[Show abstract][Hide abstract] ABSTRACT: Angiogenesis, the recruitment and re-configuration of pre-existing vasculature, is essential for tumor growth and metastasis. Increased tumor vascularization often correlates with poor patient outcomes in a broad spectrum of carcinomas. We identified four jointed box 1 (FJX1) as a candidate regulator of tumor angiogenesis in colorectal cancer. FJX1 mRNA and protein are upregulated in human colorectal tumor epithelium as compared with normal epithelium and colorectal adenomas, and high expression of FJX1 is associated with poor patient prognosis. FJX1 mRNA expression in colorectal cancer tissues is significantly correlated with changes in known angiogenesis genes. Augmented expression of FJX1 in colon cancer cells promotes growth of xenografts in athymic mice and is associated with increased tumor cell proliferation and vascularization. Furthermore, FJX1 null mice develop significantly fewer colonic polyps than wild-type littermates after combined dextran sodium sulfate (DSS) and azoxymethane (AOM) treatment. In vitro, conditioned media from FJX1 expressing cells promoted endothelial cell capillary tube formation in a HIF1-α dependent manner. Taken together our results support the conclusion that FJX1 is a novel regulator of tumor progression, due in part, to its effect on tumor vascularization.
PLoS ONE 01/2013; 8(7):e69660. · 3.53 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: The purpose of this preclinical study was to determine the effectiveness of RAF265, a multikinase inhibitor, for treatment of human metastatic melanoma and to characterize traits associated with drug response.
Advanced metastatic melanoma tumors from 34 patients were orthotopically implanted to nude mice. Tumors that grew in mice (17 of 34) were evaluated for response to RAF265 (40 mg/kg, every day) over 30 days. The relation between patient characteristics, gene mutation profile, global gene expression profile, and RAF265 effects on tumor growth, mitogen-activated protein/extracellular signal-regulated kinase (MEK)/extracellular signal-regulated kinase (ERK) phosphorylation, proliferation, and apoptosis markers was evaluated.
Nine of the 17 tumors that successfully implanted (53%) were mutant BRAF (BRAF(V600E/K)), whereas eight of 17 (47%) tumors were BRAF wild type (BRAF(WT)). Tumor implants from 7 of 17 patients (41%) responded to RAF265 treatment with more than 50% reduction in tumor growth. Five of the 7 (71%) responders were BRAF(WT), of which 1 carried c-KIT(L576P) and another N-RAS(Q61R) mutation, while only 2 (29%) of the responding tumors were BRAF(V600E/K). Gene expression microarray data from nonimplanted tumors revealed that responders exhibited enriched expression of genes involved in cell growth, proliferation, development, cell signaling, gene expression, and cancer pathways. Although response to RAF265 did not correlate with pERK1/2 reduction, RAF265 responders did exhibit reduced pMEK1, reduced proliferation based upon reduced Ki-67, cyclin D1 and polo-like kinase1 levels, and induction of the apoptosis mediator BCL2-like 11.
Orthotopic implants of patient tumors in mice may predict prognosis and treatment response for melanoma patients. A subpopulation of human melanoma tumors responds to RAF265 and can be characterized by gene mutation and gene expression profiles.
Clinical Cancer Research 02/2012; 18(8):2184-98. · 7.84 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: VAPB (VAMP- associated protein B) is an ER protein that regulates multiple biological functions. Although aberrant expression of VAPB is associated with breast cancer, its function in tumor cells is poorly understood. In this report, we provide evidence that VAPB regulates breast tumor cell proliferation and AKT activation. VAPB protein expression is elevated in primary and metastatic tumor specimens, and VAPB mRNA expression levels correlated negatively with patient survival in two large breast tumor datasets. Overexpression of VAPB in mammary epithelial cells increased cell growth, whereas VAPB knockdown in tumor cells inhibited cell proliferation in vitro and suppressed tumor growth in orthotopic mammary gland allografts. The growth regulation of mammary tumor cells controlled by VAPB appears to be mediated, at least in part, by modulation of AKT activity. Overexpression of VAPB in MCF10A-HER2 cells enhances phosphorylation of AKT. In contrast, knockdown of VAPB in MMTV-Neu tumor cells inhibited pAKT levels. Pharmacological inhibition of AKT significantly reduced three-dimensional spheroid growth induced by VAPB. Collectively, the genetic, functional and mechanistic analyses suggest a role of VAPB in tumor promotion in human breast cancer.
