The bimodality index: a criterion for discovering and ranking bimodal signatures from cancer gene expression profiling data

Multi-omic measurement of mutually exclusive loss-of-function enriches for candidate synthetic lethal gene pairs

Background

Identification of synthetic lethal interactions in cancer cells could offer promising new therapeutic targets. Large-scale functional genomic screening presents an opportunity to test large numbers of cancer synthetic lethal hypotheses. Methods enriching for candidate synthetic lethal targets in molecularly defined cancer cell lines can steer effective design of screening efforts. Loss of one partner of a synthetic lethal gene pair creates a dependency on the other, thus synthetic lethal gene pairs should never show simultaneous loss-of-function. We have developed a computational approach to mine large multi-omic cancer data sets and identify gene pairs with mutually exclusive loss-of-function. Since loss-of-function may not always be genetic, we look for deleterious mutations, gene deletion and/or loss of mRNA expression by bimodality defined with a novel algorithm BiSEp.

Results

Applying this toolkit to both tumour cell line and patient data, we achieve statistically significant enrichment for experimentally validated tumour suppressor genes and synthetic lethal gene pairings. Notably non-reliance on genetic loss reveals a number of known synthetic lethal relationships otherwise missed, resulting in marked improvement over genetic-only predictions. We go on to establish biological rationale surrounding a number of novel candidate synthetic lethal gene pairs with demonstrated dependencies in published cancer cell line shRNA screens.

Conclusions

This work introduces a multi-omic approach to define gene loss-of-function, and enrich for candidate synthetic lethal gene pairs in cell lines testable through functional screens. In doing so, we offer an additional resource to generate new cancer drug target and combination hypotheses. Algorithms discussed are freely available in the BiSEp CRAN package at http://cran.r-project.org/web/packages/BiSEp/index.html.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-016-2375-1) contains supplementary material, which is available to authorized users.

Systems approaches to predict the functions of glycoside hydrolases during the life cycle of Aspergillus niger using developmental mutants ∆brlA and ∆flbA

BACKGROUND

The filamentous fungus Aspergillus niger encounters carbon starvation in nature as well as during industrial fermentations. In response, regulatory networks initiate and control autolysis and sporulation. Carbohydrate-active enzymes play an important role in these processes, for example by modifying cell walls during spore cell wall biogenesis or in cell wall degradation connected to autolysis.

RESULTS

In this study, we used developmental mutants (ΔflbA and ΔbrlA) which are characterized by an aconidial phenotype when grown on a plate, but also in bioreactor-controlled submerged cultivations during carbon starvation. By comparing the transcriptomes, proteomes, enzyme activities and the fungal cell wall compositions of a wild type A. niger strain and these developmental mutants during carbon starvation, a global overview of the function of carbohydrate-active enzymes is provided. Seven genes encoding carbohydrate-active enzymes, including cfcA, were expressed during starvation in all strains; they may encode enzymes involved in cell wall recycling. Genes expressed in the wild-type during starvation, but not in the developmental mutants are likely involved in conidiogenesis. Eighteen of such genes were identified, including characterized sporulation-specific chitinases and An15g02350, member of the recently identified carbohydrate-active enzyme family AA11. Eight of the eighteen genes were also expressed, independent of FlbA or BrlA, in vegetative mycelium, indicating that they also have a role during vegetative growth. The ΔflbA strain had a reduced specific growth rate, an increased chitin content of the cell wall and specific expression of genes that are induced in response to cell wall stress, indicating that integrity of the cell wall of strain ΔflbA is reduced.

CONCLUSION

The combination of the developmental mutants ΔflbA and ΔbrlA resulted in the identification of enzymes involved in cell wall recycling and sporulation-specific cell wall modification, which contributes to understanding cell wall remodeling mechanisms during development.

A peculiar mutation spectrum emerging from young peruvian patients with hepatocellular carcinoma

Hepatocellular carcinoma usually afflicts individuals in their later years following longstanding liver disease. In Peru, hepatocellular carcinoma exists in a unique clinical presentation, which affects patients around age 25 with a normal, healthy liver. In order to deepen our understanding of the molecular processes ongoing in Peruvian liver tumors, mutation spectrum analysis was carried out on hepatocellular carcinomas from 80 Peruvian patients. Sequencing analysis focused on nine genes typically altered during liver carcinogenesis, i.e. ARID2, AXIN1, BRAF, CTNNB1, NFE2L2, H/K/N-RAS, and TP53. We also assessed the transcription level of factors involved in the control of the alpha-fetoprotein expression and the Hippo signaling pathway that controls contact inhibition in metazoans. The mutation spectrum of Peruvian patients was unique with a major class of alterations represented by Insertions/Deletions. There were no changes at hepatocellular carcinoma-associated mutation hotspots in more than half of the specimens analyzed. Furthermore, our findings support the theory of a consistent collapse in the Hippo axis, as well as an expression of the stemness factor NANOG in high alpha-fetoprotein-expressing hepatocellular carcinomas. These results confirm the specificity of Peruvian hepatocellular carcinoma at the molecular genetic level. The present study emphasizes the necessity to widen cancer research to include historically neglected patients from South America, and more broadly the Global South, where cancer genetics and tumor presentation are divergent from canonical neoplasms.

