Cancer Outlier Profile Analysis Research Articles

A new statistical approach for the identification of outlier genes in cancer microarray data

The aim of microarrays technology is to discover genes, which are differentially expressed as outliers between two or more groups of patients are an important task in the genomics community. The regular pattern of genes may often breakdown due to the presence of outliers and it is essential to detect those genes whose behavior looks abnormal in experimental and biological conditions. Several statistical techniques - t-statistic, cancer outlier profile analysis (COPA), outlier sums (OS), outlier robust t-statistic (ORT), maximum ordered subset t-statistics (MOST) and least sum of ordered subset square t-statistics (LSOSS) were developed to address the problem of detecting outlier genes in microarray data but these methods are affected by some problems especially if there is an unusual observation in such dataset then the standard assumptions of distribution parameter may be violated and these techniques might not be suitable to detect outliers genes as well. For these consequences, I have developed a new statistical technique that is “Propose t-statistic (PT)”. The performance of the newly proposed method PT statistic compare with the other existing methods applied to the monte carlo simulation data, package data, and real cancer datasets. The result shows that the outlier genes are identified by using the proposed method PT as well and will give the best and identical results than other methods. The performance of the proposed approach significantly improves than the traditional methods and it can extensively contribute to the medical as well as the genomic community.
 Asian J. Med. Biol. Res. December 2020, 6(4): 795-801

A meta-analysis of BRF2 as a prognostic biomarker in invasive breast carcinoma

BackgroundDeregulation of the RNA polymerase III specific TFIIIB subunit BRF2 occurs in subtypes of human cancers. However, correlations between BRF2 alterations and clinical outcomes in breast cancer are limited. We conducted this review to analyze BRF2 alterations in genomic data sets housed in Oncomine and cBioPortal to identify potential correlations between BRF2 alterations and clinical outcomes.MethodsThe authors queried both Oncomine and cBioPortal for alterations in BRF2 in human cancers and performed meta-analyses identifying significant correlations between BRF2 and clinical outcomes in invasive breast cancer (IBC).ResultsA meta cancer outlier profile analysis (COPA) of 715 data sets (86,733 samples) in Oncomine identified BRF2 as overexpressed in 60% of breast cancer data sets. COPA scores in IBC data sets (3594 patients) are comparable for HER2 (24.211, median gene rank 60) and BRF2 (29.656, median gene rank 36.5). Overall survival in IBC patients with BRF2 alterations (21%) is significantly decreased (p = 9.332e-3). IBC patients with BRF2 alterations aged 46 to 50 have a significantly poor survival outcome (p = 7.093e-3). Strikingly, in metastatic breast cancer, BRF2 is altered in 33% of women aged 45–50. BRF2 deletions are predominant in this age group.ConclusionThis study suggests BRF2 may be an prognostic biomarker in invasive breast carcinoma.

PAX8 regulon in human ovarian cancer links lineage dependency with epigenetic vulnerability to HDAC inhibitors.

PAX8 is a prototype lineage-survival oncogene in epithelial ovarian cancer. However, neither its underlying pro-tumorigenic mechanisms nor potential therapeutic implications have been adequately elucidated. Here, we identified an ovarian lineage-specific PAX8 regulon using modified cancer outlier profile analysis, in which PAX8-FGF18 axis was responsible for promoting cell migration in an autocrine fashion. An image-based drug screen pinpointed that PAX8 expression was potently inhibited by small-molecules against histone deacetylases (HDACs). Mechanistically, HDAC blockade altered histone H3K27 acetylation occupancies and perturbed the super-enhancer topology associated with PAX8 gene locus, resulting in epigenetic downregulation of PAX8 transcripts and related targets. HDAC antagonists efficaciously suppressed ovarian tumor growth and spreading as single agents, and exerted synergistic effects in combination with standard chemotherapy. These findings provide mechanistic and therapeutic insights for PAX8-addicted ovarian cancer. More generally, our analytic and experimental approach represents an expandible paradigm for identifying and targeting lineage-survival oncogenes in diverse human malignancies.

Canonical Wnt/β-catenin signaling activation in soft-tissue sarcomas: A comparative study of synovial sarcoma and leiomyosarcoma.