PLoS ONE 01/2012; 7(10):e46281. · 3.53 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Most estrogen receptor α (ER)-positive breast cancers initially respond to antiestrogens, but many eventually become estrogen-independent and recur. We identified an estrogen-independent role for ER and the CDK4/Rb/E2F transcriptional axis in the hormone-independent growth of breast cancer cells. ER downregulation with fulvestrant or small interfering RNA (siRNA) inhibited estrogen-independent growth. Chromatin immunoprecipitation identified ER genomic binding activity in estrogen-deprived cells and primary breast tumors treated with aromatase inhibitors. Gene expression profiling revealed an estrogen-independent, ER/E2F-directed transcriptional program. An E2F activation gene signature correlated with a lesser response to aromatase inhibitors in patients' tumors. siRNA screening showed that CDK4, an activator of E2F, is required for estrogen-independent cell growth. Long-term estrogen-deprived cells hyperactivate phosphatidylinositol 3-kinase (PI3K) independently of ER/E2F. Fulvestrant combined with the pan-PI3K inhibitor BKM120 induced regression of ER(+) xenografts. These data support further development of ER downregulators and CDK4 inhibitors, and their combination with PI3K inhibitors for treatment of antiestrogen-resistant breast cancers.
Cancer Discovery 09/2011; 1(4):338-51. · 15.93 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Trefoil factor 1 (TFF1) is a tumor suppressor gene that encodes a peptide belonging to the trefoil factor family of protease-resistant peptides. Although TFF1 expression is frequently lost in gastric carcinomas, the tumorigenic pathways this affects have not been determined. Here we show that Tff1-knockout mice exhibit age-dependent carcinogenic histological changes in the pyloric antrum of the gastric mucosa, progressing from gastritis to hyperplasia, low-grade dysplasia, high-grade dysplasia, and ultimately malignant adenocarcinoma. The histology and molecular signatures of gastric lesions in the Tff1-knockout mice were consistent with an inflammatory phenotype. In vivo, ex-vivo, and in vitro studies showed that TFF1 expression suppressed TNF-α-mediated NF-κB activation through the TNF receptor 1 (TNFR1)/IκB kinase (IKK) pathway. Consistent with these mouse data, human gastric tissue samples displayed a progressive decrease in TFF1 expression and an increase in NF-κB activation along the multi-step carcinogenesis cascade. Collectively, these results provide evidence that loss of TFF1 leads to activation of IKK complex-regulated NF-κB transcription factors and is an important event in shaping the NF-κB-mediated inflammatory response during the progression to gastric tumorigenesis.
The Journal of clinical investigation 05/2011; 121(5):1753-67. · 15.39 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Regulatory T cells (Tregs) are a suppressive subset of CD4(+) T lymphocytes implicated in the prevention of acute GVHD (aGVHD) after allo-SCT (ASCT). To determine whether increased frequency of Tregs with a skin-homing (cutaneous lymphocyte Ag, CLA(+)) or a gut-homing (α(4)β(7)(+)) phenotype is associated with reduced risk of skin or gut aGVHD, respectively, we quantified circulating CLA(+) or α(4)β(7)(+) on Tregs at the time of neutrophil engraftment in 43 patients undergoing ASCT. Increased CLA(+) Tregs at engraftment was associated with the prevention of skin aGVHD (2.6 vs 1.7%; P=0.038 (no skin aGVHD vs skin aGVHD)), and increased frequencies of CLA(+) and α(4)β(7)(+) Tregs were negatively correlated with severity of skin aGVHD (odds ratio (OR), 0.67; 95% confidence interval (CI), 0.46-0.98; P=0.041) or gut aGVHD (OR, 0.93; 95% CI, 0.88-0.99; P=0.031), respectively. This initial report suggests that Treg tissue-homing subsets help to regulate organ-specific risk and severity of aGVHD after human ASCT. These results need to be validated in a larger, multicenter cohort.