Molecular and functional definition of the developing human striatum

The complexity of the human brain derives from the intricate interplay of molecular instructions during development. Here we systematically investigated gene expression changes in the prenatal human striatum and cerebral cortex during development from post-conception weeks 2 to 20. We identified tissue-specific gene coexpression networks, differentially expressed genes and a minimal set of bimodal genes, including those encoding transcription factors, that distinguished striatal from neocortical identities. Unexpected differences from mouse striatal development were discovered. We monitored 36 determinants at the protein level, revealing regional domains of expression and their refinement, during striatal development. We electrophysiologically profiled human striatal neurons differentiated in vitro and determined their refined molecular and functional properties. These results provide a resource and opportunity to gain global understanding of how transcriptional and functional processes converge to specify human striatal and neocortical neurons during development.

Bayesian nonparametric estimation of targeted agent effects on biomarker change to predict clinical outcome

The effect of a targeted agent on a cancer patient's clinical outcome putatively is mediated through the agent's effect on one or more early biological events. This is motivated by pre-clinical experiments with cells or animals that identify such events, represented by binary or quantitative biomarkers. When evaluating targeted agents in humans, central questions are whether the distribution of a targeted biomarker changes following treatment, the nature and magnitude of this change, and whether it is associated with clinical outcome. Major difficulties in estimating these effects are that a biomarker's distribution may be complex, vary substantially between patients, and have complicated relationships with clinical outcomes. We present a probabilistically coherent framework for modeling and estimation in this setting, including a hierarchical Bayesian nonparametric mixture model for biomarkers that we use to define a functional profile of pre-versus-post-treatment biomarker distribution change. The functional is similar to the receiver operating characteristic used in diagnostic testing. The hierarchical model yields clusters of individual patient biomarker profile functionals, and we use the profile as a covariate in a regression model for clinical outcome. The methodology is illustrated by analysis of a dataset from a clinical trial in prostate cancer using imatinib to target platelet-derived growth factor, with the clinical aim to improve progression-free survival time.

OPG and PgR show similar cohort specific effects as prognostic factors in ER positive breast cancer

The RANK/RANKL/OPG pathway is well known for bone destruction in skeletal metastases but has also been implicated in osteoclast-independent roles in tumorigenesis and de novo metastasis. Experimental data suggest contribution of progesterone to tumorigenesis may be mediated by RANKL. Importantly, modulation of this pathway became possible through the availability of denosumab, an artificial counterpart of OPG, but significant gaps remain in the translation of preclinical findings on the pathway. We analyzed gene expression of RANK, RANKL and OPG from 40 Affymetrix datasets encompassing 4467 primary breast cancers and focused on ER positive disease. We did not observe a significant prognostic value of RANK and RANKL mRNA expression. In contrast, OPG was associated with a better prognosis among 1941 ER positive cancers (HR 0.64, 95% CI 0.53-0.77; P < 0.0001) using a cutoff from its highly bimodal expression. We detected considerable heterogeneity regarding the prognostic value of OPG between different datasets. This heterogeneity could neither be attributed to technical reasons nor to differences in standard clinical parameters or treatments of the cohorts. Interestingly, the prognostic value of the progesterone receptor and of OPG showed similar cohort specific effects. Still both factors were no surrogates for each other but contributed independent prognostic value in multivariate analyses. Thus, both OPG and PgR are independently associated with good prognosis in ER positive breast cancer. However both markers share common cohort specific differences in contrast to proliferation markers as Ki67 which may be based on the underlying biology.

Immunohistochemistry as a valuable tool to assess CD30 expression in peripheral T-cell lymphomas: high correlation with mRNA levels

The extended use of brentuximab-vedotin was reported for CD30(+) nonanaplastic peripheral T-cell lymphomas (PTCLs) with promising efficacy. CD30 status assessment is thus a critical factor for therapeutic decision, but the reliability of immunohistochemistry (IHC) in evaluating its expression remains to be defined. This prompted us to investigate the correlation between semiquantitative CD30 protein assessment by IHC and messenger RNA (mRNA) assessment by microarrays in a cohort of 376 noncutaneous PTCLs representative of the main entities. By IHC, CD30 expression was heterogeneous across and within entities and significantly associated with large tumor cell size. In addition to 100% anaplastic large-cell lymphomas, 57% of other PTCL entities were CD30-positive at a 5% threshold. CD30 protein expression was highly correlated to mRNA levels. mRNA levels were bimodal, separating high from low CD30-expressing PTCL cases. We conclude that IHC is a valuable tool in clinical practice to assess CD30 expression in PTCLs.

The Italian primary school-size distribution and the city-size: a complex nexus

We characterize the statistical law according to which Italian primary school-size distributes. We find that the school-size can be approximated by a log-normal distribution, with a fat lower tail that collects a large number of very small schools. The upper tail of the school-size distribution decreases exponentially and the growth rates are distributed with a Laplace PDF. These distributions are similar to those observed for firms and are consistent with a Bose-Einstein preferential attachment process. The body of the distribution features a bimodal shape suggesting some source of heterogeneity in the school organization that we uncover by an in-depth analysis of the relation between schools-size and city-size. We propose a novel cluster methodology and a new spatial interaction approach among schools which outline the variety of policies implemented in Italy. Different regional policies are also discussed shedding lights on the relation between policy and geographical features.

