Significant Gene Sets Research Articles

Enrichment of rare genetic variants in astrocyte gene enriched co-expression modules altered in postmortem brain samples of schizophrenia

The transcriptome profiles of the cingulate gyrus region from the postmortem brain tissues of a set of well-characterized patients with schizophrenia (SCZ) and matched controls were investigated using an integrated approach that analyzed both the alterations in transcription expression pattern and rare genetic variants in expressed genes. We demonstrated increased expression of astrocyte-related genes using spatiotemporal co-expression modules that have previously been established for developing human brain, and showed these results are independent of medication dosage. The relationship between genetic variants and expression pattern in the context of neurodevelopment was further investigated, and we identified an enrichment of rare genetic variants in a set of signature genes that were specific to astrocytes and up-regulated in the patients with SCZ. Our result suggested the involvement of astrocyte malfunction in SCZ pathophysiology. In addition, our approach indicated a novel strategy of narrowing down genetic variants that might contribute to the pathophysiology in the patients with SCZ to a subset of genes that are highly expressed in an affected brain region.

Gastric Cancer Cell Lines Have Different MYC-Regulated Expression Patterns but Share a Common Core of Altered Genes.

MYC is an oncogene responsible for excessive cell growth in cancer, enabling transcriptional activation of genes involved in cell cycle regulation, metabolism, and apoptosis, and is usually overexpressed in gastric cancer (GC). By using siRNA and Next-Generation Sequencing (NGS), we identified MYC-regulated differentially expressed Genes (DEGs) in three Brazilian gastric cancer cell lines representing the histological subtypes of GC (diffuse, intestinal, and metastasis). The DEGs were picked using Sailfish software, followed by Gene Set Enrichment Analysis (GSEA) and Kyoto Encyclopedia of Gene and Genome (KEGG) pathway analysis using KEGG. We found 11 significantly enriched gene sets by using enrichment score (ES), False Discovery Rate (FDR), and nominal P-values. We identified a total of 5.471 DEGs with correlation over (80%). In diffuse-type and in metastatic GC cell lines, MYC-silencing caused DEGs downregulation, while the intestinal-type GC cells presented overall DEGs upregulation after MYC siRNA depletion. We were able to detect 11 significant gene sets when comparing our samples to the hallmark collection of gene expression, enriched mostly for the following hallmarks: proliferation, pathway, signaling, metabolic, and DNA damage response. When we analyzed our DEGs considering KEGG metabolic pathways, we found 12 common branches covering a wide range of biological functions, and three of them were common to all three cell lines: ubiquitin-mediated proteolysis, ribosomes, and system and epithelial cell signaling in Helicobacter pylori infection. The GC cell lines used in this study share 14 MYC-regulated genes, but their gene expression profile is different for each histological subtype of GC. Our results present a computational analysis of MYC-related signatures in GC, and we present evidence that GC cell lines representing distinct histological subtypes of this disease have different MYC-regulated expression profiles but share a common core of altered genes. This is an important step towards the understanding of MYC's role in gastric carcinogenesis and an indication of probable new drug targets in stomach cancer.

Matrix mineralization controls gene expression in osteoblastic cells

Osteoblasts are adherent cells, and under physiological conditions they attach to both mineralized and non-mineralized osseous surfaces. However, how exactly osteoblasts respond to these different osseous surfaces is largely unknown. Our hypothesis was that the state of matrix mineralization provides a functional signal to osteoblasts. To assess the osteoblast response to mineralized compared to demineralized osseous surfaces, we developed and validated a novel tissue surface model. We demonstrated that with the exception of the absence of mineral, the mineralized and demineralized surfaces were similar in molecular composition as determined, for example, by collagen content and maturity. Subsequently, we used the human osteoblastic cell line MG63 in combination with genome-wide gene set enrichment analysis (GSEA) to record and compare the gene expression signatures on mineralized and demineralized surfaces. Assessment of the 5 most significant gene sets showed on mineralized surfaces an enrichment exclusively of genes sets linked to protein synthesis, while on the demineralized surfaces 3 of the 5 enriched gene sets were associated with the matrix. Focusing on these three gene sets, we observed not only the expected structural components of the bone matrix, but also gene products, such as HMCN1 or NID2, that are likely to act as temporal migration guides. Together, these findings suggest that in osteoblasts mineralized and demineralized osseous surfaces favor intracellular protein production and matrix formation, respectively. Further, they demonstrate that the mineralization state of bone independently controls gene expression in osteoblastic cells.

