biomarker patterns algorithm program to generate a representative decision classification tree. As shown in Figure 2, the resulting decision tree utilized the peak at 5909 Da as a primary splitter in the classification process. In the learning data set, all 30 samples were correctly classified as either BRCA-1 with breast cancer or BCRA1 carrier without breast cancer, for a sensitivity and specificity of 100%. Cross-validation analysis of this final decision tree correctly identified 13 of the 15 BRCA-1 with breast cancer and 13 of the 15 BRCA-1 carriers. This resulted in a final sensitivity of 87% and a specificity of 87% (Table 2).
|
|
Next, the serum protein profiles from the BRCA-1 carriers who developed cancer (n = 15) were compared with the serum profiles obtained from 16 women with sporadic breast cancers. With use of the copper-coated affinity ProteinChips, a total of 122 peaks were generated with the SELDI-TOF mass spectrometer. Thirty-five of these proteins were overexpressed in the BRCA-1 carrier-with-cancer group, in comparison with the sporadic breast cancer cohort, all in the 6000-Da to 15,000-Da mass range. Again, a protein peak of 8.1 kDa was identified that was highly overexpressed in the BRCA-1 cancer samples in comparison with the sporadic breast cancer cohort. The decision tree generated from these protein peaks was able to correctly identify 14 of the 15 BRCA-1 breast cancer samples and all 16 sporadic breast cancers, for a sensitivity of 94% and specificity of 100% on cross-validation. For visual comparison, the spectra ranging from 2000 to 10,000 Da from representative samples of each group are shown in Figure 3.
|
Finally, a comparison of the normal healthy volunteers to the other three cohorts was performed, for several reasons. A variety of factors, namely the number of freeze/thaws and sample storage, can affect the protein profile of a sample and thus create the illusion of differences in the spectra generated. Because the samples were obtained from two geographically separate locations, it was necessary to evaluate whether any apparent institutional differences were present. Thus, the serum from the BRCA-1 noncancer cohort was compared with serum obtained from 15 normal healthy volunteers. The protein spectra were interchangeable to such a degree that accurate discrimination was not possible without overfitting the data (Fig. 3). Additionally, the spectra appeared quite different from both the BRCA-1 cancer and sporadic cancer cohorts (Fig. 3) and confirmed prior results of distinction of normal from sporadic breast cancer.9 Consistent with the previous study,9 the 15 women with BRCA-1 cancer were separated from the normal healthy volunteers with a sensitivity and specificity of 87% and 94%, respectively.
DISCUSSION
Over 211,000 new cases of breast cancer are expected in the United States this year, making breast cancer the second leading cause of cancer death. Genetic mutations are known to increase a woman’s risk for cancer development by 10-fold and account for 7% to 10% of all breast cancer cases, with the BRCA-1 gene mutations being the most common (52%).1
The BRCA-1 gene was first identified by Hall in 1990 and subsequently was found to be a tumor suppressor.2 This gene, located on the long arm of chromosome 17, encodes a 190-kDa protein that functions as part of a large enzymatic complex. Although all the functions of BRCA-1 are not known, it has been demonstrated to play a role in DNA repair, cell cycle regulation, and transcription regulation. There are over 250 known mutations in the BRCA-1 gene. Although the gene is inherited in an autosomal dominant fashion, a second mutation in the wild-type gene is required for carcinogenesis. Clinically, King et al. showed that BRCA-1 mutations are more common in Ashkenazi Jewish women and overall mutation carriers have a 82% lifetime risk of developing breast cancer, compared with 13% in the general population.1,11 Furthermore, the BRCA-1 mutation confers onset of disease at an earlier age (37% of cases occur by age 50 years), with a higher incidence of recurrent, multifocal, bilateral, and atypical breast carcinomas, i.e., medullary.10 Screening for BRCA-1 mutation is currently recommended for women with a family history of (1) early onset breast cancer (i.e., before age 50 years); (2) breast and ovarian cancer, especially if in the same person; (3) bilateral breast cancer; and (4) male breast cancer. Once a BRCA-1 mutation is identified, determining which carriers are at the highest risk for breast cancer is not currently possible.