Bone marrow transplantation 03/2011; 46(3):436-42. · 3.00 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: The transcription factor p73 plays critical roles during development and tumorigenesis. It exhibits sequence identity and structural homology with p53, and can engage p53-like tumor-suppressive programs. However, different pathways regulate p53 and p73, and p73 is not mutated in human tumors. Therefore, p73 represents a therapeutic target, and there is a critical need to understand genes and noncoding RNAs regulated by p73 and how they change during treatment regimens. Here, we define the p73 genomic binding profile and demonstrate its modulation by rapamycin, an inhibitor of mammalian target of rapamycin (mTOR) and inducer of p73. Rapamycin selectively increased p73 occupancy at a subset of its binding sites. In addition, multiple determinants of p73 binding, activity, and function were evident, and were modulated by mTOR. We generated an mTOR-p73 signature that is enriched for p73 target genes and miRNAs that are involved in mesenchymal differentiation and tumorigenesis, can classify rhabdomyosarcomas by clinical subtype, and can predict patient outcome.
Proceedings of the National Academy of Sciences 02/2011; 108(5):2076-81. · 9.81 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: It is well known that tumor-derived proangiogenic factors induce neovascularization to facilitate tumor growth and malignant progression. However, the concept of "angiocrine" signaling, in which signals produced by endothelial cells elicit tumor cell responses distinct from vessel function, has been proposed, yet remains underinvestigated. Here, we report that angiocrine factors secreted from endothelium regulate tumor growth and motility. We found that Slit2, which is negatively regulated by endothelial EphA2 receptor, is one such tumor suppressive angiocrine factor. Slit2 activity is elevated in EphA2-deficient endothelium. Blocking Slit activity restored angiocrine-induced tumor growth/motility, whereas elevated Slit2 impaired growth/motility. To translate our findings to human cancer, we analyzed EphA2 and Slit2 expression in human cancer. EphA2 expression inversely correlated with Slit2 in the vasculature of invasive human ductal carcinoma samples. Moreover, analysis of large breast tumor data sets revealed that Slit2 correlated positively with overall and recurrence-free survival, providing clinical validation for the tumor suppressor function for Slit2 in human breast cancer. Together, these data support a novel, clinically relevant mechanism through which EphA2 represses Slit2 expression in endothelium to facilitate angiocrine-mediated tumor growth and motility by blocking a tumor suppressive signal.
Cancer Research 02/2011; 71(3):976-87. · 9.28 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Metallothionein 3 (MT3) maintains intracellular metal homeostasis and protects against reactive oxygen species (ROS)-induced DNA damage. In this study, we investigated the epigenetic alterations and gene expression of the MT3 gene in esophageal adenocarcinomas (EACs).
Using quantitative bisulfite pyrosequencing, we detected unique DNA methylation profiles in the MT3 promoter region. The CpG nucleotides from -372 to -306 from the transcription start site (TSS) were highly methylated in tumor (n = 64) and normal samples (n = 51), whereas CpG nucleotides closest to the TSS (-4 and +3) remained unmethylated in all normal and most tumor samples. Conversely, CpG nucleotides in two regions (from -139 to -49 and +296 to +344) were significantly hypermethylated in EACs as compared to normal samples [FDR<0.001, -log10(FDR)>3.0]. The DNA methylation levels from -127 to -8 CpG sites showed the strongest correlation with MT3 gene expression (r = -0.4, P<0.0001). Moreover, the DNA hypermethylation from -127 to -8 CpG sites significantly correlated with advanced tumor stages and lymph node metastasis (P = 0.005 and P = 0.0313, respectively). The ChIP analysis demonstrated a more repressive histone modification (H3K9me2) and less active histone modifications (H3K4me2, H3K9ace) in OE33 cells than in FLO-1 cells; concordant with the presence of higher DNA methylation levels and silencing of MT3 expression in OE33 as compared to FLO-1 cells. Treatment of OE33 cells with 5-Aza-deoxycitidine restored MT3 expression with demethylation of its promoter region and reversal of the histone modifications towards active histone marks.
In summary, EACs are characterized by frequent epigenetic silencing of MT3. The choice of specific regions in the CpG island is a critical step in determining the functional role and prognostic value of DNA methylation in cancer cells.