Abstract B045: Gene expression pathways associated with vigorous physical activity in prostate cancer

Abstract Background: Epidemiologic studies have consistently shown that men engaged in high levels of physical activity have lower risks of advanced and fatal prostate cancer. The mechanisms underlying this association are not well understood but may include changes in hormonal and anti-inflammatory pathways. Our objective was to investigate the link between prediagnostic physical activity and gene expression alterations in prostate tumor and adjacent normal tissue. Methods: We performed whole-gene expression profiling on tumor (N=248) and adjacent normal tissue (N=117) among a subset of men diagnosed with indolent and lethal prostate cancer between 1986 and 2005 in the prospective Health Professionals Follow-up Study. Men were followed prospectively for prostate cancer outcomes through 2011. Physical activity was ascertained by self-report of the average number of hours per week spent engaged in various forms of recreational physical activity at baseline and every two years thereafter. We examined cumulative average physical activity of vigorous intensity in units of metabolic equivalent of task (MET)-hours/week updated from baseline to the time prostate cancer diagnosis. Gene set enrichment analysis was performed among KEGG and hallmark gene sets to identify pathways with differential expression by vigorous activity. Results: We identified 33 KEGG and four Hallmark gene sets enriched in the adjacent normal tissue comparing men in the highest compared to the lowest quintile of vigorous activity (false-discovery rate < 0.25). Among the significant gene sets identified were cancer-related pathways as well as vascular smooth muscle contraction, angiogenesis, and insulin-signaling pathways. No gene sets were significantly enriched in the tumor tissue. Conclusions: These findings suggest that physical activity may influence prostate cancer development through epigenetic alterations in the prostate tissue. Future studies are needed to confirm these findings and further investigate potential mechanisms linking physical activity to risk of lethal prostate cancer. Citation Format: Claire H. Pernar, Giovanni Parmigiani, Edward L. Giovannucci, Eric B. Rimm, Stephen Finn, Michelangelo Fiorentino, Ericka M. Ebot, Lorelei A. Mucci. Gene expression pathways associated with vigorous physical activity in prostate cancer [abstract]. In: Proceedings of the AACR Special Conference: Prostate Cancer: Advances in Basic, Translational, and Clinical Research; 2017 Dec 2-5; Orlando, Florida. Philadelphia (PA): AACR; Cancer Res 2018;78(16 Suppl):Abstract nr B045.

MiREM: an expectation-maximization approach for prioritizing miRNAs associated with gene-set

BackgroundThe knowledge of miRNAs regulating the expression of sets of mRNAs has led to novel insights into numerous and diverse cellular mechanisms. While a single miRNA may regulate many genes, one gene can be regulated by multiple miRNAs, presenting a complex relationship to model for accurate predictions.ResultsHere, we introduce miREM, a program that couples an expectation-maximization (EM) algorithm to the common approach of hypergeometric probability (HP), which improves the prediction and prioritization of miRNAs from gene-sets of interest. miREM has been made available through a web-server (https://bioinfo-csi.nus.edu.sg/mirem2/) that can be accessed through an intuitive graphical user interface. The program incorporates a large compendium of human/mouse miRNA-target prediction databases to enhance prediction. Users may upload their genes of interest in various formats as an input and select whether to consider non-conserved miRNAs, amongst filtering options. Results are reported in a rich graphical interface that allows users to: (i) prioritize predicted miRNAs through a scatterplot of HP p-values and EM scores; (ii) visualize the predicted miRNAs and corresponding genes through a heatmap; and (iii) identify and filter homologous or duplicated predictions by clustering them according to their seed sequences.ConclusionWe tested miREM using RNAseq datasets from two single “spiked” knock-in miRNA experiments and two double knock-out miRNA experiments. miREM predicted these manipulated miRNAs as having high EM scores from the gene set signatures (i.e. top predictions for single knock-in and double knock-out miRNA experiments). Finally, we have demonstrated that miREM predictions are either similar or better than results provided by existing programs.

Abstract A21: Characterization of the immune environment in the in situ to invasive breast carcinoma transition