Molecular biomarkers of residual disease after surgical debulking of high-grade serous ovarian cancer

PURPOSE

Residual disease following primary cytoreduction is associated with adverse overall survival in patients with epithelial ovarian cancer. Accurate identification of patients at high risk of residual disease has been elusive, lacking external validity and prompting many to undergo unnecessary surgical exploration. Our goal was to identify and validate molecular markers associated with high rates of residual disease.

METHODS

We interrogated two publicly available datasets from chemonaïve primary high-grade serous ovarian tumors for genes overexpressed in patients with residual disease and significant at a 10% false discovery rate (FDR) in both datasets. We selected genes with wide dynamic range for validation in an independent cohort using quantitative RT-PCR to assay gene expression, followed by blinded prediction of a patient subset at high risk for residual disease. Predictive success was evaluated using a one-sided Fisher exact test.

RESULTS

Forty-seven probe sets met the 10% FDR criterion in both datasets. These included FABP4 and ADH1B, which tracked tightly, showed dynamic ranges >16-fold and had high expression levels associated with increased incidence of residual disease. In the validation cohort (n = 139), FABP4 and ADH1B were again highly correlated. Using the top quartile of FABP4 PCR values as a prespecified threshold, we found 30 of 35 cases of residual disease in the predicted high-risk group (positive predictive value = 86%) and 54 of 104 among the remaining patients (P = 0.0002; OR, 5.5).

CONCLUSION

High FABP4 and ADH1B expression is associated with significantly higher risk of residual disease in high-grade serous ovarian cancer. Patients with high tumoral levels of these genes may be candidates for neoadjuvant chemotherapy.

Gene co-expression modules as clinically relevant hallmarks of breast cancer diversity

Co-expression modules are groups of genes with highly correlated expression patterns. In cancer, differences in module activity potentially represent the heterogeneity of phenotypes important in carcinogenesis, progression, or treatment response. To find gene expression modules active in breast cancer subpopulations, we assembled 72 breast cancer-related gene expression datasets containing ∼5,700 samples altogether. Per dataset, we identified genes with bimodal expression and used mixture-model clustering to ultimately define 11 modules of genes that are consistently co-regulated across multiple datasets. Functionally, these modules reflected estrogen signaling, development/differentiation, immune signaling, histone modification, ERBB2 signaling, the extracellular matrix (ECM) and stroma, and cell proliferation. The Tcell/Bcell immune modules appeared tumor-extrinsic, with coherent expression in tumors but not cell lines; whereas most other modules, interferon and ECM included, appeared intrinsic. Only four of the eleven modules were represented in the PAM50 intrinsic subtype classifier and other well-established prognostic signatures; although the immune modules were highly correlated to previously published immune signatures. As expected, the proliferation module was highly associated with decreased recurrence-free survival (RFS). Interestingly, the immune modules appeared associated with RFS even after adjustment for receptor subtype and proliferation; and in a multivariate analysis, the combination of Tcell/Bcell immune module down-regulation and proliferation module upregulation strongly associated with decreased RFS. Immune modules are unusual in that their upregulation is associated with a good prognosis without chemotherapy and a good response to chemotherapy, suggesting the paradox of high immune patients who respond to chemotherapy but would do well without it. Other findings concern the ECM/stromal modules, which despite common themes were associated with different sites of metastasis, possibly relating to the “seed and soil” hypothesis of cancer dissemination. Overall, co-expression modules provide a high-level functional view of breast cancer that complements the “cancer hallmarks” and may form the basis for improved predictors and treatments.

Gene expression in teratogenic exposures: a new approach to understanding individual risk

The phenomenon of partial or incomplete penetrance is common to many paradigms of exposure to teratogens, where only some of the exposed individuals exhibit developmental defects. We here argue that the most widely used experimental approaches in reproductive toxicology do not take partial penetrance into account, and are thus likely to miss differences between affected and unaffected individuals that contribute to susceptibility for teratogenesis. We propose that focus on the variation between exposed individuals could help to discover factors that may play a causative role for abnormal developmental processes that occur with incomplete penetrance.

Analysis considerations for utilizing RNA-Seq to characterize the brain transcriptome

RNA-Seq allows one to examine only gene expression as well as expression of noncoding RNAs, alternative splicing, and allele-specific expression. With this increased sensitivity and dynamic range, there are computational and statistical considerations that need to be contemplated, which are highly dependent on the biological question being asked. We highlight these to provide an overview of their importance and the impact they can have on downstream interpretation of the brain transcriptome.

DNA repair gene patterns as prognostic and predictive factors in molecular breast cancer subtypes