Background:Previous studies have shown that aberrant activation of the Wnt/β-catenin pathway is associated with many malignant neoplasms. This includes some soft-tissue sarcoma phenotypes, most notably synovial sarcoma, implicating potential targets for novel molecular therapies.Objective:We investigate the level of Wnt/β-catenin pathway activation present in leiomyosarcomas relative to synovial sarcomas, using expression of LEF1 and β-catenin as surrogates.Methods:Cancer outlier profile analysis was performed on messenger RNA expression datasets in Oncomine (70 synovial sarcomas, 178 leiomyosarcomas). Results for LEF1 and β-catenin messenger RNA expression were reported in terms of median-centered intensity. Separate immunohistochemical studies were performed on tissue microarrays created from 77 synovial sarcomas and 89 leiomyosarcomas using antibodies to LEF1 and β-catenin. Tumors with unequivocal strong nuclear staining involving ⩾5% of cells were interpreted as positive.Results:Cancer outlier profile analysis demonstrated a higher level of LEF1 messenger RNA expression in synovial sarcomas than in leiomyosarcomas (p < 0.0001), but showed no significant difference in β-catenin messenger RNA expression (p = 0.868). Immunohistochemistry showed most synovial sarcomas had strong nuclear expression of LEF1 (79%) and β-catenin (84%), while a small minority of leiomyosarcomas had strong nuclear expression of LEF1 (5%) and β-catenin (6%).Conclusion:These results provide further evidence that aberrant activation of the Wnt/β-catenin pathway is present in most synovial sarcomas, but not in most leiomyosarcomas. While targeting the constituents of this pathway might be effective in the treatment of synovial sarcomas, it is not likely to be an effective strategy in the treatment of leiomyosarcomas.

Downregulation of SAFB Sustains the NF-κB Pathway by Targeting TAK1 during the Progression of Colorectal Cancer.

Purpose: To investigate the role and the underlying mechanism of scaffold attachment factor B (SAFB) in the progression of colorectal cancer (CRC).Experimental Design: SAFB expression was analyzed in the Cancer Outlier Profile Analysis of Oncomine and in 175 paraffin-embedded archived CRC tissues. Gene Ontology analyses were performed to explore the mechanism of SAFB in CRC progression. Western blot, RT-PCR, luciferase assay, and chromatin immunoprecipitation (ChIP) were used to detect the regulation of transforming growth factor-β-activated kinase 1 (TAK1) and NF-κB signaling by SAFB The role of SAFB in invasion, metastasis, and angiogenesis was investigated using in vitro and in vivo assays. The relationship between SAFB and TAK1 was analyzed in CRC tissues.Results: SAFB was downregulated in CRC tissues, and low expression of SAFB was significantly associated with an aggressive phenotype and poorer survival of CRC patients. The downregulation of SAFB activated NF-κB signaling by targeting the TAK1 promoter. Ectopic expression of SAFB inhibited the development of aggressive features and metastasis of CRC cells both in vitro and in vivo The overexpression of TAK1 could rescue the aggressive features in SAFB-overexpressed cells. Furthermore, the expression of SAFB in CRC tissues was negatively correlated with the expression of TAK1- and NF-κB-related genes.Conclusions: Our results show that SAFB regulated the activity of NF-κB signaling in CRC by targeting TAK1 This novel mechanism provides a comprehensive understanding of both SAFB and the NF-κB signaling pathway in the progression of CRC and indicates that the SAFB-TAK1-NF-κB axis is a potential target for early therapeutic intervention in CRC progression. Clin Cancer Res; 23(22); 7108-18. ©2017 AACR.

Abstract 4552: ssCOPA: a novel method with improved sensitivity to assess statistical significance for cancer outlier profile analysis of gene expression

