• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. Jun 15, 2004; 101(24): 9067–9072.
PMCID: PMC428474
Medical Sciences

High-resolution characterization of the pancreatic adenocarcinoma genome

Abstract

The pancreatic adenocarcinoma genome harbors multiple amplifications and deletions, pointing to the existence of numerous oncogenes and tumor suppressor genes driving the genesis and progression of this lethal cancer. Here, array comparative genomic hybridization on a cDNA microarray platform and informatics tools have been used to define the copy number alterations in a panel of 24 pancreatic adenocarcinoma cell lines and 13 primary tumor specimens. This high-resolution genomic analysis has identified all known regional gains and losses as well as many previously uncharacterized highly recurrent copy number alterations. A systematic prioritization scheme has selected 64 focal minimal common regions (MCRs) of recurrent copy number change. These MCRs possess a median size of 2.7 megabases (Mb), with 21 (33%) MCRs spanning 1 Mb or less (median of 0.33 Mb) and possessing an average of 15 annotated genes. Furthermore, complementary expression profile analysis of a significant fraction of the genes residing within these 64 prioritized MCRs has enabled the identification of a subset of candidates with statistically significant association between gene dosage and mRNA expression. Thus, the integration of DNA and RNA profiles provides a highly productive entry point for the discovery of genes involved in the pathogenesis of pancreatic adenocarcinoma.

Keywords: array comparative genomic hybridization, expression profile

Pancreatic adenocarcinoma is among the most lethal of human cancers, typically presenting as advanced inoperable disease with a rapidly progressive clinical course characterized by intense resistance to all therapeutic modalities. Significant effort has been directed toward charting the molecular genetic events in this cancer with the goals of improving early detection and providing new therapeutic targets. The current compendium of validated genetic mutations has provided a multistep model for the initiation and progression of pancreatic adenocarcinoma that is typified by the near-universal and early occurrence of activating mutations in KRAS and frequent later-stage inactivation of p16INK4A, p53, and/or SMAD4 (1).

These stereotypic genetic lesions take place against the backdrop of a high level of genomic instability that is evident in the earliest stages of the disease (2-4). Indeed, a hallmark genomic feature of this cancer is the presence of numerous complex chromosome structural aberrations, including nonreciprocal translocations, amplifications, and deletions. To date, karyotype analyses (5-10), chromosomal comparative genomic hybridization (CGH) (11-17), and loss of heterozygosity mapping (18-20) have identified recurrent regions of copy number change or allelic loss. In particular, frequent gains have been mapped to 3q, 5p, 7p, 8q, 11q, 12p, 17q, and 20q and losses to 3p, 4q, 6q, 8p, 9p, 10q, 12q, 13q, 17p, 18q, 21q, and 22q. In some instances, validated oncogenes and tumor suppressor genes residing within these loci have been identified, including MYC (8q24), p16INK4A (9p21), p53 (17p13), SMAD4 (18q21), and AKT2 (19q13). However, for the majority of amplified and deleted loci, the presumed cancer-relevant targets remain to be discovered.

In this study, the development of optimized protocols and bioinformatic tools has enabled the use of a cDNA-based platform for the high-resolution characterization of copy number alterations (CNAs) in the pancreatic adenocarcinoma genome, leading to the rediscovery of known cytogenetic alterations and the identification of many focal and previously undescribed CNAs. Furthermore, the integration of these copy number data with expression profiles and other cancer database information provides for a highly efficient entry point for cancer gene discovery.

Materials and Methods

Primary Tumors and Cell Lines. All cell lines were acquired from the American Type Culture Collection (ATCC) or the German Collection of Microorganisms and Cell Cultures (DSMZ). All fresh-frozen specimens of primary pancreatic ductal adenocarcinoma were obtained from the Memorial Sloan-Kettering Cancer Center tumor bank, and histology was confirmed by hematoxylin/eosin before inclusion in this study (Tables 2 and 3, which are published as supporting information on the PNAS web site).

Array-CGH Profiling on cDNA Microarrays. Genomic DNA was fragmented and random-prime labeled according to published protocols (21) with modifications (for details, see http://genomic.dfci.harvard.edu/array_cgh.htm). Labeled DNAs were hybridized to human cDNA microarrays containing 14,160 cDNA clones (Agilent Technologies, Palo Alto, CA, Human 1 clone set), for which ≈9,420 unique map positions were defined (National Center for Biotechnology Information, Build 33). The median interval between mapped elements is 100.1 kb, 92.8% of intervals are <1 megabases (Mb), and 98.6% are <3 Mb.