DNA repair pathways can enable tumor cells to survive DNA damage induced by chemotherapy and thus provide prognostic and/or predictive value. We evaluated Affymetrix gene expression profiles for 145 DNA repair genes in untreated breast cancer (BC) patients (n = 684) and BC patients treated with regimens containing neoadjuvant taxane/anthracycline (n = 294) or anthracycline (n = 210). We independently assessed estrogen receptor (ER)-positive/HER2-negative, HER2-positive, and ER-negative/HER2-negative subgroups for differential expression, bimodal distribution, and the prognostic and predictive value of DNA repair gene expression. Twenty-two genes were consistently overexpressed in ER-negative tumors, and five genes were overexpressed in ER-positive tumors, but no differences in expression were associated with HER2 status. In ER-positive/HER2-negative tumors, the expression of nine genes (BUB1, FANCI, MNAT1, PARP2, PCNA, POLQ, RPA3, TOP2A, and UBE2V2) was associated with poor prognosis, and the expression of one gene (ATM) was associated with good prognosis. Furthermore, the prognostic value of specific genes did not correlate with proliferation. A few genes were associated with chemotherapy response in BC subtypes and treatment-specific manner. In ER-negative/HER2-negative tumors, the MSH2, MSH6, and FAN1 (previously MTMR15) genes were associated with pathological complete response and residual invasive cancer in taxane/anthracycline-treated patients. Conversely, PMS2 expression was associated with residual invasive cancer in treatments using anthracycline as a single agent. In HER2-positive tumors, TOP2A was associated with patient response to anthracyclines but not to taxane/anthracycline regimens. In genes expressed in a bimodal fashion, RECQL4 was significantly associated with clinical outcome. In vitro studies showed that defects in RECQL4 impair homologous recombination, sensitizing BC cells to DNA-damaging agents.

An atypical age-specific pattern of hepatocellular carcinoma in Peru: a threat for Andean populations

BACKGROUND

In South America, the highest incidence of primary liver cancer is observed in Peru. However, national estimations on hepatocellular carcinoma incidence and mortality are approximated using aggregated data from surrounding countries. Thus, there is a lack of tangible information from Peru that impairs an accurate description of the local incidence, presentation, and outcomes of hepatocellular carcinoma. The present study attempts to fill this gap and assesses the clinical epidemiology of hepatocellular carcinoma in this country.

METHODS

A retrospective cohort study was conducted by analysing the medical charts of 1,541 patients with hepatocellular carcinoma admitted between 1997 and 2010 at the Peruvian national institute for cancer. The medical records including liver function, serologic status, and tumor pathology and stage were monitored. Statistical analyses were performed in order to characterize tumor presentation according to demographic features, risk factors, and regional origin.

RESULTS

Surprisingly, the age distribution of the patient population displayed bimodality corresponding to two distinct age-based subpopulations. While an older group was in keeping with the age range observed for hepatocellular carcinoma around the world, a younger population displayed an abnormally juvenile mean age of 25.5 years old. In addition, each subpopulation displayed age-specific pathophysiological and clinical characteristics.

CONCLUSIONS

The analysis suggests two different age-specific natural histories of hepatocellular carcinoma in the Peruvian patient population. This otherwise unusual tumor process that is ongoing in younger patients leads to the hypothesis that there may be a Peru-endemic risk factor driving hepatocarcinogenesis in the local population.

Lactate dehydrogenase B: a metabolic marker of response to neoadjuvant chemotherapy in breast cancer

PURPOSE

Although breast cancers are known to be molecularly heterogeneous, their metabolic phenotype is less well-understood and may predict response to chemotherapy. This study aimed to evaluate metabolic genes as individual predictive biomarkers in breast cancer.

EXPERIMENTAL DESIGN

mRNA microarray data from breast cancer cell lines were used to identify bimodal genes-those with highest potential for robust high/low classification in clinical assays. Metabolic function was evaluated in vitro for the highest scoring metabolic gene, lactate dehydrogenase B (LDHB). Its expression was associated with neoadjuvant chemotherapy response and relapse within clinical and PAM50-derived subtypes.

RESULTS

LDHB was highly expressed in cell lines with glycolytic, basal-like phenotypes. Stable knockdown of LDHB in cell lines reduced glycolytic dependence, linking LDHB expression directly to metabolic function. Using patient datasets, LDHB was highly expressed in basal-like cancers and could predict basal-like subtype within clinical groups [OR = 21 for hormone receptor (HR)-positive/HER2-negative; OR = 10 for triple-negative]. Furthermore, high LDHB predicted pathologic complete response (pCR) to neoadjuvant chemotherapy for both HR-positive/HER2-negative (OR = 4.1, P < 0.001) and triple-negative (OR = 3.0, P = 0.003) cancers. For triple-negative tumors without pCR, high LDHB posttreatment also identified proliferative tumors with increased risk of recurrence (HR = 2.2, P = 0.006).

CONCLUSIONS

Expression of LDHB predicted response to neoadjuvant chemotherapy within clinical subtypes independently of standard prognostic markers and PAM50 subtyping. These observations support prospective clinical evaluation of LDHB as a predictive marker of response for patients with breast cancer receiving neoadjuvant chemotherapy.

Biasogram: visualization of confounding technical bias in gene expression data

Gene expression profiles of clinical cohorts can be used to identify genes that are correlated with a clinical variable of interest such as patient outcome or response to a particular drug. However, expression measurements are susceptible to technical bias caused by variation in extraneous factors such as RNA quality and array hybridization conditions. If such technical bias is correlated with the clinical variable of interest, the likelihood of identifying false positive genes is increased. Here we describe a method to visualize an expression matrix as a projection of all genes onto a plane defined by a clinical variable and a technical nuisance variable. The resulting plot indicates the extent to which each gene is correlated with the clinical variable or the technical variable. We demonstrate this method by applying it to three clinical trial microarray data sets, one of which identified genes that may have been driven by a confounding technical variable. This approach can be used as a quality control step to identify data sets that are likely to yield false positive results.