Abstract To understand the genetic mechanism of disease, finding biomarkers and drug targets are a major focus of academic and pharmaceutical research. However, it has become clear that many important genes driving diseases like cancer, are deregulated in a relatively small group of outliers, e.g. amplified and fusion genes. Cancer outlier profile analysis (COPA), introduced by Tomlins et al, has raised interest in detecting genes with outlier expression within a group, particularly in cancer patients. However, the magnitudes of the COPA scores are not directly comparable across different studies or with different input parameters. Also, integrating COPA scores across multiple studies in a meta-analysis is problematic. Current solutions to performing meta-analyses, including rank-based methods and binomial test, require large study numbers without adjustment for the differences in study size. Also, within a single study COPA scores alone do not provide a P-value, and it is impossible to know which outlier genes are statistically significant. Here we developed three novel methods to address these problems by computing a significance for COPA scores, including Bootstrap resampling, Bootstrap + generalized Pareto distribution (GPD) and Binomial method for statistical significance of COPA (ssCOPA), to enable the comparison of COPA score across studies or parameters. We first evaluated our methods using two breast cancer studies. The P-values generated by Binomial method match closely to Bootstrap resampling using 10,000 randomizations, better than Bootstrap + GPD. In addition, the Binomial method requires significantly less computing time than the other methods. We further validated our methods with a meta-analysis of five breast cancer studies and compared the results with the top-ranking genes of a large-scale meta-analysis of 31 breast cancer studies by ONCOMINE. We show that our P-value-based approaches (P = 10-17~10-24) achieves significantly better concordance than the Binomial test (P = 10-14). Data and methods are available for use on the Illumina BaseSpace® CohortAnalyzer platform (formerly known as NextBio Clinical). Citation Format: Hong Gao, Sam Ng, Hinco J. Gierman. ssCOPA: a novel method with improved sensitivity to assess statistical significance for cancer outlier profile analysis of gene expression [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4552. doi:10.1158/1538-7445.AM2017-4552

Eya2, a Target Activated by Plzf, Is Critical for PLZF-RARA-Induced Leukemogenesis.

PLZF is a transcription factor that confers aberrant self-renewal in leukemogenesis, and the PLZF-RARA fusion gene causes acute promyelocytic leukemia (APL) through differentiation block. However, the molecular mechanisms of aberrant self-renewal underlying PLZF-mediated leukemogenesis are poorly understood. To investigate these mechanisms, comprehensive expression profiling of mouse hematopoietic stem/progenitor cells transduced with Plzf was performed, which revealed the involvement of a key transcriptional coactivator, Eya2, a target molecule shared by Plzf and PLZF-RARA, in the aberrant self-renewal. Indeed, PLZF-RARA as well as Plzf rendered those cells immortalized through upregulation of Eya2. Eya2 also led to immortalization without differentiation block, while depletion of Eya2 suppressed clonogenicity in cells immortalized by PLZF-RARA without influence on differentiation and apoptosis. Interestingly, cancer outlier profile analysis of human samples of acute myeloid leukemia (AML) in The Cancer Genome Atlas (TCGA) revealed a subtype of AML that strongly expressed EYA2 In addition, gene set enrichment analysis of human AML samples, including TCGA data, showed that this subtype of AML was more closely associated with the properties of leukemic stem cells in its gene expression signature than other AMLs. Therefore, EYA2 may be a target for molecular therapy in this subtype of AML, including PLZF-RARA APL.

Identification of MAGEA12 as a prognostic outlier gene in gastric cancers.

Melanoma antigen (MAGE) family genes are frequently over-expressed in asubset population of multiple cancers, and serve as idea therapeutic targets; however, their distribution pattern in gastric cancers has not yet been evaluated. In this study, we first performed acancer outlier profile analysis (COPA) on aseries of public gene expression datasets of gastric cancer, and identified MAGEA12 showing asignificant outlier expression model reproducibly. We further in silico validated that MAGEA12 outlier over-expression were associated with poor clinical outcome using six microarray datasets from GEO database. We then experimentally detected the MAGEA12 expression in an independent cohort of gastric cancer samples by immunohistochemistry, and showed that over-expression of MAGEA12 in asubset of cancers was associated with later stage and reduced survival; furthermore, MAGEA12 was an independent prognostic factor in an outlier manner. Our results indicate that MAGEA12 is anovel prognostic outlier gene in gastric cancers and patterns of MAGE expression may inform individualized targeted immunotherapies.

