Approximately 2% to 5% of advanced cancers that present as metastatic cases are considered tumors of unknown origin when the primary tumor site is not determined, also known as carcinoma of unknown primary (CUP).1, 2 The rationale for identifying the primary tumor for such metastatic cases traditionally has been for the purpose of providing patients with cancer-specific treatment recommendations.3 This was especially relevant in the era before targeted therapy, when histotype and site of origin were the key parameters to define the treatment regimen. However, as targeted therapies increase in number and breadth, the concept of treatment on the basis of site of origin is likely to be replaced progressively by treatment on the basis of molecular signature. In this editorial, we suggest that, although the CUP test is just gaining popularity and use, it is also about to become obsolete. We also argue that next-generation sequencing (NGS) and therapeutic choice based on molecular signature will eliminate the need for site-of-origin–type tests.

Traditionally, site-of-origin information has been an important part of clinical management. Although some CUPs fall into a “favorable subset” characterized by similarity to good-prognosis tumor types, such as germ cell tumors or lymphoma, the remaining 80% of patients with CUP fall into a poor prognosis group characterized by a median survival of approximately 7 to 12 months from diagnosis.4, 5 CUPs are predominantly adenocarcinomas (90%), and the remaining tumors are divided evenly between squamous carcinomas and neuroendocrine carcinomas.2 Among adenocarcinomas, pancreatic and lung cancers are the most common primary sites (approximately 25% each), followed by colon, stomach, breast, ovarian, prostate, and solid organ tumors (kidney, thyroid, and liver)2; the similar appearance among adenocarcinoma metastatic biopsies is confounding to the diagnosis of these tumors. Nonetheless, CUP tests (whether based on expression levels or immunohistochemistry [IHC]) had clinical utility because they helped guide therapy.

In a recent edition of Cancer (Cancer Cytopathology), Stancel et al6 evaluated the feasibility of applying the Pathwork tissue-of-origin gene expression test (Pathwork Diagnostics, Redwood City, Calif) on cellblocks prepared from cytologic body fluid specimens. In this follow-up commentary, we describe the current methodologies for identifying the primary tumor site for CUPs along with the advantages and disadvantages of each method. We illustrate that the lack of a criterion (gold) standard to compare the accuracy of each methodology represents a major challenge, because, even upon autopsy, a primary tumor may not be identified in 70% to 80% of patients with CUPs.1, 5 To complicate matters further, the molecular assays described sometimes use IHC results, the previous criterion standard, for assessing the accuracy of their tumor site classifications. This is a poor standard, because IHC predicts the primary tumor site in only approximately 30% to 40% of cases.3, 5 Thus, the evaluation of this technology is inherently less than perfect.

Tissue Sample-Based Methods for Identifying the Primary Tumor Site

  1. Top of page
  2. Tissue Sample-Based Methods for Identifying the Primary Tumor Site
  3. Classification of Tumors Based on Actionable Mutations

In addition to the pathology assessment performed on the CUP sample, patients with CUPs also often undergo full-body imaging and other testing (such as a mammogram for a woman and a serum prostate-specific antigen [PSA] test for a man) to assess potential primary tumor sites. Beyond those tests, there are 2 main approaches available for determining the primary tumor site using the resected sample, cytology sample, or biopsy: IHC and molecular-based assays. Molecular-based assays include the use of technologies like quantitative real-time polymerase chain reaction (qRT-PCR) and oligonucleotide microarrays.

Immunohistochemistry panels

IHC evaluations complement the pathologist's diagnostic review of the hematoxylin and eosin-stained slide by testing a series of markers in a systematic approach so as to determine the likely primary site of CUPs.3 According to Dabbs7 and National Comprehensive Cancer Network guidelines,3 evaluation of CUPs by IHC typically begins with an evaluation of cytokeratin (CK) expression (CK7 and CK20), which is then followed by some organ-specific IHC panels. A partial listing of common markers used in these IHC panels for providing a primary tumor site diagnosis for CUPs is provided in Table 1; note that these markers are used by the pathologist as a resource to assist in diagnosis but are usually not uniformly specific or sensitive.3