PLoS ONE 01/2011; 6(7):e22009. · 3.53 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Pre-clinical studies provide compelling evidence that Eph family receptor tyrosine kinases (RTKs) and ligands promote cancer growth, neovascularization, invasion, and metastasis. Tumor suppressive roles have also been reported for the receptors, however, creating a potential barrier for clinical application. Determining how these observations relate to clinical outcome is a crucial step for translating the biological and mechanistic data into new molecularly targeted therapies. We investigated eph and ephrin expression in human breast cancer relative to endpoints of overall and/or recurrence-free survival in large microarray datasets. We also investigated protein expression in commercial human breast tissue microarrays (TMA) and Stage I prognostic TMAs linked to recurrence outcome data. We found significant correlations between ephA2, ephA4, ephA7, ephB4, and ephB6 and overall and/or recurrence-free survival in large microarray datasets. Protein expression in TMAs supported these trends. While observed no correlation between ephrin ligand expression and clinical outcome in microarray datasets, ephrin-A1 and EphA2 protein co-expression was significantly associated with recurrence in Stage I prognostic breast cancer TMAs. Our data suggest that several Eph family members are clinically relevant and tractable targets for intervention in human breast cancer. Moreover, profiling Eph receptor expression patterns in the context of relevant ligands and in the context of stage may be valuable in terms of diagnostics and treatment.
PLoS ONE 01/2011; 6(9):e24426. · 3.53 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: To examine the effects of Roux-en-Y gastric bypass (RYGB) surgery with and without laparoscopic removal of omental fat (omentectomy) on the temporal gene expression profiles of skeletal muscle.
Previously reported were the whole-body metabolic effects of a randomized, single-blinded study in patients receiving RYGB surgery stratified to receive or not receive omentectomy. In this follow up study we report on changes in skeletal muscle gene expression in a subset of 21 patients, for whom biopsies were collected preoperatively and at either 6 months or 12 months postoperatively.
RNA isolated from skeletal muscle biopsies of 21 subjects (8 without omentectomy and 13 with omentectomy) taken before RYGB or at 6 and 12 months postoperatively were subjected to gene expression profiling via Exon 1.0 S/T Array and Taqman Low Density Array. Robust Multichip Analysis and gene enrichment data analysis revealed 84 genes with at least a 4-fold expression difference after surgery. At 6 and 12 months the RYGB with omentectomy group displayed a greater reduction in the expression of genes associated with skeletal muscle inflammation (ANKRD1, CDR1, CH25H, CXCL2, CX3CR1, IL8, LBP, NFIL3, SELE, SOCS3, TNFAIP3, and ZFP36) relative to the RYGB non-omentectomy group. Expressions of IL6 and CCL2 were decreased at all postoperative time points. There was differential expression of genes driving protein turnover (IGFN1, FBXW10) in both groups over time and increased expression of PAAF1 in the non-omentectomy group at 12 months. Evidence for the activation of skeletal muscle satellite cells was inferred from the up-regulation of HOXC10. The elevated post-operative expression of 22 small nucleolar RNAs and the decreased expression of the transcription factors JUNB, FOS, FOSB, ATF3 MYC, EGR1 as well as the orphan nuclear receptors NR4A1, NR4A2, NR4A3 suggest dramatic reorganizations at both the cellular and genetic levels.
These data indicate that RYGB reduces skeletal muscle inflammation, and removal of omental fat further amplifies this response.
PLoS ONE 01/2011; 6(12):e28577. · 3.53 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Small-cell lung cancer (SCLC) is the most aggressive subtype of lung cancer in its clinical behavior, with a 5-year overall survival as low as 5%. Despite years of research in the field, molecular determinants of SCLC behavior are still poorly understood, and this deficiency has translated into an absence of specific diagnostics and targeted therapeutics. We hypothesized that tumor DNA copy number alterations would allow the identification of molecular pathways involved in SCLC progression. Array comparative genomic hybridization was performed on DNA extracted from 46 formalin-fixed paraffin-embedded SCLC tissue specimens. Genomic profiling of tumor and sex-matched control DNA allowed the identification of 70 regions of copy number gain and 55 regions of copy number loss. Using molecular pathway analysis, we found a strong enrichment in these regions of copy number alterations for 11 genes associated with the focal adhesion pathway. We verified these findings at the genomic, gene expression and protein level. Focal Adhesion Kinase (FAK), one of the central genes represented in this pathway, was commonly expressed in SCLC tumors and constitutively phosphorylated in SCLC cell lines. Those were poorly adherent to most substrates but not to laminin-322. Inhibition of FAK phosphorylation at Tyr(397) by a small-molecule inhibitor, PF-573,228, induced a dose-dependent decrease of adhesion and an increase of spreading in SCLC cell lines on laminin-322. Cells that tended to spread also showed a decrease in focal adhesions, as demonstrated by a decreased vinculin expression. These results support the concept that pathway analysis of genes in regions of copy number alterations may uncover molecular mechanisms of disease progression and demonstrate a new role of FAK and associated adhesion pathways in SCLC. Further investigations of FAK at the functional level may lead to a better understanding of SCLC progression and may have therapeutic implications.