Abstract 4555: Identification of bone morphogenetic protein 2 splice variant in pancreatic cancer reveals cross-talk between endocytic and autophagy pathways

Abstract The inherent heterogeneity in pancreatic cancer (PC) confers complex transcription patterns that require a subset approach to identifying novel gene targets. Thus, we performed meta-analysis on 11 COPA (Cancer Outlier Profile Analysis) transformed datasets (www.oncomine.org) depicting the expression profiles of 416 pancreatic tumors. This analysis revealed Bone Morphogenetic Protein 2 (BMP-2), a pivotal regulator of cellular proliferation, motility, &amp; differentiation, to be increased in a subset of PC tumors. Canonical BMP-2 signaling is mediated via type 1 &amp; type 2 serine/threonine kinase receptors, which upon ligand binding, phosphorylate SMAD proteins leading to initiation or attenuation of gene expression programs. Receptor mediated endocytosis has been shown to be an integral regulator of BMP signaling intensity and duration whereby endocytic internalization of activated receptors leads to their subsequent inactivation and recycling back to the plasma membrane or degradation via the lysosomal pathway. Our investigation of BMP-2 in PC has revealed a novel splice variant (BMP-2 SV), which due to exclusion of its signal peptide bearing exon, escapes the secretory pathway and is endowed with an intracellular pattern of expression. Furthermore, this BMP-2 isoform undergoes preferential degradation via the autophagy lysosomal pathway (ALP). Endosomal and autophagosomal trafficking has been previously shown to converge, through endosome and autophagosome vesicular fusion, to form a hybrid structure described as an amphisome. We hypothesize that endocytosed bone morphogenetic receptors and autophagocytosed BMP-2 SV converge in the amphisome thereby promoting signal transduction. Indeed, immunofluorescence analysis has revealed colocalization of the exogenous BMP-2 SV with components of both the endocytic and autophagic pathways. Furthermore, overexpression of BMP-2 SV in HEK-293 and pancreatic cancer cell lines was performed, which led to a corresponding increased Smad 1/5/8 phosphorylation. In contrast, biochemical inhibition of receptor mediated endocytosis with PitStop ® 2 or disruption of microtubule-dependent vesicular transport with vinblastine was found to diminish the BMP-SV dependent increase in Smad 1/5/8 phosphorylation. Likewise, increasing autophagy through Rapamycin treatment led to a corresponding increase in Smad 1/5/8 phosphorylation. Altogether, our results suggest a novel amphisome-mediated mechanism of BMP-2 signaling due to convergence of endocytosed receptors and autophagocytosed BMP-2 SV. Consequently, this intracellular mode of signaling may endow pancreatic tumor cells with a cell-autonomous means of driving its own growth and metastasis. Citation Format: Eric Cruz, Surinder K. Batra. Identification of bone morphogenetic protein 2 splice variant in pancreatic cancer reveals cross-talk between endocytic and autophagy pathways. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 4555.

Identification of a new subclass of ALK-negative ALCL expressing aberrant levels of ERBB4 transcripts

AbstractAnaplastic large-cell lymphoma (ALCL) is a clinical and biological heterogeneous disease that includes systemic anaplastic lymphoma kinase (ALK)-positive and ALK-negative entities. To discover biomarkers and/or genes involved in ALK-negative ALCL pathogenesis, we applied the cancer outlier profile analysis algorithm to a gene expression profiling data set including 249 cases of T-cell non-Hodgkin lymphoma and normal T cells. Ectopic coexpression of ERBB4 and COL29A1 genes was detected in 24% of ALK-negative ALCL patients. RNA sequencing and 5′ RNA ligase-mediated rapid amplification of complementary DNA ends identified 2 novel ERBB4-truncated transcripts displaying intronic transcription start sites. By luciferase assays, we defined that the expression of ERBB4-aberrant transcripts is promoted by endogenous intronic long terminal repeats. ERBB4 expression was confirmed at the protein level by western blot analysis and immunohistochemistry. Lastly, we demonstrated that ERBB4-truncated forms show oncogenic potentials and that ERBB4 pharmacologic inhibition partially controls ALCL cell growth and disease progression in an ERBB4-positive patient-derived tumorgraft model. In conclusion, we identified a new subclass of ALK-negative ALCL characterized by aberrant expression of ERBB4-truncated transcripts carrying intronic 5′ untranslated regions.

WISP3 is highly expressed in a subset of colorectal carcinomas with a better prognosis.

Outlier genes with marked overexpression in subsets of cancers like ERBB2 have potential for the identification of gene classifiers and therapeutic targets for the appropriate subpopulation. In this study, using the cancer outlier profile analysis strategy, we identified WNT1-inducible-signaling pathway protein 3 (WISP3) as an outlier gene that is highly expressed in a subset of colorectal cancers (CRCs) from The Cancer Genome Atlas dataset. A meta-cancer outlier profile analysis and immunohistochemistry experiment to validate the outlier expression model of WISP3 in CRC was then performed. Our immunohistochemical results indicated that WISP3 was more frequently seen in the small tumors, and there was a significant association between its overexpression with a good prognosis. Furthermore, in the multivariable model, WISP3 outlier expression retained significance for overall survival. In summary, in this study, we identified an outlier gene WISP3 overexpressed in a subset of CRC having less aggressive characteristics and a better prognosis. We suggest WISP3 may provide more accurate and precise information regarding CRC population classification.