Fluorescence ratios of scanned images of the arrays were calculated, and the raw array-CGH profiles were processed to identify statistically significant transitions in copy number using a segmentation algorithm, which uses permutation to determine the significance of change points in the raw data (22). Each segment is assigned a log2 ratio that is the median of the contained probes. The data are centered by the tallest mode in the distribution of the segmented values. After mode centering, we defined gains and losses as log2 ratios of ≥+0.13 or -0.13 (±4 standard deviations of the middle 50% quantile of data), and amplification and deletion as a ratio >0.52 or less than -0.58, respectively (i.e., 97% or 3% quantiles) (Fig. 4, which is published as supporting information on the PNAS web site). Automated Locus Definition. Loci are defined by an automated algorithm applied to the segmented data based on the following rules:

  1. Segments above the 97th or below the 3rd percentile are identified as altered.
  2. If two or more altered segments are adjacent in a single profile or separated by <500 kb, the entire region spanned by the segments is considered to be an altered span.
  3. Highly altered segments or spans that are shorter than 20 Mb are retained as “informative spans” for defining discrete locus boundaries. Longer regions are not discarded but are not included in defining locus boundaries.
  4. Informative spans are compared across samples to identify overlapping groups of positive- or negative-value segments; each group defines a locus.
  5. Minimal common regions (MCRs) are defined as contiguous spans having at least 75% of the peak recurrence as calculated by counting the occurrence of highly altered segments. If two MCRs are separated by a gap of only one probe position, they are joined. If there are more than three MCRs in a locus, the whole region is reported as a single complex MCR.

MCR Characterization. For each MCR, the peak segment value is identified. Recurrence of gain or loss is calculated across all samples based on the lower thresholds previously defined (≈±0.13). As an additional measure of recurrence independent of thresholds for segment value or length, median aberration (MA) is calculated for each probe position by taking the median of all segment values above zero for amplified regions and below zero for deleted regions. This pair of values is compared with the distribution of values obtained after permuting the probe labels independently in each sample profile. Where the magnitude of the MA exceeds 95% of the permuted averages, the region is marked as significantly gained or lost, and this is used in the voting system for prioritization. The number of known genes and genscan (Massachusetts Institute of Technology, Cambridge, MA) model predicted genes is counted based on the April 2003 human assembly at the University of California at Santa Cruz (http://genome.ucsc.edu).

Quantitative PCR (QPCR) Verification. PCR primers were designed to amplify products of 100-150 bp within target and control sequences (available upon request). Primers for control sequences in each cell line were designed within a region of euploid copy number as shown by array-CGH analysis. QPCR was performed by monitoring in real-time the increase in fluorescence of SYBR green dye (Qiagen, Chatsworth, CA) with an ABI 7700 (Applied Biosystems) sequence detection system (PerkinElmer). The relative gene copy number was calculated by the comparative Ct method (23).

Expression Profiling on Affymetrix GeneChip. Biotinylated target cRNA was generated from total sample RNA and hybridized to human oligonucleotide probe arrays (U133A, Affymetrix, Santa Clara, CA) according to standard protocols (24). Expression values for each gene were standardized by log2 ratio to a middle value for the sample set, defined as the midpoint between 25% and 75% quantiles, and mapped to genomic positions based on National Center for Biotechnology Information Build 33 of the human genome.

Integrated Copy Number and Expression Analysis. Array-CGH data are interpolated such that each expression value can be mapped to its corresponding copy number value. To maximize detection of focal CNAs, two separate interpolations are calculated: one selecting the higher bounding CGH probe and one choosing the lower. For each gene position, the samples are grouped based on whether or not array-CGH shows an altered copy number according to interpolated CGH value. The effect of gene dosage on expression is measured by calculating a gene weight defined as the difference of the means of the expression value in the altered and unaltered sample groups divided by the sum of the standard deviations of the expression values in altered and unaltered sample groups (25). The significance of the weight for each gene is estimated by permuting the sample labels 10,000 times and applying an α threshold of 0.05.