Table 1. Partial List of Immunohistochemical Markers Used for the Potential Diagnosis of Major Tumor Site of Origin in Carcinoma of Unknown Primary Origin
Major Tumor TypeaKey Screening Antibodies for Positive IHC Staining to Aid in Diagnosis
  • Abbreviations: CA 19-9, carbohydrate antigen 19-9; CAM5.2, anticytokeratin antibody 5.2; CD10, cluster of differentiation 10 (neprilysin, membrane metalloendopeptidase, neutral endopeptidase, and common acute lymphoblastic leukemia antigen); CDX2, caudal type homeobox 2; CEA, carcinoembryonic antigen; CK, cytokeratin; D2-40, anti-M2A monoclonal antibody; EMA, epithelial membrane antigen; ER, estrogen receptor; GCDFP-15, gross cystic disease fluid protein 15; HepPar-1, hepatocyte paraffin 1; IHC, immunohistochemistry; LCA, leukocyte common antigen; MUC5-AC, mucin 5AC, oligomeric mucus/gel-forming; PAP, prostatic acid phosphatase; PLAP, placenta-like alkaline phosphatase; PR, progesterone receptor; PSA, prostate-specific antigen; RCC, renal cell carcinoma; S100, S-100 protein; TTF-1, thyroid transcription factor 1; WT-1, Wilms tumor 1.

  • a

    CK7/CK20 status is used first to suggest potential included or excluded tumor types.

AdrenocorticalInhibin, melan-A, calretinin
BreastER, PR, GCDFP-15, mammaglobin
EndometriumER, PR
Germ cell (nonseminoma)EMA, PLAP
Germ cell (seminoma)CAM5.2, PLAP
MelanomaS100, melan-A
MesotheliomaCalretinin, mesothelin, D2-40
OvaryER, PR, WT-1
PancreasCEA, CA 19-9, MUC5-AC
ProstatePSA, PAP
RenalRCC marker, vimentin, CD10
ThyroidTTF-1, thyroglobulin
UrothelialUroplakin III, thrombomodulin, p63, CK5/6

Each of these IHC assays in the panels is performed on a serial 5-μm-thick section from a formalin-fixed, paraffin-embedded (FFPE) tumor block, which is problematic when the tumor resection, cytology sample, or biopsy sample is limited. A study by Dennis et al8 was able to classify adenocarcinomas in approximately 88% of cases using a panel of 10 IHC stains (PSA, thyroid transcription factor 1 [TTF1], gross cystic disease fluid protein 15 [GCDFP-15], caudal type homeobox 2 [CDX2], CK20, CK7, estrogen receptor [ER], mesothelin, cancer antigen 125 [CA125], lysozyme). However, as noted above, only 30% to 40% of patients with metastatic disease who present with a CUP have a primary tumor site identified through IHC testing3, 5; molecular-based assays were developed to fill this gap in providing an alternative and a more robust approach for identifying the primary tumor origin.

Molecular assays

Advancements in microarray technology have allowed for the development of gene expression signatures of known tumor types for predicting the primary tumor site. Validation of CUP assays is inherently complex; because, by defining the tumor as a CUP, the primary site may remain unknown even after thorough investigation (and, in many cases, even after autopsy).1 Currently, there are 5 companies with commercially available tests for predicting the primary tumor site for FFPE CUP samples, as described in Table 2 including Agendia (Amsterdam, the Netherlands), bioTheranostics (San Diego, Calif), Pathwork Diagnostics (Redwood City, Calif), Rosetta (Philadelphia, Pa), and Veridex (La Jolla, Calif). An in-depth history of each test, other research publications related to test validation and development, and the classification methods used have been described previously by Monzon and Koen.1 The claimed accuracy of these tests in predicting the primary origin site of CUPs ranges from 78% to 88.5% compared with IHC and/or autopsy results. It is also important to note that some of the assays have decreased accuracy for poorly differentiated tumors or for specific tumor types, such as lung and pancreatic cancers, which is problematic, because those are the top 2 most prevalent tumor types identified from CUPs.

Table 2. List of Commercial Molecular Assays for Determining Tumor Origin by Gene Expression Profiling for FFPE Samples
  Assay TypeApproval StatusGenes in AssayClaimed Tissue Classification: No. of Major Types (No. of Histologic Subtypes)Claimed Accuracy, %aSelected Reference(s)
ManufacturerTest NamePlatformRNA TypeFDACE    
  • Abbreviations: CE, European Conformity; CUP, carcinoma of unknown primary origin; FDA, US Food and Drug Administration; mRNA, messenger RNA; miRNA, microRNA; qRT-PCR, quantitative reverse transcriptase-polymerase chain reaction.