Abstract B2-21: Identification of somatic RNA splicing alterations in human cancers

Abstract Recent whole-exome sequencing studies have found that somatic mutations frequently occur in splicing factors across multiple cancer types, supporting the need to systematically and globally characterize splicing alterations across human cancers. Somatic alterations that affect RNA processing of cancer genes may be caused by multiple mechanisms, including mutations in splicing factors, mutations in cis-acting splice sites and other regulatory sequences, alterations in methylation, or alterations in chromatin structure. For a global look at somatic RNA splicing alterations, we are applying the Cancer Outlier Profile Analysis (COPA) approach on ~7,000 cancer transcriptomes from 12 tumor types in The Cancer Genome Atlas (TCGA) and investigating the underlying somatic mutations that cause these splicing alterations using matched whole-exome, whole-genome, and methylation data from these samples. To identify and quantify alternative splicing in RNA-Seq data, including unannotated splicing events, we have further developed a computational pipeline called JuncBASE. To distinguish between cancer-specific splicing alterations and normal transcriptome variation, we are utilizing ~700 RNA-Seq libraries from healthy individuals from the Genotype-Tissue Expression (GTEx) project. COPA analysis of outlier splicing events in lung adenocarcinomas and glioblastomas successfully identified known altered splicing events in MET and EGFR, respectively, that are caused by DNA level somatic mutations. The genomic mechanisms for novel somatic splicing events in known cancer genes such as FGFR3, as well as genes previously uncharacterized in cancer genomes, are currently being investigated. To identify splicing events that may be novel somatic driver alterations, these events have been profiled using RNA-Seq data from the Cancer Cell Line Encyclopedia and are being used as biomarkers to identify genetic vulnerabilities in high-throughput shRNA screens. This work will have a significant impact on the identification of somatic splicing events that contribute to cancer pathogenesis. Citation Format: Angela N. Brooks, Yawei Ge, Kevin Chau, Samuel S. Freeman, Gordon Saksena, Chandra Sekhar Pedamallu, Matthew Meyerson. Identification of somatic RNA splicing alterations in human cancers. [abstract]. In: Proceedings of the AACR Special Conference on Computational and Systems Biology of Cancer; Feb 8-11 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 2):Abstract nr B2-21.

Identification of a New Subclass of ALK Negative Anaplastic Large Cell Lymphoma Expressing Aberrant Levels of ERBB4 Transcripts

Anaplastic Large Cell Lymphoma (ALCL) is a clinical and biological heterogeneous disease including the ALK+ and ALK- systemic forms. While ALK+ ALCL are molecularly characterized and can be readily diagnosed, no specific markers and molecular events leading to ALK- ALCL transformation have been identified so far.To discover biomarkers and/or genes potentially involved in ALK- ALCL pathogenesis, we applied the Cancer Outlier Profile Analysis (COPA) algorithm to a large gene expression profiling (GEP) data set including 249 cases of T-NHLs and normal T-cells. Among the top outliers, ERBB4 and COL29A1 genes were exclusively expressed in a subset of ALK- ALCL cases, and resulted highly correlated. Differential analysis comparing the expression profiles of ERBB4/COL29A1+ to ALK+ ALCL patients identified 48 genes up- and 37 down-regulated (qval < 0.0076). The most significantly over-represented functional category included transcriptional regulators. Gene-set-enrichment-analysis (GSEA) indicated that ERBB4/COL29A1+ samples shared activation of specific pathways such as inflammatory response, TNF-NF-κB, JAK-STAT, cytokine, and angiogenesis. RNA sequencing and 5’RNA Ligase Mediated Rapid Amplification of cDNA Ends (RLM-RACE) identified two ERBB4 truncated transcripts (referred as I20ΔERBB4 and I12ΔERBB4), displaying Transcription Starting Sites (TSS) in intron 12 and 20, respectively. On the contrary, full length COL29A1 mRNA was expressed in ERBB4/COL29A1+ samples. RT-qPCR analysis performed on 170 T-NHL samples (51 PCTL-NOS, 44 ALK+ ALCL, and 75 ALK- ALCL) highlighted specific expression of ERBB4 aberrant transcripts in 25% of ALK- ALCL (18 out of 75). ERBB4+ patients showed higher expression of I20ΔERBB4 as compared to I12ΔERBB4 transcript. ERBB4 expression at protein level was confirmed by immunohistochemistry and Western Blotting on selected cases. Inspection of intronic regions spanning two hypothetical TSS revealed the presence of Human Endogenous Retrovirus (HERV) Long Terminal Repeats (LTR) displaying promoter activity, as demonstrated by luciferase assays.In conclusion, we defined a new subclass of ALK- ALCL characterized by ectopic expression of ERBB4 and COL29A1 genes. To the best of our knowledge, ERBB4 truncated transcripts carrying intronic 5’UTR have never been described before. Further studies are required to address whether the expression of ERBB4 aberrant transcripts may contribute to the ALCL transformation and lymphoma maintenance. This information might lead to more rational therapeutic approaches for ERBB4+ ALCL patients. DisclosuresNo relevant conflicts of interest to declare.