Abstract Reactivation of immune responses against cancer cells—immunotherapy—is one of the few cancer therapies that can successfully eliminate even metastatic disease in a relatively nontoxic manner. However, its success has been limited to a subset of patients. For example, in breast cancer only ~20% of triple-negative breast cancer (TNBC) patients benefit from anti-PDL1 therapy. One reason for this limited success can be that different tumors evade the immune system via different mechanisms, which suggests that they may respond to different types of immunotherapies. Epithelial cancer cells in ductal carcinoma in situ (DCIS) are physically separated from the tumor-infiltrating leukocytes by the myoepithelial cell layer and the basement membrane, whereas in invasive ductal carcinoma (IDC), the epithelial cancer cells are intermingled with leukocytes. Therefore, we hypothesize that the DCIS to IDC transition is a key step in tumor progression as cancer cells are under different selection pressures, and only those that can evade the immune system can continue tumor progression, hence shaping subsequent tumor evolution. To dissect the role of leukocytes in the DCIS to IDC transition, we began by analyzing the composition and molecular profiles of leukocytes, with special emphasis on T cells, in normal breast tissues, DCIS, and IDC. We found that the relative frequency of leukocytes increases during tumor progression but the CD8/CD4 T cell ratio decreases. In addition, the gene expression profile of CD45+CD3+ T cells is different in DCIS compared to those isolated from normal breast tissue and IDCs. We found that gene set signatures corresponding to CD8+ T cells and NKT cells were enriched over regulatory T-cell signatures in DCIS compared to IDC. This result suggested that DCIS had a more activated immune environment compared to IDC. We further examined T-cell activation by immunofluorescence (IF) analysis and found a higher percentage of activated GZMB+CD8+ T cells in DCIS compared to IDC including a set of matched DCIS and locally recurrent IDC. We also found that the TCR clonotype was more diverse in DCIS than in IDCs. Interestingly, we detected a few relatively frequent clones that were shared among different DCIS patients, one of which was previously shown to recognize a protein from the Epstein-Bar virus. In order to dissect mechanisms of immune evasion in IDC, we analyzed immune checkpoint genes and proteins by FISH and IF. We found that TIGIT+ T cells were slightly more frequent in DCIS than in IDC. In triple-negative IDC, there was high expression of PD-L1 in epithelial cells and in 3/10 cases amplification of CD274 (encoding PD-L1), whereas DCIS had lower expression of PD-L1 and no amplification of CD274. To further elucidate mechanisms of immune evasion, we explored the significance of a cluster of genes encoding several chemokines that are located in close proximity of ERBB2 (encoding HER2). When analyzing the HER2+ samples from the TCGA, we found that coamplification of the 17q12 chemokine cluster (CC) with ERBB2 was enriched in HER2+ER+ luminal-like tumors, whereas there was either no gain or loss of the cluster in the HER2+ER breast tumors. Interestingly, we found higher expression of both T-cell activation and exhaustion-related genes in tumors that lack CC gain. Moreover, when assessing a cohort of HER2+ samples by multicolor FISH and IF, we found an inverse correlation between CC amplification and activation of CD8+ T cells. There was no correlation between CC amplification and recruitment of macrophages or myeloid-derived suppressor cells. Overall our results show coevolution of cancer cells and the immune microenvironment during tumor progression. Citation Format: Carlos R. Gil del Alcazar, SungJin Huh, Muhammad B. Ekram, Anne Trinh, Lin L. Liu, Francisco Beca, Zi Xiaoyuan, Misuk Kwak, Helga Bergholtz, Ying Su, Lina Ding, Hege G. Russnes, Andrea L. Richardson, Kirsten Babski, Elizabeth Min Hui Kim, Charles H. McDonnell, III, Jon Wagner, Ron Rowberry, Gordon J. Freeman, Deborah Dillon, Therese Sorlie, Lisa M. Coussens, Judy E. Garber, Rong Fan, Kristie Bobolis, D. Craig Allred, Joon Jeong, So Yeon Park, Franziska Michor, Kornelia Polyak. Characterization of the immune environment in the in situ to invasive breast carcinoma transition [abstract]. In: Proceedings of the AACR Special Conference: Advances in Breast Cancer Research; 2017 Oct 7-10; Hollywood, CA. Philadelphia (PA): AACR; Mol Cancer Res 2018;16(8_Suppl):Abstract nr A21.

PO-340 Dysregulation of the TP53 network and PRC2 activity are independent poor prognostic factors in patients with malignant peripheral nerve sheath tumours