  • a

    Information obtained from literature or company website.

  • b

    Available at:

  • c
  • d

    Available at:

  • e
  • f

    Available at:

  • All web sited accessed on June 1, 2012.

Agendia, Amsterdam, the NetherlandsbCupPrintOligonucleotide microarraymRNANoYes4959 (48)83Ma 2006,9 Horlings 200810
bioTheranostics, Inc., San Diego, CalifcCancerTYPE IDqRT-PCRmRNANoNo9230 (54)82Ma 20069
Pathwork Diagnostics, Redwood City, CalifdPathwork Tissue of Origin TestOligonucleotide microarraymRNAYesNo>15001588.5Pillai 201111
Rosetta Genomics Laboratories, Philadelphia, PaeProOncTumorSourceDx/ miRview metsqRT-PCRmiRNANoNo482585Rosenfeld 2008,12 Rosenwald 201013
Veridex, La Jolla, CaliffCUP AssayqRT-PCRmRNANo (not clinically available)No (not clinically available)10678Varadhachary 2008,14 Talantov 200615

Two companies use an oligonucleotide microarray approach: Agendia and Pathwork Diagnostics. The CupPrint test from Agendia was developed using published data sets, including that published by Ma et al,9 which is the same data set but with more genes than are included in the bioTheranostics assay. The CupPrint oligonucleotide microarray-based assay for FFPE samples uses a 495-gene set10 to classify tumor samples into 1 of 9 major tumor types, and it has been the only test to date to receive a European Conformity marking. Also using an oligonucleotide microarray approach initially developed for fresh frozen samples, the Tissue of Origin Test by Pathwork Diagnostics is the only 1 of these tests that has been approved by the US Food and Drug Administration. The analytic performance and reproducibility of this test across 4 different laboratories using a 1668-gene classifier on 60 samples had an overall concordance rate of 89.1% between laboratories and overall 86.7% agreement with the known tissue of origin.16 This assay was validated for use on FFPE samples using a 2000-gene classification model with sample requirements >60% tumor content, <20% necrosis, and a minimum of 30 ng total RNA. This test demonstrated a reported 89.3% concordance between 3 laboratories and overall agreement with reference diagnoses of 88.5%.11

Three companies have taken a qRT-PCR approach, with 2 using an mRNA classifier (bioTheranostics and Veridex) and 1 using a microRNA (miRNA) classifier (Rosetta). bioTheranostics' CancerTYPE ID test for FFPE CUP samples is based on gene expression profiles generated by Ma et al9 distilled down to a 92-gene assay. The Veridex assay is a qRT-PCR assay composed of 10 genes [HUMPB], TTF1, desmoglein 3 [DSG3], prostate stem cell antigen [PSCA], coagulation factor V [F5], cadherin 11 [CDH11], mammaglobin B [MGB], prostate epithelium-specific Ets transcription factor [PDEF], PSA, and Wilms tumor 1 [WT1]) used for FFPE samples to discriminate between a diagnosis of 6 adenocarcinoma tumor types (breast, colon, ovarian, lung, pancreatic, and prostate cancer).14, 15 Although most of the commercially available molecular assays are used to evaluate mRNA from FFPE samples, a decrease in mRNA stability in older FFPE samples has been noted,10 although some of the studies have reported that nearly 80% of FFPE samples were evaluable for mRNA.11 Therefore, miRNA-based classifiers may represent an alternative approach for the analysis of older FFPE samples with more robust retention in older samples12 than mRNA. An example of this is Rosetta's miRview mets assay, which is based on an miRNA classifier using 48 genes; Rosetta has also launched a second-generation test called miRview mets2.

There have been some limited studies indicating that defining the primary tumor site origin may have an impact on outcomes, such as the response of patients with colon cancer profiles to colon cancer-specific therapies.14 However, to our knowledge, there have not been any definitive, large-scale studies demonstrating an impact on patient outcome by such diagnosis.4 In addition, even with the use of molecular-based CUP assays, up to 40% of CUPs still may be left without a primary tumor site designation.1 Combining together IHC panels with a molecular-based assay (although not commercially available) has been described for use in refining the diagnosis of CUPs17 and may be 1 approach to integrating molecular CUP assays with current practices. However, further studies with larger sample numbers of most of these molecular assays would be required for a clearer picture of assay reproducibility, specificity, and sensitivity compared with autopsy results or IHC diagnostic workups. The true measure of the value of these tests will be the assessment of their clinical impact (response to therapy) in a randomized clinical trial. The value of such large studies to define tumor origin site may now be dubious in the current climate, in which it appears to be more important to define actionable mutations, regardless of tumor site, than to define site of origin.