DNA methylation profiling of medulloblastoma allows robust subclassification and improved outcome prediction using formalin-fixed biopsies

Molecular subclassification is rapidly informing the clinical management of medulloblastoma. However, the disease remains associated with poor outcomes and therapy-associated late effects, and the majority of patients are not characterized by a validated prognostic biomarker. Here, we investigated the potential of epigenetic DNA methylation for disease subclassification, particularly in formalin-fixed biopsies, and to identify biomarkers for improved therapeutic individualization. Tumor DNA methylation profiles were assessed, alongside molecular and clinical disease features, in 230 patients primarily from the SIOP-UKCCSG PNET3 clinical trial. We demonstrate by cross-validation in frozen training and formalin-fixed test sets that medulloblastoma comprises four robust DNA methylation subgroups (termed WNT, SHH, G3 and G4), highly related to their transcriptomic counterparts, and which display distinct molecular, clinical and pathological disease characteristics. WNT patients displayed an expected favorable prognosis, while outcomes for SHH, G3 and G4 were equivalent in our cohort. MXI1 and IL8 methylation were identified as novel independent high-risk biomarkers in cross-validated survival models of non-WNT patients, and were validated using non-array methods. Incorporation of MXI1 and IL8 into current survival models significantly improved the assignment of disease risk; 46 % of patients could be classified as 'favorable risk' (>90 % survival) compared to 13 % using current models, while the high-risk group was reduced from 30 to 16 %. DNA methylation profiling enables the robust subclassification of four disease subgroups in frozen and routinely collected/archival formalin-fixed biopsy material, and the incorporation of DNA methylation biomarkers can significantly improve disease-risk stratification. These findings have important implications for future risk-adapted clinical disease management.

SIBER: systematic identification of bimodally expressed genes using RNAseq data

MOTIVATION

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.

RESULTS

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.

AVAILABILITY

The R package SIBER is available at the website http://bioinformatics.mdanderson.org/main/OOMPA:Overview.

An epithelial-mesenchymal transition gene signature predicts resistance to EGFR and PI3K inhibitors and identifies Axl as a therapeutic target for overcoming EGFR inhibitor resistance

PURPOSE

Epithelial-mesenchymal transition (EMT) has been associated with metastatic spread and EGF receptor (EGFR) inhibitor resistance. We developed and validated a robust 76-gene EMT signature using gene expression profiles from four platforms using non-small cell lung carcinoma (NSCLC) cell lines and patients treated in the Biomarker-Integrated Approaches of Targeted Therapy for Lung Cancer Elimination (BATTLE) study.

EXPERIMENTAL DESIGN

We conducted an integrated gene expression, proteomic, and drug response analysis using cell lines and tumors from patients with NSCLC. A 76-gene EMT signature was developed and validated using gene expression profiles from four microarray platforms of NSCLC cell lines and patients treated in the BATTLE study, and potential therapeutic targets associated with EMT were identified.

RESULTS

Compared with epithelial cells, mesenchymal cells showed significantly greater resistance to EGFR and PI3K/Akt pathway inhibitors, independent of EGFR mutation status, but more sensitivity to certain chemotherapies. Mesenchymal cells also expressed increased levels of the receptor tyrosine kinase Axl and showed a trend toward greater sensitivity to the Axl inhibitor SGI-7079, whereas the combination of SGI-7079 with erlotinib reversed erlotinib resistance in mesenchymal lines expressing Axl and in a xenograft model of mesenchymal NSCLC. In patients with NSCLC, the EMT signature predicted 8-week disease control in patients receiving erlotinib but not other therapies.

CONCLUSION

We have developed a robust EMT signature that predicts resistance to EGFR and PI3K/Akt inhibitors, highlights different patterns of drug responsiveness for epithelial and mesenchymal cells, and identifies Axl as a potential therapeutic target for overcoming EGFR inhibitor resistance associated with the mesenchymal phenotype.

Identification of sepsis subtypes in critically ill adults using gene expression profiling

Introduction

Sepsis is a syndromic illness that has traditionally been defined by a set of broad, highly sensitive clinical parameters. As a result, numerous distinct pathophysiologic states may meet diagnostic criteria for sepsis, leading to syndrome heterogeneity. The existence of biologically distinct sepsis subtypes may in part explain the lack of actionable evidence from clinical trials of sepsis therapies. We used microarray-based gene expression data from adult patients with sepsis in order to identify molecularly distinct sepsis subtypes.

Methods

We used partitioning around medoids (PAM) and hierarchical clustering of gene expression profiles from neutrophils taken from a cohort of septic patients in order to identify distinct subtypes. Using the medoids learned from this cohort, we then clustered a second independent cohort of septic patients, and used the resulting class labels to evaluate differences in clinical parameters, as well as the expression of relevant pharmacogenes.

Results

We identified two sepsis subtypes based on gene expression patterns. Subtype 1 was characterized by increased expression of genes involved in inflammatory and Toll receptor mediated signaling pathways, as well as a higher prevalence of severe sepsis. There were differences between subtypes in the expression of pharmacogenes related to hydrocortisone, vasopressin, norepinephrine, and drotrecogin alpha.

Conclusions

Sepsis subtypes can be identified based on different gene expression patterns. These patterns may generate hypotheses about the underlying pathophysiology of sepsis and suggest new ways of classifying septic patients both in clinical practice, and in the design of clinical trials.