Results and Discussion

Comprehensive Catalog of CNAs in the Pancreatic Adenocarcinoma Genome. From a total of 75 primary pancreatic tumor specimens, we identified 13 samples that possessed >60% neoplastic cellularity (Table 2). Genomic DNAs from these primary tumor samples, along with DNAs derived from 24 established pancreatic cancer cell lines (Table 3), were subjected to genome-wide array-CGH profiling by using a cDNA-based array platform that offers a median resolution of 100 kB (see Materials and Methods). To facilitate identification of significant copy number events in these array-CGH profiles, this study has used a modified version of the circular binary segmentation methodology developed by Olshen and colleagues (22, 26). This algorithm applies nonparametric statistical testing to identify and distinguish discrete copy number transition points from chance noise events in the primary dataset. As shown in Fig. 1A, the segmented array-CGH profiles readily identified large regional changes that are typically of low amplitude, hereafter referred to as “gain” or “loss” (see Materials and Methods). Similarly, focal high-amplitude alterations representing “amplification” or “deletion” are evident in both primary tumor specimens and tumor cell lines (Fig. 1). Recurrence frequencies of the CNAs reported here match the frequencies described in the published literature (11-17) (Fig. 1B). There is also strong concordance between primary tumors and cell lines with respect to gains on 3q, 8q, and 20q and losses on 1p, 3p, 6q, 9p, 17p, and 18q (Fig. 5, which is published as supporting information on the PNAS web site). However, some differences were evident between primary tumor and cell line datasets and are likely attributable to the cellular heterogeneity within primary tumor samples and/or culture-induced genetic adaptation in the cell lines.

Fig. 1.
Genomic profiles from pancreatic adenocarcinoma samples. Array-CGH profiles with x axis coordinates representing cDNA probes ordered by genomic map positions. Segmented data are displayed in red, median filtered (three nearest neighbors) in blue, and ...

The identification of many CNAs, along with the high degree of structural complexity within each CNA, prompted the implementation of objective criteria to define and prioritize CNAs across the dataset. To that end, a locus-identification algorithm was developed that defines informative CNAs on the basis of size and achievement of a high-significance threshold for the amplitude of change (see Materials and Methods). Overlapping CNAs from multiple profiles are then merged in an automated fashion to define a discrete “locus” of regional copy number change, the bounds of which represent the combined physical extent of these overlapping CNAs (Fig. 1C). Each locus is characterized by a peak profile, the width and amplitude of which reflect the contour of the most prominent amplification or deletion for that locus. Furthermore, within each locus, one or more MCRs can be identified across multiple tumor samples (Fig. 1C), with each MCR potentially harboring a distinct cancer-relevant gene targeted for copy number alteration across the sample set.

The locus identification algorithm appears to be highly effective in delineating more discrete CNAs within previously described larger regions of gain or loss. For example, chromosome 6q has been reported as one of the most frequently deleted regions in pancreatic adenocarcinoma, but to our knowledge, no validated tumor suppressor gene has yet been assigned to this locus. Analysis of 6q loss in our dataset has identified four distinct MCRs that range in size from 2.4 to 12.8 Mb, raising the possibility that there may be multiple targets for 6q loss. Notably, two of these MCRs (Table 4, which is published as supporting information on the PNAS web site, locus nos. 74 and 75) coincide with previously identified regions of common allelic loss (27), an observation that provides a measure of validation for the analytical approach developed in this study. Selection of High-Priority Loci. The above locus-identification algorithm defined 287 discrete MCRs (from 256 independent autosomal loci) within this dataset and annotated each in terms of recurrence, amplitude of change, and representation in both cell lines and primary tumors. Based on our extensive experience with this platform across many tumor types (unpublished data), recurrence across multiple independent samples and high-amplitude signals are the two features most predictive of verification by independent assays. Hence, we prioritized these discrete MCRs based on four criteria that include (i) recurrence of high-threshold amplification or deletion (above the 97th percentile or below the 3rd percentile) in at least two specimens, (ii) presence of a high-threshold event in at least one primary tumor specimen, (iii) statistically significant median aberration (see Materials and Methods), and (iv) a peak amplitude of equal to or greater than absolute log2 value of 0.8 in either a cell line or primary tumor (beyond 0.5% quantiles).