Classification of Tumors Based on Actionable Mutations

  1. Top of page
  2. Tissue Sample-Based Methods for Identifying the Primary Tumor Site
  3. Classification of Tumors Based on Actionable Mutations

Ideally, the identification of the primary tumor site would lead to further testing to determine patient qualification for targeted therapeutics based on mutation or other biomarker status. However, the most rapid approach to obtain this information on molecular pathway alterations for CUP samples would not necessarily be using an IHC panel and/or molecular assays for determining the primary tumor site. Instead, an analysis of relevant, actionable mutations is an approach for classifying CUP tumors that can either replace or complement determining the site of origin. Similar to the molecular assays described above for tumor type identification, multiplexed, targeted genomic analyses like sequencing allow for the extraction of maximal information from small amounts of FFPE tumor samples even when mutations are rare in the heterogeneous mix of tissue present in a biopsy. In fact, the alterations identified by sequencing may or may not point to the site of origin but would allow for targeted therapeutic treatment regardless of origin site, which ultimately may improve patient outcomes more than the tumor site identification alone.

Cancer gene mutations may be present across multiple tumor types (even if at low frequencies) that may be indicators of benefit from targeted therapeutics. A partial list of such actionable mutations is provided in Table 3, which was adapted from MacConaill et al18 and Lipson et al.19 For example, MacConaill et al18 identified v-raf murine sarcoma viral oncogene homolog B1 (BRAF) codon 600 valine-to-glutamic acid (V600E) mutations in 4 different tumor types (colon cancer, ovarian cancer, thyroid cancer, and endometrial cancer), which may be responsive to selective BRAF inhibitors, as well as 7 different tumor types with either phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha (PIK3CA) or phosphatase and tensin homolog (PTEN) mutations (breast cancer, colon cancer, endometrial cancer, esophageal cancer, gastric cancer, prostate cancer, and pediatric astrocytoma), which may be responsive to phosphoinositide 3-kinase (PI3K) inhibitors. The identification of v-erb-b2 erythroblastic leukemia viral oncogene homolog 2, neuroblastoma/glioblastoma-derived oncogene homolog (avian) (ERBB2) mutations in gastric cancer18 has led to the use of ERBB2 inhibitors like trastuzumab, which is currently approved for patients with human epidermal growth factor receptor 2 (HER2)-positive breast cancer as well as certain HER2-expressing, advanced gastric or gastroesophageal junction cancers.20 There is a growing list of targeted therapeutics for which NGS-based testing can be used to assess the likelihood of drug responsiveness or drug resistance, ranging from biomarkers that are useful for a variety of signaling pathways, such as tyrosine kinase signaling (eg, EGFR inhibitors, ERBB2 inhibitors, and pan-tyrosine kinase inhibitors), mitogen-activated protein kinase (MAPK) signaling (eg, mitogen-activated protein kinase kinase/extracellular signal-regulated kinase [MEK/ERK] inhibitors), mechanistic target of rapamycin (serine/threonine kinase) (mTOR) signaling, and others listed in Table 3. It is expected that further cross-tumor–type, actionable mutations may be identified through worldwide collaborative efforts across cancer types, such as the Cancer Genome Atlas project and other genome projects.

Table 3. Partial List of Potentially “Actionable” Gene Mutations Regardless of Tumor Sitea
Drug Class or DrugRelevant Gene(s)
  • Abbreviations: ALK, Anaplastic lymphoma receptor tyrosine kinase; ATM, ataxia telangiectasia mutated; BRAF, v-raf murine sarcoma viral oncogene homolog B1; BRCA1/2, breast cancer genes 1 and 2, early onset; CCNE1, cyclin E1; CDK, cyclin-dependent kinase; CDK4, cyclin-dependent kinase 4; CDK8, cyclin-dependent kinase 8; CDKN2A, cyclin-dependent kinase inhibitor 2A; EGFR, epidermal growth factor receptor; ERBB2, v-erb-b2 erythroblastic leukemia viral oncogene homolog 2, neuroblastoma/glioblastoma-derived oncogene homolog (avian); FBXW7, F-box and WD repeat domain containing 7, E3 ubiquitin protein ligase; GNAS, guanine nucleotide binding protein, alpha stimulating activity polypeptide 1; JAK2, Janus kinase 2; KIT, v-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog; KRAS, v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog; MDM2, p53 E3 ubiquitin protein ligase homolog (mouse); MEK/ERK, mitogen-activated protein kinase kinase/extracellular signal-regulated kinase; mTOR, mechanistic target of rapamycin (serine/threonine kinase); PARP, poly-(ADP-ribose) polymerase; PDGFRA, platelet-derived growth factor receptor alpha; PI3K, phosphatidylinositol-4,5-bisphosphate 3-kinase; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha; PTEN, phosphatase and tensin homolog; RET, ret proto-oncogene; TSC1, tuberous sclerosis 1.