Inter-donor variation in cell subset specific immune signaling responses in healthy individuals

Single cell network profiling (SCNP) is a multi-parameter flow cytometry based approach that allows for the simultaneous interrogation of intracellular signaling pathways in multiple cell subpopulations within heterogeneous tissues, without the need for individual cell subset isolation. Thus, the technology is extremely well-suited for characterizing the multitude of interconnected signaling pathways and immune cell subpopulations that regulate the function of the immune system. Recently, SCNP was applied to generate a functional map of the healthy human immune cell signaling network by profiling immune signaling pathways downstream of 12 immunomodulators in 7 distinct immune cell subsets within peripheral blood mononuclear cells (PBMCs) from 60 healthy donors. In the study reported here, the degree of inter-donor variation in the magnitude of the immune signaling responses was analyzed. The highest inter-donor differences in immune signaling pathway activity occurred following perturbation of the immune signaling network, rather than in basal signaling. When examining the full panel of immune signaling responses, as one may expect, the overall degree of inter-donor variation was positively correlated (r = 0.727) with the magnitude of node response (i.e. a larger median signaling response was associated with greater inter-donor variation). However, when examining the degree of heterogeneity across cell subpopulations for individual signaling nodes, cell subset specificity in the degree of inter-donor variation was observed for several nodes. For such nodes, relatively weak correlations between inter-donor variation and the magnitude of the response were observed. Further, within the phenotypically distinct subpopulations, a fraction of the immune signaling responses had bimodal response profiles in which (a) only a portion of the cells had elevated phospho-protein levels following modulation and (b) the proportion of responsive cells varied by donor. These data exemplify the application of SCNP to provide a detailed characterization of inter-donor variation in immune signaling pathway activation in a healthy donor cohort. This dataset provides a basis for identifying cell subpopulation specific immune signaling abnormalities in cancer and immune-mediated diseases. Building upon these data in future studies may help inform on disease etiology, maintenance and therapeutic selection.

Melanoma antigen family A identified by the bimodality index defines a subset of triple negative breast cancers as candidates for immune response augmentation

BACKGROUND

Molecular markers displaying bimodal expression distribution can reveal distinct disease subsets and may serve as prognostic or predictive markers or represent therapeutic targets. Oestrogen (ER) and human epidermal growth factor receptor 2 (HER2) receptors are strongly bimodally expressed genes in breast cancer.

MATERIAL AND METHODS

We applied a novel method to identify bimodally expressed genes in 394 triple negative breast cancers (TNBC). We identified 133 bimodally expressed probe sets (128 unique genes), 69 of these correlated to previously reported metagenes that define molecular subtypes within TNBC including basal-like, molecular-apocrine, claudin-low and immune cell rich subgroups but 64 probe sets showed no correlation with these features.

RESULTS

The single most prominent functional group among these uncorrelated genes was the X chromosome derived Cancer/Testis Antigens (CT-X) including melanoma antigen family A (MAGE-A) and Cancer/Testis Antigens (CTAG). High expression of CT-X genes correlated with worse survival in multivariate analysis (HR 2.02, 95% CI 1.27-3.20; P=0.003). The only other significant variable was lymph node status. The poor prognosis of patients with high MAGE-A expression was ameliorated by the concomitant high expression of immune cell metagenes (HR 1.87, 95% CI 0.96-3.64; P=0.060), whereas the same immune metagene had lesser prognostic value in TNBC with low MAGE-A expression.

CONCLUSIONS

MAGE-A antigen defines a very aggressive subgroup of TNBC; particularly in the absence of immune infiltration in the tumour microenvironment. These observations suggest a therapeutic hypothesis; TNBC with MAGE-A expression may benefit the most from further augmentation of the immune response. Novel immune stimulatory drugs such as (anti-cytotoxic T-lymphocyte antigen-4 CTLA-4) directed therapies provide a realistic opportunity to directly test this hypothesis in the clinic.

A comparison of methods for data-driven cancer outlier discovery, and an application scheme to semisupervised predictive biomarker discovery

A core component in translational cancer research is biomarker discovery using gene expression profiling for clinical tumors. This is often based on cell line experiments; one population is sampled for inference in another. We disclose a semisupervised workflow focusing on binary (switch-like, bimodal) informative genes that are likely cancer relevant, to mitigate this non-statistical problem. Outlier detection is a key enabling technology of the workflow, and aids in identifying the focus genes.

We compare outlier detection techniques MOST, LSOSS, COPA, ORT, OS, and t-test, using a publicly available NSCLC dataset. Removing genes with Gaussian distribution is computationally efficient and matches MOST particularly well, while also COPA and OS pick prognostically relevant genes in their top ranks. Also our stability assessment is in favour of both MOST and COPA; the latter does not pair well with prefiltering for non-Gaussianity, but can handle data sets lacking non-cancer cases.

We provide R code for replicating our approach or extending it.

BiC: a web server for calculating bimodality of coexpression between gene and protein networks

Bimodal patterns of expression have recently been shown to be useful not only in prioritizing genes that distinguish phenotypes, but also in prioritizing network models that correlate with proteomic evidence. In particular, subgroups of strongly coexpressed gene pairs result in an increased variance of the correlation distribution. This variance, a measure of association between sets of genes (or proteins), can be summarized as the bimodality of coexpression (BiC). We developed an online tool to calculate the BiC for user-defined gene lists and associated mRNA expression data. BiC is a comprehensive application that provides researchers with the ability to analyze both publicly available and user-collected array data.