Comparison of Outlier Techniques Based on Simulated Data

This research work employed a simulation study to evaluate six outlier techniques: t-Statistic, Modified Z-Statistic, Cancer Outlier Profile Analysis (COPA), Outlier Sum-Statistic (OS), Outlier Robust t-Statistic (ORT), and the Truncated Outlier Robust t-Statistic (TORT) with the aim of determining the technique that has a higher power of detecting and handling outliers in terms of their P-values, true positives, false positives, False Discovery Rate (FDR) and their corresponding Receiver Operating Characteristic (ROC) curves. From the result of the analysis, it was revealed that OS was the best technique followed by COPA, t, ORT, TORT and Z respectively in terms of their P-values. The result of the False Discovery Rate (FDR) shows that OS is the best technique followed by COPA, t, ORT, TORT and Z. In terms of their ROC curves, t-Statistic and OS have the largest Area under the ROC Curve (AUC) which indicates better sensitivity and specificity and is more significant followed by COPA and ORT with the equal significant AUC while Z and TORT have the least AUC which is not significant.

Abstract 3554: Promoter hypermethylation of zinc finger proteins is the novel epigenetic biomarkers of head and neck squamous cell carcinomas.

Abstract Objective: Head and Neck Squamous Cell Carcinoma (HNSCC) is the fifth most common cancer worldwide with five-year survival rate of 50% despite use of modern treatment modalities. We have employed a comprehensive genome-wide integrated analysis of epigenetic and transcriptional alteration in primary HNSCC tissues in order to discover novel biomarkers of this disease. Experimental Design: A discovery cohort included 44 primary HNSCC and 25 normal mucosal samples was analyzed by Affymetrix GeneChip Human Exon 1.0 and Illumina Infinium Human Methylation 27 arrays. An independent validation cohort included 32 primary HNSCC and 15 normal mucosal samples to confirm differential methylation by bisulfate sequencing. A second validation cohort included salivary rinse and primary tissue from 59 HNSCC and 31 healthy patients. Detection of DNA methylation in bodily fluids and primary tissues from the later cohort was evaluated by quantitative methylation-specific PCR (qMSP). Results: We have applied novel screening approach based on Cancer Outlier Profile Analysis (COPA) and Gene Set Enrichment Analysis (GSEA) to our array data for the discovery cohort. With use of this approach we determined 81 significant differentially expressed and 37 significant differentially methylated genes. Out of total 118 candidate genes we picked 36 tumor suppressing genes with strong correlation between hypermethylation and decrease of expression in tumor samples. 26 out of 36 genes (72%) demonstrated strong differential methylation in tumor samples that was confirmed on the independent validation cohort by bisulfate sequencing. Five of these 26 genes belong to Zinc Finger Proteins (ZNF) group. We analyzed the detection of the promoter methylation of ZNF14, ZNF160 and ZNF420 in salivary rinse of HNSCC and healthy patients. DNA methylation of at least one of ZNFs promoter was detected in HNSCC patients salivary rinses with sensitivity of 22% (95% CI: 12.3%-34.7%) and specificity 100% (95% CI: 89.9%-100%). Overall frequency and concordance of detection of DNA methylation from at least one ZNFs promoter in salivary rinse with signal found in primary tissues was 35.3% and 92.3%, respectively. ZNF methylation was strongly associated with oral cavity SCC (p = 0.0049). Conclusions: Our new screening approach allowed the identification of at least 36 candidate tumor suppressor genes aberrantly methylated and transcriptionally suppressed in HNSCC. Promoter hypermethylation for three ZNF genes in salivary rinse was detected with 100% specificity to HNSCC and may be a valuable salivary rinse biomarker for HNSCC incidence and recurrence, especially among high-risk group population. Citation Format: Daria A. Gaykalova, Rajita Vatapalli, Yingying Wei, Hao Wang, Patrick Hennessey, Ryan Li, Chi Zhang, Marietta Tan, Zubair Khan, William Westra, Tanbir Khan, David Zaboli, Michael F. Ochs, Joseph A. Califano. Promoter hypermethylation of zinc finger proteins is the novel epigenetic biomarkers of head and neck squamous cell carcinomas. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 3554. doi:10.1158/1538-7445.AM2013-3554