[Show abstract][Hide abstract] ABSTRACT: Staging inadequately predicts metastatic risk in patients with colon cancer. We used a gene expression profile derived from invasive, murine colon cancer cells that were highly metastatic in an immunocompetent mouse model to identify patients with colon cancer at risk of recurrence.
This phase 1, exploratory biomarker study used 55 patients with colorectal cancer from Vanderbilt Medical Center (VMC) as the training dataset and 177 patients from the Moffitt Cancer Center as the independent dataset. The metastasis-associated gene expression profile developed from the mouse model was refined with comparative functional genomics in the VMC gene expression profiles to identify a 34-gene classifier associated with high risk of metastasis and death from colon cancer. A metastasis score derived from the biologically based classifier was tested in the Moffitt dataset.
A high score was significantly associated with increased risk of metastasis and death from colon cancer across all pathologic stages and specifically in stage II and stage III patients. The metastasis score was shown to independently predict risk of cancer recurrence and death in univariate and multivariate models. For example, among stage III patients, a high score translated to increased relative risk of cancer recurrence (hazard ratio, 4.7; 95% confidence interval, 1.566-14.05). Furthermore, the metastasis score identified patients with stage III disease whose 5-year recurrence-free survival was >88% and for whom adjuvant chemotherapy did not increase survival time.
A gene expression profile identified from an experimental model of colon cancer metastasis predicted cancer recurrence and death, independently of conventional measures, in patients with colon cancer.
[Show abstract][Hide abstract] ABSTRACT: Tobacco smoking is responsible for 85% of all lung cancers. To further our understanding of the molecular pathogenesis of lung cancer, we determined whether smoking history leads to the emergence of specific genomic alterations found in non-small cell lung cancer (NSCLC).
To identify gene copy number alterations in NSCLCs associated with smoking history or DNA repair capacity.
Seventy-five NSCLCs were selected for this study from patients with current, none, or past smoking history, including pack year information. Tissue sections were microdissected, and DNA was extracted, purified, and labeled by random priming before hybridization onto bacterial artificial chromosome (BAC) arrays. Normalized ratios were correlated with smoking history and DNA repair capacity was measured by an in vitro lymphocyte assay in the same patients.
We identified smoking-related genomic signatures in NSCLCs that could be predicted with an overall 74% accuracy. Lung tumors arising from current-smokers had the greatest number of copy number alterations. The genomic regions most significantly associated with smoking were located within 60 regions and were functionally associated with genes controlling the M phase of the cell cycle, the segregation of chromosomes, and the methylation of DNA. Verification of the data is provided from data in the public domain and by quantitative real-time polymerase chain reaction. The associations between genomic abnormalities and DNA repair capacity did not reach statistical significance.
These findings indicate that smoking history leaves a specific genomic signature in the DNA of lung tumors and suggest that these alterations may reflect new molecular pathways to cancer development.
American Journal of Respiratory and Critical Care Medicine 10/2008; 178(11):1164-72. · 11.04 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Genetic differences in endothelial biology could underlie development of phenotypic heterogeneity among persons afflicted with vascular diseases. We obtained blood outgrowth endothelial cells from 20 subjects with sickle cell anemia (age, 4-19 years) shown to be either at-risk (n=11) or not-at-risk (n=9) for ischemic stroke because of, respectively, having or not having occlusive disease at the circle of Willis. Gene expression profiling identified no significant single gene differences between the 2 groups, as expected. However, analysis of Biological Systems Scores, using gene sets that were predetermined to survey each of 9 biologic systems, showed that only changes in inflammation signaling are characteristic of the at-risk subjects, as supported by multiple statistical approaches. Correspondingly, subsequent biologic testing showed significantly exaggerated RelA activation on the part of blood outgrowth endothelial cells from the at-risk subjects in response to stimulation with interleukin-1beta/tumor necrosis factoralpha. We conclude that the pathobiology of circle of Willis disease in the child with sickle cell anemia predominantly involves inflammation biology, which could reflect differences in genetically determined endothelial biology that account for differing host responses to inflammation.