Introduction Malignant peripheral nerve sheath tumours (MPNSTs) are rare and aggressive tumours with neuroectodermal origin, mainly affecting adolescents and young adults. Genetic aberrations in the EED and SUZ12 genes, encoding core components of the Polycomb repressive complex 2 (PRC2), have been found in up to 80% of MPNSTs. Homozygous losses of PRC2 components have been shown to result in complete loss of trimethylation of lysine 27 (H2K27m3) and this has been found to be an indicator of poor prognosis for MPNST patients. Furthermore, TP53 is one of few recurrently mutated genes, but the clinical relevance of the TP53 network in MPNSTs remains inconclusive. Here, we have analysed the prognostic impact of dysregulated PRC2 and TP53 activity on the gene expression-level in an aggregated series of MPNSTs. Material and methods Frozen samples from 60 MPNSTs and 15 neurofibromas from patients treated at specialised sarcoma centres in Norway, Sweden, and Italy were subjected to gene expression analyses (Genome-Wide GeneChip Human Transcriptome Array 2.0, Affymetrix). Single-sample gene set enrichment analysis was performed for a gene set upregulated in breast tumours with TP53 mutation [Miller et al. Proc Natl Acad Sci U S A. 2005;102(38:13550–5)] and a gene set upregulated in MPNSTs with PRC2 inactivation [Lee et al. Nat Genet. 2014;46(11:1227–32)]. Results and discussions Both the TP53 network and PRC2 activity were down-regulated in MPNSTs compared with the benign neurofibromas, as determined by a significantly higher gene set score (p=3 × 10–10 and p=0.0009, respectively). Furthermore, down-regulation of both processes were associated with inferior survival for MPNST patients (HR=4.1 [95% CI, 1.73–9.79], p=0.001 and HR=4.9 [95% CI, 1.65–14.69], p=0.004, respectively). In multivariable analyses, including the clinical variables NF1 status, sex, age at diagnosis, tumour location, tumour size, and remission status, both gene sets were independent prognostic markers, however the TP53 gene set signature had the strongest prognostic impact. Patients with down-regulation of both processes (n=25, 42%) had a particularly poor survival (HR=9.1, CI:2.1–39.9, p=0.003). Conclusion Dysregulation of the TP53 network and PRC2 function distinguish MPNSTs from benign neurofibromas and are independently associated with inferior patient survival.

Generalized estimating equations by considering additive terms for analyzing time-course gene sets data

Time-course gene sets are collections of predefined groups of genes in some patients gathered over time. The analysis of time-course gene sets for testing gene sets which vary significantly over time is an important context in genomic data analysis. In this paper, the method of generalized estimating equations (GEEs), which is a semi-parametric approach, is applied to time-course gene set data. We propose a special structure of working correlation matrix to handle the association among repeated measurements of each patient over time. Also, the proposed working correlation matrix permits estimation of the effects of the same gene among different patients. The proposed approach is applied to an HIV therapeutic vaccine trial (DALIA-1 trial). This data set has two phases: pre-ATI and post-ATI which depend on a vaccination period. Using multiple testing, the significant gene sets in the pre-ATI phase are detected and data on two randomly selected gene sets in the post-ATI phase are also analyzed. Some simulation studies are performed to illustrate the proposed approaches. The results of the simulation studies confirm the good performance of our proposed approach.

Identification of functionally unique CD4 T cells that are the circulating counterparts of epidermal resident memory T cells

Abstract As a barrier organ, the skin contains specialized T cell populations that combat infection and also help maintain tissue homeostasis and promote wound repair. Circulating skin-tropic T cells can also be identified in the blood based on their expression of the cutaneous lymphocyte antigen (CLA), but the developmental and functional relationships between these circulating CLA+ T cells and tissue-resident T cells in the skin are not fully understood. Using 33-parameter Cytometry by time-of-flight (CyTOF) analysis we identified a novel population of skin-tropic CD4+ T cells in human blood that expresses the CD103 integrin and phenotypically and functionally resembles epidermal resident memory T cells (TRM) in the skin. RNA sequencing identified a set of signature genes shared by circulating CD103+CLAhi T cells in the blood and epidermal TRM cells in the skin, and suggested that CD103+CLAhiT cells contribute to normal skin function and response to tissue damage. Finally, using humanized NSG mice carrying full-thickness human skin grafts we demonstrated that, despite their identification as TRM, CD103+CLAhi T cells in the skin are capable of exiting the tissue and re-entering circulation. Thus, CD103+CLAhi T cells in the blood represent a circulating T cell population of cutaneous TRM, and this provides novel opportunities for the study and therapeutic manipulation of TRM cells in the contexts of cutaneous infection, inflammation, and tissue-repair.

Two new functions in the WormBase Enrichment Suite.

Genome-wide experiments routinely generate large amounts of data that can be hard to interpret biologically. A common approach to interpreting these results is to employ enrichment analyses of controlled languages, known as ontologies, that describe various biological parameters such as gene molecular or biological function. In C. elegans, three distinct ontologies, the Gene Ontology (GO), Anatomy Ontology (AO), and the Worm Phenotype Ontology (WPO) are used to annotate gene function, expression and phenotype, respectively (Ashburner et al. 2000; Lee and Sternberg, 2003; Schindelman et al. 2011). Previously, we developed software to test datasets for enrichment of anatomical terms, called the Tissue Enrichment Analysis (TEA) tool (Angeles-Albores and Sternberg, 2016). Using the same hypergeometric statistical method, we extend enrichment testing to include WPO and GO, offering a unified approach to enrichment testing in C. elegans. The WormBase Enrichment Suite can be accessed via a user-friendly interface at http://www.wormbase.org/tools/enrichment/tea/tea.cgi. To validate the tools, we analyzed a previously published extracellular vesicle (EV)-releasing neuron (EVN) signature gene set derived from dissociated ciliated EV neurons (Wang et al. 2015) using WormBase Enrichment Suite based on the WS262 WormBase release. TEA correctly identified the CEM, hook sensillum and IL2 neuron as enriched tissues. The top phenotype associated with the EVN signature was chemosensory behavior. Gene Ontology enrichment analysis showed that cell projection and cell body were the most enriched cellular components in this gene set, followed by the biological processes neuropeptide signaling pathway and vesicle localization further down. The tutorial script used to generate the figure above can be viewed at: https://github.com/dangeles/TissueEnrichmentAnalysis/blob/master/tutorial/Tutorial.ipynb The addition of Gene Enrichment Analysis (GEA) and Phenotype Enrichment Analysis (PEA) to WormBase marks an important step towards a unified set of analyses that can help researchers to understand genomic datasets. These enrichment analyses will allow the community to fully benefit from the data curation ongoing at WormBase.