SIBER: systematic identification of bimodally expressed genes using RNAseq data

Identification of bimodally expressed genes is an important task, as genes with bimodal expression play important roles in cell differentiation, signalling and disease progression. Several useful algorithms have been developed to identify bimodal genes from microarray data. Currently, no method can deal with data from next-generation sequencing, which is emerging as a replacement technology for microarrays. We present SIBER (systematic identification of bimodally expressed genes using RNAseq data) for effectively identifying bimodally expressed genes from next-generation RNAseq data. We evaluate several candidate methods for modelling RNAseq count data and compare their performance in identifying bimodal genes through both simulation and real data analysis. We show that the lognormal mixture model performs best in terms of power and robustness under various scenarios. We also compare our method with alternative approaches, including profile analysis using clustering and kurtosis (PACK) and cancer outlier profile analysis (COPA). Our method is robust, powerful, invariant to shifting and scaling, has no blind spots and has a sample-size-free interpretation. The R package SIBER is available at the website http://bioinformatics.mdanderson.org/main/OOMPA:Overview.

BOPA: A Bayesian hierarchical model for outlier expression detection

In many cancer studies, a gene may be expressed in some but not all of the disease samples, reflecting the complexity of the underlying disease. The traditional t-test assumes a mean shift for the tumor samples compared to normal samples and is thus not structured to capture partial differential expressions. More powerful tests specially designed for this situation can find genes with heterogeneous expressions associated with possible subtypes of the cancer. This article proposes a Bayesian model for cancer outlier profile analysis (BOPA). We build on the Gamma–Gamma model introduced in Newton et al. (2001), Kendziorski et al. (2003), and Newton et al. (2004), by using a five-component mixture model to represent various differential expression patterns. The hierarchical mixture model explicitly accounts for outlier expressions, and inferences are based on samples from posterior distributions generated from the Markov chain Monte Carlo algorithm we have developed. We present simulation and real-life dataset analyses to demonstrate the proposed methodology.

Abstract 4097: Integrative analysis of epigenetic and transcriptional alterations in head and neck squamous cell carcinomas

Abstract Objective: Head and Neck Squamous Cell Carcinoma (HNSCC) is the sixth most common cancer worldwide with a 50% patient survival rate 5 years post-diagnosis despite current therapeutic strategies. A direct comprehensive genome wide integrated analysis of epigenetic and transcriptional alteration in primary HNSCC has not been performed before. Such data would greatly enrich our knowledge about HNSCC biology and would be beneficial for the development of the individual targeted therapy. Experimental Design: 44 primary HNSCC and 25 normal mucosal samples were used in the Affymetrix GeneChip Human Exon 1.0 array and for the Illumina Infinium Human Methylation 27 array. To analyze the data we employed a novel screening approach based on Cancer Outlier Profile Analysis (COPA). With use of this approach we determined 81 significant differentially expressed and 37 significant differentially methylated genes with both oncogenic and tumor suppressing properties. Results: Out of total 118 candidate genes we picked 36 candidate tumor suppressing genes with strong correlation between hypermethylation and decrease of expression in tumor samples, such as BANK1, ADFP, MAPK4, ZNF71 and others. 20 genes out of total 36 were validated by bisulfite sequencing in a separate cohort of 32 primary HNSCC tumor and 16 normal mucosal tissues. Out of the total 20 genes, 13 demonstrated at least 40% difference in methylation, most notably for BANK1 81% of tumors were more methylated as compared to normals. 7 of these newly discovered genes out of the 36 belong to Zinc Finger Proteins (ZNF) group and they have never been reported as tumor suppressors in HNSCC before. Bisulfite sequencing and quantitative real time PCR (qRT-PCR) analysis of ZNFs revealed that they are strongly hypermethylated and underexpressed in HNSCC tumor samples. Conclusions: Our new screening approach allowed identification of 36 candidate tumor suppressor genes aberrantly methylated and transcriptionally suppressed in HNSCC. Validation of gene expression and methylation allowed the detection of ZNF genes whose downregulation may play a significant role in HNSCC tumor development. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 4097. doi:1538-7445.AM2012-4097