Implementation of this prioritization scheme yielded 64 MCRs within 54 independent loci that satisfied at least three of the four criteria (Table 1). Notably, genes known to play important roles in the pathogenesis of pancreatic adenocarcinoma, the p16INK4A and TP53 tumor suppressors and the MYC, KRAS2, and AKT2 oncogenes, were present within these high-confidence loci (Table 1). Within the prioritized MCRs, there was an average recurrence rate for gain/loss of 38% across the entire dataset and the maxima or minima absolute log2 values for 34 of these 64 MCRs are >1.0, placing them significantly above the threshold defined for amplification or deletion (Fig. 4). It is noteworthy that in the majority of cases, the peak profile of a locus coincided with one of the MCRs (47 of 54 loci, Table 1), an observation that reinforces the notion that important targets are likely to reside within the genomic region defined by the peak of a CNA (28). The median size of these 64 prioritized MCRs is 2.7 Mb, with 21 MCRs (33%) spanning 1 Mb or less (Table 1). Residing within these 21 highly focal MCRs with a median size of 0.33 Mb, there are on average 15 annotated and 8 genscan predicted genes, rendering them highly attractive for target identification.

Table 1.
High-confidence recurrent CNAs in pancreatic adenocarcinoma

The confidence level ascribed to these prioritized loci has been further validated by real-time QPCR, which demonstrated 100% concordance with 16 selected MCRs defined by array-CGH (Table 1). For example, the MCR of an amplified locus at 7q21.11-7q32.2 was readily confirmed by QPCR (Fig. 2A). Furthermore, QPCR analyses also verified the structural details of complex CNAs reported by array-CGH. As shown in Fig. 2B, QPCR precisely mirrored each component of the complex 9p21 locus in HUP-T3, including homozygous deletion of p16INK4A, the known target for this CNA. Such detailed structural information may prove useful in dissecting the mechanisms responsible for the genesis of these cancer-associated chromosomal aberrations.

Fig. 2.
QPCR verifies complexity within CNAs. (A) Chromosome 7 CGH profiles (Left) showing amplification of a discrete region of 7q22 in both the AsPC-1 cell line and PA.T.14172 (locus no. 9, Table 1), with MCR defined by both samples (outlined by dashed lines). ...

Altogether, when high-priority MCRs in Table 1 are combined with an additional 80 moderate-priority MCRs (within 65 distinct loci) satisfying two of four criteria, our genomic characterization has produced a list of 144 MCRs within 119 independent loci that warrant further validation and detailed characterization (Table 4). Of note, the boundaries of the MCRs of each locus have been defined based on conservative parameters. In many cases, the focal and informative nature of these recurrent aberrations can best be appreciated by examination of the primary array-CGH profiles and consideration of such factors as the peak amplitude and morphology of the individual CNAs within an MCR. In addition, it is worth noting that the segmentation algorithm discards CNAs reported by single probes in exchange for an improved false-positive rate. Consequently, there may be highly focal alterations that are not captured by the analysis reported here. A case in point is the loss of the SMAD4 tumor suppressor, which is represented by a single probe on the cDNA microarray. Although the raw CGH profiles showed loss of copy number for this SMAD4 probe, the segmentation algorithm discarded this CNA. Data such as this should provide impetus for development of array platforms with increased density of coverage across the genome.

Integrated Analysis of Copy Number and Expression Information. Copy number aberrations and their associated impact on gene expression patterns represent common mechanisms of oncogene activation and tumor suppressor inactivation. Indeed, integration of copy number and transcription profile datasets revealed a consistent influence of gene dosage on mRNA expression globally across the genome (Fig. 6, which is published as supporting information on the PNAS web site) (25, 29). Conversely, as previously demonstrated (30), only a subset of genes within any given CNA show copy-number-driven expression changes, a feature that provides a first-pass means of distinguishing bystanders from potential cancer gene targets within the CNA. As a case in point, a locus of amplification on chromosome 17 in the cell line Hup-T3 (locus no. 21, Table 1) contains 455 genes of which 151 are present on the Affymetrix U133A array. Of these 151 genes, only 19 exhibited increased transcript levels >2-fold. Moreover, these 19 genes reside within the peak of this locus (Fig. 3A). Similar correlations can be established in regions of deletion. For example, the 9p21 deletion locus in the BxPC-3 cell line demonstrated that only 5 of 91 genes residing within the MCR show undetectable or decreased expression <2-fold (Fig. 3B). Examination of p16INK4A, the known target for deletion, across the entire sample set demonstrates that 11 of 24 cell lines show low or absent expression, of which five had homozygous deletion, whereas the remaining six were present at the DNA level (Fig. 3C). In the latter, epigenetic silencing is the presumed mechanism of p16INK4a inactivation.