AVAILABILITY

The freely available web service and the documentation can be accessed at http://gurkan.case.edu/software.

CONTACT

gurkan@case.edu.

Bimodal distribution of RNA expression levels in human skeletal muscle tissue

Background

Many human diseases and phenotypes are related to RNA expression, levels of which are influenced by a wide spectrum of genetic and exposure-related factors. In a large genome-wide study of muscle tissue expression, we found that some genes exhibited a bimodal distribution of RNA expression, in contrast to what is usually assumed in studies of a single healthy tissue. As bimodality has classically been considered a hallmark of genetic control, we assessed the genome-wide prevalence, cause, and association of this phenomenon with diabetes-related phenotypes in skeletal muscle tissue from 225 healthy Pima Indians using exon array expression chips.

Results

Two independent batches of microarrays were used for bimodal assessment and comparison. Of the 17,881 genes analyzed, eight (GSTM1, HLA-DRB1, ERAP2, HLA-DRB5, MAOA, ACTN3, NR4A2, and THNSL2) were found to have bimodal expression replicated in the separate batch groups, while 24 other genes had evidence of bimodality in only one group. Some bimodally expressed genes had modest associations with pre-diabetic phenotypes, of note ACTN3 with insulin resistance. Most of the other bimodal genes have been reported to be involved with various other diseases and characteristics. Association of expression with cis genetic variation in a subset of 149 individuals found all but one of the confirmed bimodal genes and nearly half of all potential ones to be highly significant expression quantitative trait loci (eQTL). The rare prevalence of these bimodally expressed genes found after controlling for batch effects was much lower than the prevalence reported in other studies. Additional validation in data from separate muscle expression studies confirmed the low prevalence of bimodality we observed.

Conclusions

We conclude that the prevalence of bimodal gene expression is quite rare in healthy muscle tissue (<0.2%), and is much lower than limited reports from other studies. The major cause of these clearly bimodal expression patterns in homogeneous tissue appears to be cis-polymorphisms, indicating that such bimodal genes are, for the most part, eQTL. The high frequency of disease associations reported with these genes gives hope that this unique feature may identify or actually be an underlying factor responsible for disease development.

MMP1 bimodal expression and differential response to inflammatory mediators is linked to promoter polymorphisms

Background

Identifying the functional importance of the millions of single nucleotide polymorphisms (SNPs) in the human genome is a difficult challenge. Therefore, a reverse strategy, which identifies functionally important SNPs by virtue of the bimodal abundance across the human population of the SNP-related mRNAs will be useful. Those mRNA transcripts that are expressed at two distinct abundances in proportion to SNP allele frequency may warrant further study. Matrix metalloproteinase 1 (MMP1) is important in both normal development and in numerous pathologies. Although much research has been conducted to investigate the expression of MMP1 in many different cell types and conditions, the regulation of its expression is still not fully understood.

Results

In this study, we used a novel but straightforward method based on agglomerative hierarchical clustering to identify bimodally expressed transcripts in human umbilical vein endothelial cell (HUVEC) microarray data from 15 individuals. We found that MMP1 mRNA abundance was bimodally distributed in un-treated HUVECs and showed a bimodal response to inflammatory mediator treatment. RT-PCR and MMP1 activity assays confirmed the bimodal regulation and DNA sequencing of 69 individuals identified an MMP1 gene promoter polymorphism that segregated precisely with the MMP1 bimodal expression. Chromatin immunoprecipation (ChIP) experiments indicated that the transcription factors (TFs) ETS1, ETS2 and GATA3, bind to the MMP1 promoter in the region of this polymorphism and may contribute to the bimodal expression.

Conclusions

We describe a simple method to identify putative bimodally expressed RNAs from transcriptome data that is effective yet easy for non-statisticans to understand and use. This method identified bimodal endothelial cell expression of MMP1, which appears to be biologically significant with implications for inflammatory disease. (271 Words)

Minimising immunohistochemical false negative ER classification using a complementary 23 gene expression signature of ER status

Background

Expression of the oestrogen receptor (ER) in breast cancer predicts benefit from endocrine therapy. Minimising the frequency of false negative ER status classification is essential to identify all patients with ER positive breast cancers who should be offered endocrine therapies in order to improve clinical outcome. In routine oncological practice ER status is determined by semi-quantitative methods such as immunohistochemistry (IHC) or other immunoassays in which the ER expression level is compared to an empirical threshold[1], [2]. The clinical relevance of gene expression-based ER subtypes as compared to IHC-based determination has not been systematically evaluated. Here we attempt to reduce the frequency of false negative ER status classification using two gene expression approaches and compare these methods to IHC based ER status in terms of predictive and prognostic concordance with clinical outcome.

Methodology/Principal Findings

Firstly, ER status was discriminated by fitting the bimodal expression of ESR1 to a mixed Gaussian model. The discriminative power of ESR1 suggested bimodal expression as an efficient way to stratify breast cancer; therefore we identified a set of genes whose expression was both strongly bimodal, mimicking ESR expression status, and highly expressed in breast epithelial cell lines, to derive a 23-gene ER expression signature-based classifier. We assessed our classifiers in seven published breast cancer cohorts by comparing the gene expression-based ER status to IHC-based ER status as a predictor of clinical outcome in both untreated and tamoxifen treated cohorts. In untreated breast cancer cohorts, the 23 gene signature-based ER status provided significantly improved prognostic power compared to IHC-based ER status (P = 0.006). In tamoxifen-treated cohorts, the 23 gene ER expression signature predicted clinical outcome (HR = 2.20, P = 0.00035). These complementary ER signature-based strategies estimated that between 15.1% and 21.8% patients of IHC-based negative ER status would be classified with ER positive breast cancer.