Abstract 4174: Identification of PRPF4 as a novel cancer promoter through AKT signaling in lung cancer

Abstract We recently constructed a large-scale gene expression database containing more than 33,000 tumors samples by applying cancer outlier profile analysis, and identified many candidate genes as therapeutic targets (Shin G et al., 2011). One of potential target candidates is PRPF4, which is over-expressed in a variety of human cancer, particularly in the breast, colon, ovary, skin and lung cancer patients. PRPF4 (PRP4 pre-mRNA processing factor 4 homolog), also called PRP4, HPRP4, Prp4p, or HPRP4P, is involved in pre-mRNA splicing and interacts with U4/U6-U5 tri-snRNP complex. Moreover, there have been a few studies on the roles of PRPF4 in diseases including cancer. In one study, PRPF4, along with MDR1 (ABCB1) and survivins, was involved in drug resistance in ovarian cancer patients (Duan Z et al., 2008), but its role in cancer is mostly unknown. In this study, we found that PRPF4 was over-expressed in breast, colon, and lung cancers, and involved in cancer progression, invasion, and migration, and interacted with AKT. Knock-down of PRPF4 by small interfering RNA in both lung and breast cancer cell lines dramatically reduced invasion, migration, and cell proliferation in vitro. We confirmed the up-regulation of PRPF4 in about 30% of lung and breast cancer specimens by RT-PCR. We also found that PRPF4 interacts with AKT and thus activates PI3K/AKT signaling pathway and increases proliferation, invasion, and migration of lung cancer cells. Collectively, we suggest PRPF4 is a novel up-regulated gene involved in various cancer progression and propose it as a novel cancer biomarker for targeted therapeutics. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 4174. doi:1538-7445.AM2012-4174

MCOPA: analysis of heterogeneous features in cancer expression data

BackgroundCancer outlier profile analysis (COPA) has proven to be an effective approach to analyzing cancer expression data, leading to the discovery of the TMPRSS2 and ETS family gene fusion events in prostate cancer. However, the original COPA algorithm did not identify down-regulated outliers, and the currently available R package implementing the method is similarly restricted to the analysis of over-expressed outliers. Here we present a modified outlier detection method, mCOPA, which contains refinements to the outlier-detection algorithm, identifies both over- and under-expressed outliers, is freely available, and can be applied to any expression dataset.ResultsWe compare our method to other feature-selection approaches, and demonstrate that mCOPA frequently selects more-informative features than do differential expression or variance-based feature selection approaches, and is able to recover observed clinical subtypes more consistently. We demonstrate the application of mCOPA to prostate cancer expression data, and explore the use of outliers in clustering, pathway analysis, and the identification of tumour suppressors. We analyse the under-expressed outliers to identify known and novel prostate cancer tumour suppressor genes, validating these against data in Oncomine and the Cancer Gene Index. We also demonstrate how a combination of outlier analysis and pathway analysis can identify molecular mechanisms disrupted in individual tumours.ConclusionsWe demonstrate that mCOPA offers advantages, compared to differential expression or variance, in selecting outlier features, and that the features so selected are better able to assign samples to clinically annotated subtypes. Further, we show that the biology explored by outlier analysis differs from that uncovered in differential expression or variance analysis. mCOPA is an important new tool for the exploration of cancer datasets and the discovery of new cancer subtypes, and can be combined with pathway and functional analysis approaches to discover mechanisms underpinning heterogeneity in cancers.