Fig. 3.
Combined array-CGH and expression analysis facilitates identification of candidate genes. (A) Analysis of 17q23.2-25.3 locus (locus no. 21, Table 1) in cell line Hup T3. (Upper) Array-CGH profile of HUP-T3. (Lower) Expression profile of genes on Affymetrix ...

It is important to emphasize that a major issue in interpretation of expression information is the challenge of defining over- or underexpressed levels. In many cancer types, including pancreatic adenocarcinoma, the true cell of origin remains unknown, and thus a premalignant physiological frame of reference is not available. In the examples above, we have applied one model for interfacing copy number and expression profiles by midpoint centering of the expression data and calculating a weighted statistic for assignment of significance values to genes with correlated copy number and expression (24, 25) (see Materials and Methods). Using this approach, we next sought to prioritize the genes residing within the 64 high-confidence MCRs (Table 1) based on the correlation of their expression with gene dosage. Although only a subset of genes are represented, the Affymetrix U133A array permitted inclusion of 1,926 genes of a total of 4,742 genes residing within these MCRs for this analysis. By weighing each of these 1,926 genes based on the magnitude of its expression alteration and representation within CNAs across the dataset, the integrated copy number and expression analysis yielded a list of 603 genes that show a statistically significant association between gene copy number and mRNA expression (P < 0.05; Table 5, which is published as supporting information on the PNAS web site). Of these, 336 are located within regions of amplifications and 267 within regions of deletions. Importantly, among these 603 genes were known pancreatic cancer genes such as MYC (13), p16INK4A (31, 32), and DUSP6 (33) (Table 5), thus reinforcing the value of integrating both copy number and expression information.

Although incomplete representation of known and predicted genes on the Affymetrix U133A expression array precluded assessment of all possible target genes, the complementary analysis of array-CGH and expression profiles presented above serves to prioritize the list of available cancer gene candidates and provides a basis for focus on a subset of high-probability candidates. In addition, integrating genomic datasets across species may also prove effective in facilitating cancer gene identification. A particularly productive path for oncogene identification may be the analysis of common integration sites (CISs) present in retrovirally promoted leukemias and lymphomas (34). Consistent with the paradigm that proviral integration primarily serves to activate endogenous protooncogene (34), syntenic mapping of 232 CISs to the human genome (35) uncovered 19 CISs residing within MCRs of amplified loci in Table 1, whereas only 10 would be expected by chance alone (P < 0.006). On the contrary, MCRs within regions of loss or deletion contained only 16 CISs, whereas 14.4 would have been expected by chance alone. Thus, it is tempting to speculate that the abundance of CISs mapping to amplified loci represents genes with pathogenetic relevance in mouse models of tumor progression as well as in human pancreatic cancer, although we are mindful of possible cell-type-specific roles for these candidate genes. Ultimately, rigorous functional validation in cell culture-based assays and engineered mouse models will be necessary to definitively assign cancer relevance to the genes targeted by the recurrent amplifications and deletions described in this study.

Acknowledgments

We thank Drs. Ruben Carrasco, Juan Cadinanos, Giovanni Tonon, Eric Martin, Jan-Hermen Dannenberg, and Kornelia Polyak for critical comments on the manuscript and Tali Muller and Melissa Donovan for superb technical support. Array-CGH profiles were performed at the Arthur and Rochelle Belfer Cancer Genomic Center at the Dana-Farber Cancer Institute. This work is supported in part by an Extraordinary Opportunity Award (to L.C.) from the Lustgarten Foundation, by an American Cancer Society Professorship award (to R.A.D.), and by National Institutes of Health grants [RO1 CA99041 (to L.C.) and R01CA86379 and R01CA84628 (to R.A.D.)]. C.B. is supported by National Institutes of Health Training Grant T32 CA09382 and by the LeBow Fund for Myeloma Cure.

Notes

Abbreviations: CGH, comparative genomic hybridization; MCR, minimal common region; QPCR, quantitative PCR; CNA, copy number alteration; CIS, common integration site; Mb, megabase(s).