Hypoxia stimulates microenvironment in human embryonic stem cell through inflammatory signalling: An integrative analysis

Despite, several lines of evidence suggesting the possible role of hypoxia in stem cell development and differentiation its significance in conferring the stemness and pluripotency remains elusive. In the present study we sought to delineate the candidate genes and molecular pathways imposed during hypoxic microenvironment and its physiological relevance in tipping the balance between the niche and cellular differentiation. Integrated meta-analysis was performed between the hypoxia exposed and normal human embryonic stem cells, employing three transcriptomic cohorts (GSE35819, GSE9510 and GSE37761) retrieved from Gene expression omnibus (GEO) database. Results reveal that a total number of 12 genes were consistently differentially expressed (6up regulated and 6 down regulated) with FDR <0.05 and fold change >1.5. The Gene Ontology (GO) functions and Kyoto encyclopedia of genes and genomes (KEGG) pathway analysis was performed using DAVID. The GO analysis showed DEG significantly enriched in terms of Cellular process (GO:0009987), protein binding (GO:0005515) and cell part (GO:0044464). KEGG analysis indicated participation of genes associated with circadian rthyum regulation and PPAR signalling pathway. Further, gene-set signature (MsigDB) enrichment analysis showed positive regulation with inflammatory signals and negative association with PPAR and p53 pathway. Protein-protein network of gene modules suggests significant hub proteins viz. CTTNB1 (Degree = 18), IL8 (Degree = 15), NFKB1 (Degree = 15) and RELA (Degree = 15) in the PPI network. MCODE algorithm was used for subnetworks of the PPI network. Our integrative analysis documents the potential candidate genes which serves distinct roles influencing metabolic shift and induce inflammatory effectors contributing to hypoxic mediated stem cell niche.

Abstract P4-09-14: Analysis of breast cancer recurrence using gene set enrichment analysis

Abstract Background: Even after successful treatment of primary breast tumors, there is a continued risk of recurrence. The risk varies between subtypes and there are ongoing efforts that aim to improve prediction of such risks for individual patients. Detection of subclinical metastases might be achieved by biomarkers in blood. In this study, we profiled protein expression in blood plasma from patients with known clinical outcome (recurrence vs no recurrence) to identify prognostic markers of breast cancer recurrence. Methods: The subjects and specimens were made available through the Clinical Breast Care Project using IRB-approved protocols. We analyzed blood plasma samples taken at the time of diagnosis from consented patients who subsequently relapsed (33 cases) as well as those with no disease recurrence (31 controls). Based on hormone receptor and lymph node status the samples were grouped as: ER-/HER2- (17 cases/15 controls), ER+/LN+ (10/10) and ER+/LN- (6/6). We used aptamer-based SOMAscan assay platform to study the expression of 1252 proteins. We analyzed the protein expression data by using their coding genes in order to apply the Gene Set Enrichment Analysis method (GSEA v.2, Broad Institute). Pathway databases of KEGG, REACTOME, BIOCARTA and C4 collection were used. Significant gene sets were called at 5% FDR, and overlaps and low coverage gene sets (Tags &amp;lt;70%) were removed. Statistical analysis and clustering were done using R. Results: Unsupervised clustering showed some difference in signal in the ER+/LN- group. Even though there was a lack of significantly differentiated proteins between the cases and controls of this group, many significant gene sets were identified. After applying the cutoff filters and removing the overlaps, there were 5 gene sets enriched with the pathway collection, involved in B-cell receptor signaling, mRNA metabolism, tight junction and SCF-KIT signaling. Similarly, 9 gene sets from the MORF compendium were differentially expressed with the C4 collection and included neighborhood genes of NME2, ACTG1, EIF3S2, AP2M1, DAP3, UBE2I, NPM1, AATF and NPM1. In contrast, neither differentially expressed proteins nor gene sets were identified from the ER+/LN+ and ER-/HER2- groups. Since the sample size of the ER+/LN- group was small, we conducted a similar analysis by randomly choosing 6 case and control samples in the other two groups respectively. There were still no differentially expressed proteins or gene sets identified above the specified cutoff parameters. Conclusion: Using plasma protein expression data we identified underlying gene sets differentially expressed between ER+/LN- patients who had cancer recurrence and no recurrence. Many genes in these sets were already known biomarkers (e.g. PTEN, AKT1, STAT3, SET etc.). These results can be used for understanding patterns of recurrence in different cancer subtypes. Further research is needed to estimate the clinical significance of these gene products. The views expressed in this article are those of the author and do not reflect the official policy of the Department of Army/Navy/Air Force, the Department of Defense, or U.S. Government. Citation Format: Praveen Kumar A, Kovatich AJ, Biancotto A, Cheung F, Davidson-Moncada JK, Kvecher L, Liu J, Ru Y, Kovatich AW, Deyarmin B, Fantacone-Campbell JL, Hooke JA, Raj Kumar PK, Rui H, Hu H, Shriver CD. Analysis of breast cancer recurrence using gene set enrichment analysis [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P4-09-14.