Conclusion/Significance

Expression-based ER status classification may complement IHC to minimise false negative ER status classification and optimise patient stratification for endocrine therapies.

Recurrent expression signatures of cytokines and chemokines are present and are independently prognostic in acute myelogenous leukemia and myelodysplasia

The role of circulating cytokines and chemokines (C&Ckine) in activating signal transduction in leukemic cells is incompletely defined. We hypothesized that comprehensive profiling of C&Ckine expression in leukemia would provide greater insight compared with individual analyses. We used multiplex array technology to simultaneously measure the level of 27 C&Ckines in serum from 176 acute myelogenous leukemia (AML) and 114 myelodysplastic syndrome (MDS) patients and 19 normal controls. C&Ckine levels in AML and MDS differed significantly from normal controls (5 higher, 13 lower) but were similar to each other for 24 of 27 analytes, with interleukin-8 and interleukin-13 higher in AML and vascular endothelial growth factor A higher in MDS. Levels did not correlate with age, gender, infection, or blood counts; however, 3 correlated with specific cytognetic abnormalities in AML. Individually, few cytokines had any correlation with response or survival. In newly diagnosed AML, 8 C&Ckine signatures, distinct from the normal control signature, were observed. These signatures had prognostic impact, affecting remission, primary resistance, relapse rates, and overall survival, individually (P = .003) and in multivariable analysis (P = .004). These patterns suggest specific therapeutic interventions to investigate in subsets of AML patients. In conclusion, C&Ckine expression in AML and MDS differs from normal, is similar with one another, and forms recurrent patterns of expression with prognostic relevance.

Rough set soft computing cancer classification and network: one stone, two birds

Gene expression profiling provides tremendous information to help unravel the complexity of cancer. The selection of the most informative genes from huge noise for cancer classification has taken centre stage, along with predicting the function of such identified genes and the construction of direct gene regulatory networks at different system levels with a tuneable parameter. A new study by Wang and Gotoh described a novel Variable Precision Rough Sets-rooted robust soft computing method to successfully address these problems and has yielded some new insights. The significance of this progress and its perspectives will be discussed in this article.

Comparison of scores for bimodality of gene expression distributions and genome-wide evaluation of the prognostic relevance of high-scoring genes

Background

A major goal of the analysis of high-dimensional RNA expression data from tumor tissue is to identify prognostic signatures for discriminating patient subgroups. For this purpose genome-wide identification of bimodally expressed genes from gene array data is relevant because distinguishability of high and low expression groups is easier compared to genes with unimodal expression distributions.

Recently, several methods for the identification of genes with bimodal distributions have been introduced. A straightforward approach is to cluster the expression values and score the distance between the two distributions. Other scores directly measure properties of the distribution. The kurtosis, e.g., measures divergence from a normal distribution. An alternative is the outlier-sum statistic that identifies genes with extremely high or low expression values in a subset of the samples.

Results

We compare and discuss scores for bimodality for expression data. For the genome-wide identification of bimodal genes we apply all scores to expression data from 194 patients with node-negative breast cancer. Further, we present the first comprehensive genome-wide evaluation of the prognostic relevance of bimodal genes. We first rank genes according to bimodality scores and define two patient subgroups based on expression values. Then we assess the prognostic significance of the top ranking bimodal genes by comparing the survival functions of the two patient subgroups. We also evaluate the global association between the bimodal shape of expression distributions and survival times with an enrichment type analysis.

Various cluster-based methods lead to a significant overrepresentation of prognostic genes. A striking result is obtained with the outlier-sum statistic (p < 10^-12). Many genes with heavy tails generate subgroups of patients with different prognosis.

Conclusions

Genes with high bimodality scores are promising candidates for defining prognostic patient subgroups from expression data. We discuss advantages and disadvantages of the different scores for prognostic purposes. The outlier-sum statistic may be particularly valuable for the identification of genes to be included in prognostic signatures. Among the genes identified as bimodal in the breast cancer data set several have not yet previously been recognized to be prognostic and bimodally expressed in breast cancer.

Bimodal gene expression and biomarker discovery

With insights gained through molecular profiling, cancer is recognized as a heterogeneous disease with distinct subtypes and outcomes that can be predicted by a limited number of biomarkers. Statistical methods such as supervised classification and machine learning identify distinguishing features associated with disease subtype but are not necessarily clear or interpretable on a biological level. Genes with bimodal transcript expression, however, may serve as excellent candidates for disease biomarkers with each mode of expression readily interpretable as a biological state. The recent article by Wang et al, entitled "The Bimodality Index: A Criterion for Discovering and Ranking Bimodal Signatures from Cancer Gene Expression Profiling Data," provides a bimodality index for identifying and scoring transcript expression profiles as biomarker candidates with the benefit of having a direct relation to power and sample size. This represents an important step in candidate biomarker discovery that may help streamline the pipeline through validation and clinical application.