References

1. Bardeesy, N. & DePinho, R. A. (2002) Nat. Rev. Cancer 2, 897-909. [PubMed]
2. Gisselsson, D., Pettersson, L., Hoglund, M., Heidenblad, M., Gorunova, L., Wiegant, J., Mertens, F., Dal Cin, P., Mitelman, F. & Mandahl, N. (2000) Proc. Natl. Acad. Sci. USA 97, 5357-5362. [PMC free article] [PubMed]
3. Gisselsson, D., Jonson, T., Petersen, A., Strombeck, B., Dal Cin, P., Hoglund, M., Mitelman, F., Mertens, F. & Mandahl, N. (2001) Proc. Natl. Acad. Sci. USA 98, 12683-12688. [PMC free article] [PubMed]
4. van Heek, N. T., Meeker, A. K., Kern, S. E., Yeo, C. J., Lillemoe, K. D., Cameron, J. L., Offerhaus, G. J., Hicks, J. L., Wilentz, R. E., Goggins, M. G., et al. (2002) Am. J. Pathol. 161, 1541-1547. [PMC free article] [PubMed]
5. Johansson, B., Bardi, G., Heim, S., Mandahl, N., Mertens, F., Bak-Jensen, E., Andren-Sandberg, A. & Mitelman, F. (1992) Cancer 69, 1674-1681. [PubMed]
6. Bardi, G., Johansson, B., Pandis, N., Mandahl, N., Bak-Jensen, E., Andren-Sandberg, A., Mitelman, F. & Heim, S. (1993) Br. J. Cancer 67, 1106-1112. [PMC free article] [PubMed]
7. Griffin, C. A., Hruban, R. H., Long, P. P., Morsberger, L. A., Douna-Issa, F. & Yeo, C. J. (1994) Genes Chromosomes Cancer 9, 93-100. [PubMed]
8. Griffin, C. A., Hruban, R. H., Morsberger, L. A., Ellingham, T., Long, P. P., Jaffee, E. M., Hauda, K. M., Bohlander, S. K. & Yeo, C. J. (1995) Cancer Res. 55, 2394-2399. [PubMed]
9. Gorunova, L., Johansson, B., Dawiskiba, S., Andren-Sandberg, A., Jin, Y., Mandahl, N., Heim, S. & Mitelman, F. (1995) Genes Chromosomes Cancer 14, 259-266. [PubMed]
10. Gorunova, L., Hoglund, M., Andren-Sandberg, A., Dawiskiba, S., Jin, Y., Mitelman, F. & Johansson, B. (1998) Genes Chromosomes Cancer 23, 81-99. [PubMed]
11. Solinas-Toldo, S., Wallrapp, C., Muller-Pillasch, F., Bentz, M., Gress, T. & Lichter, P. (1996) Cancer Res. 56, 3803-3807. [PubMed]
12. Mahlamaki, E. H., Hoglund, M., Gorunova, L., Karhu, R., Dawiskiba, S., Andren-Sandberg, A., Kallioniemi, O. P. & Johansson, B. (1997) Genes Chromosomes Cancer 20, 383-391. [PubMed]
13. Mahlamaki, E. H., Barlund, M., Tanner, M., Gorunova, L., Hoglund, M., Karhu, R. & Kallioniemi, A. (2002) Genes Chromosomes Cancer 35, 353-358. [PubMed]
14. Fukushige, S., Waldman, F. M., Kimura, M., Abe, T., Furukawa, T., Sunamura, M., Kobari, M. & Horii, A. (1997) Genes Chromosomes Cancer 19, 161-169. [PubMed]
15. Curtis, L. J., Li, Y., Gerbault-Seureau, M., Kuick, R., Dutrillaux, A. M., Goubin, G., Fawcett, J., Cram, S., Dutrillaux, B., Hanash, S. & Muleris, M. (1998) Genomics 53, 42-55. [PubMed]
16. Ghadimi, B. M., Schrock, E., Walker, R. L., Wangsa, D., Jauho, A., Meltzer, P. S. & Ried, T. (1999) Am. J. Pathol. 154, 525-536. [PMC free article] [PubMed]
17. Armengol, G., Knuutila, S., Lluis, F., Capella, G., Miro, R. & Caballin, M. R. (2000) Cancer Genet. Cytogenet. 116, 133-141. [PubMed]
18. Seymour, A. B., Hruban, R. H., Redston, M., Caldas, C., Powell, S. M., Kinzler, K. W., Yeo, C. J. & Kern, S. E. (1994) Cancer Res. 54, 2761-2764. [PubMed]