  • a

    Adapted from MacConaill LE, et al.18; and Lipson D, et al.19

ALK inhibitors (eg, crizotinib)ALK (fusion)
BRAF inhibitors (eg, vemurafenib)BRAF
CDK inhibitors (eg, flavopiridol)CCNE1, CDK4, CDKN2A, CDK8
EGFR inhibitors (eg, cetuximab, panitumumab)BRAF, KRAS
EGFR inhibitors (eg, erlotinib, gefitinib)EGFR, PTEN
ERBB2 inhibitors (eg, trastuzumab, lapatinib)ERBB2
JAK2 inhibitorsJAK2
MDM2 inhibitors (eg, nutlins)MDM2
MEK/ERK inhibitorsGNAS
mTOR inhibitorsPIK3CA, TSC1
Pan-tyrosine kinase inhibitors (eg, imatinib, nilotinib)KIT, PDGFRA
PARP inhibitorsATM, BRCA1/2
PI3K inhibitorsPIK3CA, PTEN
RET inhibitors (eg, sorafenib, sunitinib)RET

From a platform perspective, there are currently at least 3 bench-top, high-throughput sequencing instruments commercially available: Roche (454 GS Junior; Roche, Diagnostics, Indianapolis, Ind), Illumina (MiSeq; Illumina, San Diego, Calif), and Life Technologies (Ion Torrent PGM; Life Technology, Carlsbad, Calif). Cancer gene panel-focused options exist for their appropriate platforms, such as the OncoCarta panel offered by Sequenom on the MassARRAY system (238 mutations across 19 genes), Life Technology's Ion AmpliSeq Cancer Panel (739 mutations across 46 genes), and the RainDance ONCOSeq Research Screening Panel (142 genes) and Cancer Hotspot Panel (54 genes). Many cancer centers are working to develop and validate NGS services for tumor sample analysis, with testing currently offered from institutions like Baylor College of Medicine (,681; accessed June 1, 2012) and Genomic and Pathology Services at Washington University (St. Louis, Mo; (; accessed June 1, 2012). The companies that offer these types of NGS cancer panel-based services include Foundation Medicine (Cambridge, Mass; www.foundationmedicine.com19; accessed June 1, 2012), Asuragen (Austin, Tex;; accessed June 1, 2012), AltheaDx (San Diego, Calif; accessed June 1, 2012), and Personal Genome Diagnostics, Inc. (Baltimore, Md;; accessed June 1, 2012). In fact, this list is growing so fast that it is almost certain to be obsolete within a few months after this publication. And, as the cost per base goes down at a rate exceeding that of Moore's law, and as the number of genes that are “clinically actionable” increases, it seems highly likely that NGS will have a dramatically negative effect on the clinical utility of CUP testing.

In conclusion, with advances in NGS and publications demonstrating the existence of cancer gene mutations across tumor types linked to response or lack of response to targeted therapeutics, organ system regimens are becoming outdated. Therefore, there may be great clinical benefit for a range of cancers, defined by the NGS approach, either in combination with or even possibly in lieu of methods for identifying the primary tumor site. Finally, even if no actionable mutations are identified from CUP samples through an NGS approach, traditional IHC panels and improved IHC panels have the capability to provide guidance for possible traditional tumor site-based therapeutics. Traditional pathology approaches are critical for sample classification. Cutting edge pathologists will maintain their critical position in analysis of patient samples by embracing these new approaches for optimizing patient outcomes beyond a focus on tumor type diagnosis alone.


  1. Top of page
  2. Tissue Sample-Based Methods for Identifying the Primary Tumor Site
  3. Classification of Tumors Based on Actionable Mutations

No specific funding was disclosed.