SELDI-TOF MS may offer a solution to this problem, as it has several potential advantages as a clinical assay. Using readily accessible serum samples, this technology has proven to be reproducible, has adequate throughput, poses minimal risk to patients, and is relatively inexpensive.3–7 To date, serum protein profiling with SELDI-TOF MS has been applied to early detection of cancer studies, with minimal application to prognostic assay evaluation. In particular, in prostate cancer, our laboratory was able to distinguish invasive prostate cancer from benign prostatic hypertrophy and normal healthy volunteers, using sera that were stratified for PSA levels.5–6,12 These results indicated a role for SELDI-TOF MS in the early detection of malignancy with use of minimally invasive techniques (phlebotomy). However, these highly successful results have not been replicated in breast cancer, probably reflecting the vast heterogeneity of breast cancer, including inherited subtypes.8,9 In fact, the striking differences in the clinical manifestations of BRCA-1 mutated breast cancer versus sporadic breast cancer speak to different disease entities that are not being adequately reflected under the pathologist’s microscope. Our ability to demonstrate these differences in the protein profiling using SELDI-TOF MS lends credence to the fact that sporadic breast cancer and hereditary breast cancer are not one and the same.
Investigation into identifying these differentiating proteins may allow further elucidation of the roles of the BRCA-1 gene and lead to better-targeted therapy. For example, a primary target is the protein peak at 8.1-kDa that was overexpressed in the BRCA-1 cancer cohort, compared with both the BRCA-1 noncancer and sporadic breast cancer samples. A peak of 8.1 kDa was also identified in a separate serum profiling study as being upregulated in breast cancer subjects.8 Purification and sequencing determinations of the differentially expressed peaks detected in our study are ongoing.
Our ability to differentiate the BRCA-1 mutation carriers who developed breast cancer from those who did not holds much promise. Currently, once a woman is found to have a BRCA-1 mutation, she is offered three paths: (1) high-risk surveillance, (2) chemoprevention, or (3) prophylactic surgery. High-risk surveillance involves monthly breast self-examinations, clinical breast examinations every 6 months, and annual mammograms starting at age 25 years. For a BRCA1 carrier, by the age of 40 years, the chance of developing carcinoma of the breast is 20% (vs. 0.5% in the general population), and by age 50 years the risk increases to 50% (vs. 2% in the general population). Close surveillance has the advantages of being noninvasive and minimally intrusive but is not a preventative strategy; rather, it is an earlier detection modality.
Additionally, mammography, the "gold standard" for cancer detection, has the shortcoming of an inability to detect masses in younger patients with dense breasts. Ironically, for BRCA-1 carriers, this is the very age group for which cancer detection is most needed. Chemoprevention, aside from its plethora of side effects, offers a 50% reduction in cancer development. In high-risk women, this occurs by reducing estrogen receptor–positive sporadic breast cancer, but its role in BRCA-1 mutation carriers who traditionally generate estrogen receptor–negative breast cancer is unclear. Prophylactic surgery comprises bilateral oophorectomy (50% risk reduction) and bilateral mastectomy (90% risk reduction). In a 2001 study, Meijers-Heijboer prospectively randomized 76 BRCA-1 carriers to prophylactic bilateral mastectomies versus observation.13 There were no reported cases of breast cancer among the patients who underwent surgery, at a mean follow-up of 2.9 ± 1.4 years. In the surveillance cohort (N = 63), there were eight cases of breast cancer after a mean follow-up of 3.0 ± 1.5 years. Despite the advantages of prophylactic mastectomies and oophorectomy, the effect these procedures may have on young women’s body image, quality of life, and long-term health sequela of premature menopause may make prophylactic surgery a nonviable option.14 Obviously, the path chosen by a BRCA-1 carrier comes only with much contemplation and counseling regarding the risk and benefits of each treatment strategy, thus emphasizing the need for better predictive diagnostic tests in order to have a more complete clinical picture.
Our encouraging results would allow for the small subset of women with a BRCA-1 mutation who will not develop the disease in the near future to avoid unnecessary chemoprevention and/or prophylactic surgery. For those women who will develop breast cancer, protein profiling offers reassurance that the preventative actions were necessary. Since the time frame from baseline serum sample to cancer development was relatively short (within 3 years), one could envision following a woman’s protein profile annually and, when the profile begins to change toward the BRCA-1 cancer profile, performing the prophylactic mastectomies. Performing prophylactic surgery when the woman is older may have less impact on body image and quality of life.
Although the results of our study are very provocative, several limitations must be addressed. First and foremost, the sample size is too small to draw any definitive conclusions. Second, regarding the sporadic breast cancer vs. BRCA-1 cancer, the samples were from two different institutions, and differences may have resulted in a geographic and specimen acquisition bias in the protein profile comparison. Although this bias was preliminarily addressed by comparing the normal health volunteers (from Virginia) and the BRCA-1 carriers (from Nebraska), clarification with use of samples from BRCA-1 carriers with and without cancer from Virginia for comparison with the sporadic cases is currently under way. Last, both of these potential biases (small sample size and different collection techniques) could be addressed by enrolling multiple institutions with a standard collection protocol.