TNM stage at diagnosis is more predictive of prognosis than pathological complete response in young breast cancer treated with neoadjuvant chemotherapy.

TNM stage at diagnosis is more predictive of prognosis than pathological complete response in young breast cancer treated with neoadjuvant chemotherapy.

EAMA: Empirically adjusted meta-analysis for large-scale simultaneous hypothesis testing in genomic experiments.

Recent developments in high throughput genomic assays have opened up the possibility of testing hundreds and thousands of genes simultaneously. However, adhering to the regular statistical assumptions regarding the null distributions of test statistics in such large-scale multiple testing frameworks has the potential of leading to incorrect significance testing results and biased inference. This problem gets worse when one combines results from different independent genomic experiments with a possibility of ending up with gross false discoveries of significant genes. In this article, we develop a meta-analysis method of combining p-values from different independent experiments involving large-scale multiple testing frameworks, through empirical adjustments of the individual test statistics and p-values. Even though, it is based on various existing ideas, this specific combination is novel and potentially useful. Through simulation studies and real genomic datasets we show that our method outperforms the standard meta-analysis approach of significance testing in terms of accurately identifying the truly significant set of genes.

Integrating genome-wide association study and expression quantitative trait locus study identifies multiple genes and gene sets associated with schizophrenia

Schizophrenia is a serious mental disease with high heritability. To better understand the genetic basis of schizophrenia, we conducted a large scale integrative analysis of genome-wide association study (GWAS) and expression quantitative trait loci (eQTLs) data. GWAS summary data was derived from a published GWAS of schizophrenia, containing 9394 schizophrenia patients and 12,462 healthy controls. The eQTLs dataset was obtained from an eQTLs meta-analysis of 5311 subjects, containing 923,021 cis-eQTLs for 14,329 genes and 4732 trans-eQTLs for 2612 genes. Genome-wide single gene expression association analysis was conducted by SMR software. The SMR analysis results were further subjected to gene set enrichment analysis (GSEA) to identify schizophrenia associated gene sets. SMR detected 49 genes significantly associated with schizophrenia. The top five significant genes were CRELD2 (p value=1.65×10−11), DIP2B (p value=3.94×10−11), ZDHHC18 (p value=1.52×10−10) and ZDHHC5 (p value=7.45×10−10), C11ORF75 (p value=3.70×10−9). GSEA identified 80 gene sets with p values <0.01. The top five significant gene sets were COWLING_MYCN_TARGETS (p value <0.001) and CHR16P11 (p value <0.001), ACTACCT_MIR196A_MIR196B (p value=0.002), CELLULAR_COMPONENT_DISASSEMBLY (p value=0.002) and GRAESSMANN_RESPONSE_TO_MC_AND_DOXORUBICIN_DN (p value=0.002). Our results provide useful information for revealing the genetic basis of schizophrenia.