The authors made no disclosures.


  1. Top of page
  2. Tissue Sample-Based Methods for Identifying the Primary Tumor Site
  3. Classification of Tumors Based on Actionable Mutations
  • 1
    Monzon FA, Koen TJ. Diagnosis of metastatic neoplasms: molecular approaches for identification of tissue of origin. Arch Pathol Lab Med. 2010; 134: 216-224.
  • 2
    Oien KA, Evans TR. Raising the profile of cancer of unknown primary. J Clin Oncol. 2008; 26: 4373-4375.
  • 3
    National Comprehensive Cancer Network (NCCN). NCCN Clinical Practice Guidelines in Oncology: Occult Primary, Version 1.2012. Available at: [Access date.]
  • 4
    Pentheroudakis G, Greco FA, Pavlidis N. Molecular assignment of tissue of origin in cancer of unknown primary may not predict response to therapy or outcome: a systematic literature review. Cancer Treat Rev. 2009; 35: 221-227.
  • 5
    Greco FA, Oien K, Erlander M, et al. Cancer of unknown primary: progress in the search for improved and rapid diagnosis leading toward superior patient outcomes. Ann Oncol. 2012; 23: 298-304.
  • 6
    Stancel GA, Coffe D, Alverez K, et al. Identification of tissue of origin in body fluid specimens using a gene expression microarray assay. Cancer (Cancer Cytopathol). 2012; 120: 62-70.
  • 7
    Dabbs DJ. Diagnostic Immunohistochemistry. New York: Churchill Livingstone; 2002.
  • 8
    Dennis JL, Hvidsten TR, Wit EC, et al. Markers of adenocarcinoma characteristic of the site of origin: development of a diagnostic algorithm. Clin Cancer Res. 2005; 11: 3766-3772.
  • 9
    Ma XJ, Patel R, Wang X, et al. Molecular classification of human cancers using a 92-gene real-time quantitative polymerase chain reaction assay. Arch Pathol Lab Med. 2006; 130: 465-473.
  • 10
    Horlings HM, van Laark RK, Kerst JM, et al. Gene expression profiling to identify the histogenetic origin of metastatic adenocarcinomas of unknown primary. J Clin Oncol. 2008; 26: 4435-4441.
  • 11
    Pillai R, Deeter R, Rigl CT, et al. Validation and reproducibility of a microarray-based gene expression test for tumor identification in formalin-fixed, paraffin-embedded specimens. J Mol Diagn. 2011; 13: 48-56.
  • 12
    Rosenfeld N, Aharonov R, Meiri E, et al. MicroRNAs accurately identify cancer tissue origin. Nat Biotechnol. 2008; 26: 462-469.
  • 13
    Rosenwald S, Gilad S, Benjamin S, et al. Validation of a microRNA-based qRT-PCR test for accurate identification of tumor tissue origin. Mod Pathol. 2010; 23: 814-823.
  • 14
    Varadhachary GR, Talantov D, Raber MN, et al. Molecular profiling of carcinoma of unknown primary and correlation with clinical evaluation. J Clin Oncol. 2008; 26: 4442-4448.
  • 15
    Talantov D, Baden J, Jatkoe T, et al. A quantitative reverse transcriptase-polymerase chain reaction assay to identify metastatic carcinoma tissue of origin. J Mol Diagn. 2006; 8: 320-329.
  • 16
    Dumur CI, Lyons-Weiler M, Sciulli C, et al. Interlaboratory performance of a microarray-based gene expression test to determine tissue of origin in poorly differentiated and undifferentiated cancers. J Mol Diagn. 2008; 10: 67-77.
  • 17
    Centeno BA, Bloom G, Chen DT, et al. Hybrid model integrating immunohistochemistry and expression profiling for the classification of carcinomas of unknown primary site. J Mol Diagn. 2010; 12: 476-486.
  • 18
    MacConaill LE, Campbell CD, Kehoe SM, et al. Profiling critical cancer gene mutations in clinical tumor samples [serial online]. PLoS One. 2009; 4: e7887.
  • 19
    Lipson D, Capelletti M, Yelensky R, et al. Identification of new ALK and RET gene fusions from colorectal and lung cancer biopsies. Nat Med. 2012; 18: 382-384.
  • 20
    Stern HM. Improving treatment of HER2-positive cancers: opportunities and challenges [serial online]. Sci Transl Med. 2012; 4: 127rv2.