[Show abstract][Hide abstract] ABSTRACT: Since the available microarray data of BOEC (human blood outgrowth endothelial cells), large vessel, and microvascular endothelial cells were from two different platforms, a working cross-platform normalization method was needed to make these data comparable. With six HUVEC (human umbilical vein endothelial cells) samples hybridized on two-channel cDNA arrays and six HUVEC samples on Affymetrix arrays, 64 possible combinations of a three-step normalization procedure were investigated to search for the best normalization method, which was selected, based on two criteria measuring the extent to which expression profiles of biological samples of the same cell type arrayed on two platforms were indistinguishable. Next, three discriminative gene lists between the large vessel and the microvascular endothelial cells were achieved by SAM (significant analysis of microarrays), PAM (prediction analysis for microarrays), and a combination of SAM and PAM lists. The final discriminative gene list was selected by SVM (support vector machine). Based on this discriminative gene list, SVM classification analysis with best tuning parameters and 10,000 times of validations showed that BOEC were far from large vessel cells, they either formed their own class, or fell into the microvascular class. Based on all the common genes between the two platforms, SVM analysis further confirmed this conclusion.
Journal of Bioinformatics and Computational Biology 09/2007; 5(4):875-93. · 0.93 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Promoter hypermethylation coupled with loss of heterozygosity at the same locus results in loss of gene function in many tumor cells. The "rules" governing which genes are methylated during the pathogenesis of individual cancers, how specific methylation profiles are initially established, or what determines tumor type-specific methylation are unknown. However, DNA methylation markers that are highly specific and sensitive for common tumors would be useful for the early detection of cancer, and those required for the malignant phenotype would identify pathways important as therapeutic targets.
In an effort to identify new cancer-specific methylation markers, we employed a high-throughput global expression profiling approach in lung cancer cells. We identified 132 genes that have 5' CpG islands, are induced from undetectable levels by 5-aza-2'-deoxycytidine in multiple non-small cell lung cancer cell lines, and are expressed in immortalized human bronchial epithelial cells. As expected, these genes were also expressed in normal lung, but often not in companion primary lung cancers. Methylation analysis of a subset (45/132) of these promoter regions in primary lung cancer (n = 20) and adjacent nonmalignant tissue (n = 20) showed that 31 genes had acquired methylation in the tumors, but did not show methylation in normal lung or peripheral blood cells. We studied the eight most frequently and specifically methylated genes from our lung cancer dataset in breast cancer (n = 37), colon cancer (n = 24), and prostate cancer (n = 24) along with counterpart nonmalignant tissues. We found that seven loci were frequently methylated in both breast and lung cancers, with four showing extensive methylation in all four epithelial tumors.
By using a systematic biological screen we identified multiple genes that are methylated with high penetrance in primary lung, breast, colon, and prostate cancers. The cross-tumor methylation pattern we observed for these novel markers suggests that we have identified a partial promoter hypermethylation signature for these common malignancies. These data suggest that while tumors in different tissues vary substantially with respect to gene expression, there may be commonalities in their promoter methylation profiles that represent targets for early detection screening or therapeutic intervention.
PLoS Medicine 01/2007; 3(12):e486. · 15.25 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: MOTIVATION: It is biologically interesting to address whether human blood outgrowth endothelial cells (BOECs) belong to or are closer to large vessel endothelial cells (LVECs) or microvascular endothelial cells (MVECs) based on global expression profiling. An earlier analysis using a hierarchical clustering and a small set of genes suggested that BOECs seemed to be closer to MVECs. By taking advantage of the two known classes, LVEC and MVEC, while allowing BOEC samples to belong to either of the two classes or to form their own new class, we take a semi-supervised learning approach; for high-dimensional data as encountered here, we propose a penalized mixture model with a weighted L1 penalty to realize automatic feature selection while fitting the model. RESULTS: We applied our penalized mixture model to a combined dataset containing 27 BOEC, 28 LVEC and 25 MVEC samples. Analysis results indicated that the BOEC samples appeared to form their own new class. A simulation study confirmed that, compared with the standard mixture model with or without initial variable selection, the penalized mixture model performed much better in identifying relevant genes and forming corresponding clusters. The penalized mixture model seems to be promising for high-dimensional data with the capability of novel class discovery and automatic feature selection.