19. Hahn, S. A., Seymour, A. B., Hoque, A. T., Schutte, M., da Costa, L. T., Redston, M. S., Caldas, C., Weinstein, C. L., Fischer, A., Yeo, C. J., et al. (1995) Cancer Res. 55, 4670-4675. [PubMed]
20. Kimura, M., Abe, T., Sunamura, M., Matsuno, S. & Horii, A. (1996) Genes Chromosomes Cancer 17, 88-93. [PubMed]
21. Pollack, J. R., Perou, C. M., Alizadeh, A. A., Eisen, M. B., Pergamenschikov, A., Williams, C. F., Jeffrey, S. S., Botstein, D. & Brown, P. O. (1999) Nat. Genet. 23, 41-46. [PubMed]
22. Olshen, A. B. & Venkatraman, E. S. (2002) ASA Proceedings of the Joint Statistical Meetings, pp. 2530-2535.
23. Ginzinger, D. G. (2002) Exp. Hematol. 30, 503-512. [PubMed]
24. Golub, T. R., Slonim, D. K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, J. P., Coller, H., Loh, M. L., Downing, J. R., Caligiuri, M. A., et al. (1999) Science 286, 531-537. [PubMed]
25. Hyman, E., Kauraniemi, P., Hautaniemi, S., Wolf, M., Mousses, S., Rozenblum, E., Ringner, M., Sauter, G., Monni, O., Elkahloun, A., et al. (2002) Cancer Res. 62, 6240-6245. [PubMed]
26. Lucito, R., Healy, J., Alexander, J., Reiner, A., Esposito, D., Chi, M., Rodgers, L., Brady, A., Sebat, J., Troge, J., et al. (2003) Genome Res. 13, 2291-2305. [PMC free article] [PubMed]
27. Abe, T., Makino, N., Furukawa, T., Ouyang, H., Kimura, M., Yatsuoka, T., Yokoyama, T., Inoue, H., Fukushige, S., Hoshi, M., et al. (1999) Genes Chromosomes Cancer 25, 60-64. [PubMed]
28. Albertson, D. G., Ylstra, B., Segraves, R., Collins, C., Dairkee, S. H., Kowbel, D., Kuo, W. L., Gray, J. W. & Pinkel, D. (2000) Nat. Genet. 25, 144-146. [PubMed]
29. Pollack, J. R., Sorlie, T., Perou, C. M., Rees, C. A., Jeffrey, S. S., Lonning, P. E., Tibshirani, R., Botstein, D., Borresen-Dale, A. L. & Brown, P. O. (2002) Proc. Natl. Acad. Sci. USA 99, 12963-12968. [PMC free article] [PubMed]
30. Platzer, P., Upender, M. B., Wilson, K., Willis, J., Lutterbaugh, J., Nosrati, A., Willson, J. K., Mack, D., Ried, T. & Markowitz, S. (2002) Cancer Res. 62, 1134-1138. [PubMed]
31. Rozenblum, E., Schutte, M., Goggins, M., Hahn, S. A., Panzer, S., Zahurak, M., Goodman, S. N., Sohn, T. A., Hruban, R. H., Yeo, C. J., et al. (1997) Cancer Res. 57, 1731-1734. [PubMed]
32. Caldas, C., Hahn, S. A., da Costa, L. T., Redston, M. S., Schutte, M., Seymour, A. B., Weinstein, C. L., Hruban, R. H., Yeo, C. J. & Kern, S. E. (1994) Nat. Genet. 8, 27-32. [PubMed]
33. Furukawa, T., Sunamura, M., Motoi, F., Matsuno, S. & Horii, A. (2003) Am. J. Pathol. 162, 1807-1815. [PMC free article] [PubMed]
34. Neil, J. C. & Cameron, E. R. (2002) Cancer Cell 2, 253-255. [PubMed]
35. Akagi, K., Suzuki, T., Stephens, R. M., Jenkins, N. A. & Copeland, N. G. (2004) Nucleic Acids Res. 32, D523-D527. [PMC free article] [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • MedGen
    MedGen
    Related information in MedGen
  • PubMed
    PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...