SAIC: an iterative clustering approach for analysis of single cell RNA-seq data

BackgroundResearch interests toward single cell analysis have greatly increased in basic, translational and clinical research areas recently, as advances in whole-transcriptome amplification technique allow scientists to get accurate sequencing result at single cell level. An important step in the single-cell transcriptome analysis is to identify distinct cell groups that have different gene expression patterns. Currently there are limited bioinformatics approaches available for single-cell RNA-seq analysis. Many studies rely on principal component analysis (PCA) with arbitrary parameters to identify the genes that will be used to cluster the single cells.ResultsWe have developed a novel algorithm, called SAIC (Single cell Analysis via Iterative Clustering), that identifies the optimal set of signature genes to separate single cells into distinct groups. Our method utilizes an iterative clustering approach to perform an exhaustive search for the best parameters within the search space, which is defined by a number of initial centers and P values. The end point is identification of a signature gene set that gives the best separation of the cell clusters. Using a simulated data set, we showed that SAIC can successfully identify the pre-defined signature gene sets that can correctly separated the cells into predefined clusters. We applied SAIC to two published single cell RNA-seq datasets. For both datasets, SAIC was able to identify a subset of signature genes that can cluster the single cells into groups that are consistent with the published results. The signature genes identified by SAIC resulted in better clusters of cells based on DB index score, and many genes also showed tissue specific expression.ConclusionsIn summary, we have developed an efficient algorithm to identify the optimal subset of genes that separate single cells into distinct clusters based on their expression patterns. We have shown that it performs better than PCA method using published single cell RNA-seq datasets.

Latent HIV-1 TAR Regulates 7SK-responsive P-TEFb Target Genes and Targets Cellular Immune Responses in the Absence of Tat

The transactivating response element (TAR) structure of the nascent HIV-1 transcript is critically involved in the recruitment of inactive positive transcription elongation factor b (P-TEFb) to the promoter proximal paused RNA polymerase II. The viral transactivator Tat is responsible for subsequent P-TEFb activation in order to start efficient viral transcription elongation. In the absence of the viral transactivator of transcription (Tat), e.g., during latency or in early stages of HIV transcription, TAR mediates an interaction of P-TEFb with its inhibitor hexamethylene bis-acetamide-inducible protein 1 (HEXIM1), keeping P-TEFb in its inactive form. In this study, we address the function of HIV-1 TAR in the absence of Tat by analyzing consequences of HIV-1 TAR overexpression on host cellular gene expression. An RNA chimera consisting of Epstein-Barr virus-expressed RNA 2 (EBER2) and HIV-1 TAR was developed to assure robust overexpression of TAR in HEK293 cells. The overexpression results in differential expression of more than 800 human genes. A significant proportion of these genes is involved in the suppression of cellular immune responses, including a significant set of 7SK-responsive P-TEFb target genes. Our findings identify a novel role for HIV-1 TAR in the absence of Tat, involving the interference with host cellular immune responses by targeting 7SK RNA-mediated gene expression and P-TEFb inactivation.

Diagnostic accuracy of a selected signature gene set that discriminates active pulmonary tuberculosis and other pulmonary diseases

We validated the accuracy of host selected signature gene set using unstimulated whole blood (WB), and peripheral blood mononuclear cells (PBMC) in the diagnosis of tuberculosis (TB). The unstimulated WB and PBMC from 1417 individuals with active pulmonary TB patients, other lung diseases and healthy participants were analyzed using real time polymerase chain reaction (RT-PCR). The WB cohort test demonstrates that the combination of GBP5 and KLF2 can differentiate active TB versus HC with sensitivity and specificity of 77.8% and 87.1%, respectively; but most importantly active TB versus OD with sensitivity and specificity of 96.1% and 85.2%, respectively. Again during treatment course, the TB score of GBP5 and KLF2, analytes secretion and clinical parameters were found to be associated in disease progression. In the PBMC cohort test, we found that the only and best discriminatory combination was GBP5, DUSP3 and KLF2 inthe active TB versus HC with a sensitivity and specificity of 76.4% and 85.9%, respectively. Our study reveals that GBP5 and KLF2 may be useful as a diagnostic tool for active TB, also the two-gene set may serve as surrogate biomarkers for monitoring TB therapy.

Abstract 552: Multi-scale omics integration using parallel heatmap clustering for the systemic analysis and biomarker discovery of drug sensitivity in lymphoma cell lines