The last limitation involves the BRCA-1 cohorts with respect to timing. Since most breast cancers take years of growth to become clinically apparent, one could postulate that the protein profile of the BRCA-1 cancer group reflected occult disease rather than prediction of pending cancer development. Because they all developed cancer within 3 years of the baseline specimen, perhaps the cancer was there all along and escaped our detection by clinical examination and mammography. It is not uncommon for a woman with a BRCA-1 mutation to have a negative clinical breast examination and imaging studies, including a breast MRI, and then have breast cancer diagnosed within a few short months. To address this concern, a longer window of time—say, 10 years—would be needed between baseline specimen and cancer discovery. Even if SELDI-TOF is only detecting occult cancer and not predicting for development of cancer, this would still be a powerful tool. Earlier detection of any cancer is known to increase survival. Our prophylactic surgery would thus become "therapeutic" in cases found to have occult disease but would still be prophylactic for the contralateral side.
Conversely, the BRCA-1 carriers who did not develop cancer were followed for 7 years. Ideally, a much longer follow-up would be necessary to truly identify the carriers who will never develop cancer. Unfortunately, the incidence of breast cancer continues to increase with age, and thus a "long enough" follow-up may never be achieved. Perhaps it would be better to say that for those women with a BRCA-1 carrier profile, we predict they will not develop breast cancer within the next few years (in our study that would be up to 7 years). As our follow-up increases and the woman remains cancer-free, the prediction of disease-free interval would lengthen.
In summary, this exploratory small pilot study demonstrated that the serum protein profiles generated by SELDI-TOF MS differentiated BRCA-1 mutation carriers who developed breast cancer from those carriers who did not with reasonable sensitivity and specificity. In addition, our results showed that the protein profiles of sporadic breast cancer differ significantly from hereditary breast cancer attributed to BRCA-1 mutations and gives credence to the molecular heterogeneity of breast cancer. Perhaps a more feasible approach toward developing a serum test for early detection of breast cancer should involve the analysis of more homogeneous cohorts such as BRCA-1 mutations and then progression toward the common thread that represents "breast cancer" as a whole. Furthermore, determining the identities of the differentially expressed protein peaks could provide new biomarkers for disease or a potential target for therapy.
ACKNOWLEDGMENTS
This study was supported by a grant from the National Cancer Institute Early Detection Research Network (CA85067).
FOOTNOTES
Received March 19, 2004; accepted July 5, 2004.
Address correspondence and reprint requests to: Christine Laronga, MD, Department of Surgery, Eastern Virginia Medical School, 825 Fairfax Avenue, Suite 610, Norfolk, Virginia 23507; Fax: 757-446-8951; e-mail: larongc{at}evms.edu.
REFERENCES
This article has been cited by other articles:
 |
|
 |
|
  L. Harris, H. Fritsche, R. Mennel, L. Norton, P. Ravdin, S. Taube, M. R. Somerfield, D. F. Hayes, and R. C. Bast Jr American Society of Clinical Oncology 2007 Update of Recommendations for the Use of Tumor Markers in Breast Cancer J. Clin. Oncol., November 20, 2007; 25(33): 5287 - 5312. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. Bertucci, D. Birnbaum, and A. Goncalves Proteomics of Breast Cancer: Principles and Potential Clinical Applications Mol. Cell. Proteomics, October 1, 2006; 5(10): 1772 - 1786. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. S. Azad, N. Rasool, C. M. Annunziata, L. Minasian, G. Whiteley, and E. C. Kohn Proteomics in Clinical Trials and Practice: Present Uses and Future Promise Mol. Cell. Proteomics, October 1, 2006; 5(10): 1819 - 1829. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. F. Lopez, A. Mikulskis, S. Kuzdzal, D. A. Bennett, J. Kelly, E. Golenko, J. DiCesare, E. Denoyer, W. F. Patton, R. Ediger, et al. High-Resolution Serum Proteomic Profiling of Alzheimer Disease Samples Reveals Disease-Specific, Carrier-Protein-Bound Mass Signatures Clin. Chem., October 1, 2005; 51(10): 1946 - 1954. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. Meric-Bernstam Serum Proteomics for BRCA1-associated Breast Cancer Ann. Surg. Oncol., October 1, 2004; 11(10): 883 - 884. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
the intensity value. The question forming the first splitting rule is the following: Is the intensity levels of the peak at 5909 Da lower or equal to 1.871? Samples that follow the rule go to the left "yes" terminal node, and samples that do not follow the rule go to a "no" daughter node to the right. The numbers of BRCA-1 (Ca) and (No Ca) samples in each node are shown.