Abstract Background. Pharmacogenomics studies the role of genomics in drug response. By measuring the individual genome, it is hoped that pharmaceutical drug treatments can deviate from a "one-dose-fits-all" approach to a more "personalized" treatment. To achieve this goal, we need to explain which genomic differences cause the lack of response and to be able to predict the response from baseline omics data, that is from data available before starting treatment. Here, we present an integrated omics approach to analyze drug sensitivity based on in vitro experiments. Methods. We measured the drug response in 61 lymphoma cell lines for a number of anti-cancer drugs using a standard MTT cell proliferation assay. The baseline genomics of these cell lines were fully profiled for gene expression, copy number and methylation. Pathway and gene set signatures were computed using GO, KEGG, Biocarta and lymphoma databases to provide gene set level features. Additionally, we computed high level "biological concept" features. These multi-scale features were directly correlated with drug sensitivity but also correlated between the multiple data types. Using a novel parallel clustering approach, we integrated our data for the different omics types and at different scales: gene, gene set and concept level. Finally, a shortest path algorithm extracted the most probable explanation between genotype and response phenotype. Results. As a proof of concept, we first applied our method to the classification of activated B- cell like (ABC) and germinal center B (GCB) subtypes in diffuse large B-cell lymphomas (DLBCL). In accordance to previous knowledge, our method showed the MYD88/CD40/NFKB axis as strongly upregulated in the ABC, while, albeit less pronounced, the PI3K/MTOR and NOTCH signalling pathways were more enriched in the GCB subtype. We then applied our method to the drug sensitivity data of our lymphoma cell line panel. The analysis showed that drug response was largely driven by differential methylation rather than copy number aberrations. We found that B cell receptor signaling (BCR) and CD40 signalling were among the most correlated to drug sensitivity in a drug-specific manner. Genes in these pathways (such as SYK, SPIB and CD79A) appeared as good biomarkers for drug-specific response in lymphoma. Conclusion. Using an integrated analysis of multiple omics data at multiple scales, we successfully identified functional modules related with drug sensitivity in lymphoma cell lines. The same method can be applied to study the sensitivity to other drugs in either hematological and/or solid cancer types. Citation Format: Ivo Kwee, Andrea Rinaldi, Alberto J. Arribas, Eugenio Gaudio, Chiara Tarantelli, Filippo Spriano, Petra Hillmann, Francesco Bertoni. Multi-scale omics integration using parallel heatmap clustering for the systemic analysis and biomarker discovery of drug sensitivity in lymphoma cell lines [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 552. doi:10.1158/1538-7445.AM2017-552

Abstract 3872: Retinoic acid (RA) relieves EZH2-mediated epigenetic suppression in high-risk neuroblastoma (HR-NB)

Abstract High risk neuroblastoma tumors (HR-NB) are characterized by the presence of undifferentiated histopathology, while tumors from low risk patients have histologic evidence of differentiation and a transcriptome enriched in differentiation genes. Determining the cellular and molecular underpinnings that contribute to this undifferentiated state will lead to better understanding of the disease pathogenesis. Retinoic acid (RA) is known to prime embryonic stem cells for lineage specific differentiation. To elucidate the early events important for reprogramming of NB cells leading to changes in the degree of cellular differentiation, we treated HR-NB cells (KCNR : MYCN amplified) with 5µM RA for 24 hrs. RNA-Seq analysis, to assess early changes in the transcriptome, showed significant changes in gene expression [2fold-up-regulated-446, 2fold-down-regulated-250, FDR≤0.01] which were enriched for biological functions like neuron projection, neuron differentiation, indicating onset of neurogenesis as a marker of differentiation. Gene set enrichment analysis identified significant positive enrichment of NB differentiation (p=0.001) and PRC2 (Polycomb Repressor Complex2) gene sets (p=0.02) along with downregulation of MYCN signature gene set (p=0.001). PRC2 component EZH2 (Histone Methyltransferase) is highly expressed in NB and represses differentiation genes. Though, EZH2 protein levels did not change after treatment, RA caused 40% reduction in H3 bound EZH2. To gain insight into how this reduction affects subsequent gene expression; we evaluated changes in genome wide binding of EZH2 and H3K27me3 after treatment. Indeed, significant correlation was observed between the changes in expression levels of genes after RA treatment and relative binding of EZH2 (p=3.88e-6) and H3K27me3 (p=3.88e-6) at their promoter regions. One-third (n=147) of the upregulated genes (eg. HEY1, RARB) and half (n=117) of the downregulated genes (eg. INSM2, CAMK4) were EZH2 targets. Upregulated genes were related to biological functions like negative regulation of cell growth and transcription, nervous system development. To probe the underlying mechanism, in-silico analysis indicated RA decreased Wnt/ß-catenin, eNOS and PTEN pathways while stimulating cAMP, Integrin linked kinase and G-coupled protein pathways. Evaluation of the cAMP pathway candidates identified 13 genes as EZH2 targets. eg CNR1, DUSP6, PKIA, RSG4. In primary NB tumors, microarray studies indicate that low levels of CNR1 and DUSP6 are associated with poor prognosis. RA stimulates 10-fold and 2-fold increases in CNR1 and DUSP6 levels, respectively. RA treatment inhibits EZH2 binding to its target genes thereby activating pathways involved in reprogramming of HR-NBs to a more differentiated state. Studies are underway to further evaluate the molecular mechanism leading to decreased EZH2 binding upon RA treatment. Citation Format: Deblina Banerjee, Zhihui Liu, Doo-Yi Oh, Maggie Cam, Bong-Hyun Kim, Carol Thiele. Retinoic acid (RA) relieves EZH2-mediated epigenetic suppression in high-risk neuroblastoma (HR-NB) [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 3872. doi:10.1158/1538-7445.AM2017-3872