Curated Gene Sets Research Articles

A novel gene selection method for gene expression data for the task of cancer type classification

Cancer is a poligenetic disease with each cancer type having a different mutation profile. Genomic data can be utilized to detect these profiles and to diagnose and differentiate cancer types. Variant calling provide mutation information. Gene expression data reveal the altered cell behaviour. The combination of the mutation and expression information can lead to accurate discrimination of different cancer types. In this study, we utilized and transferred the information of existing mutations for a novel gene selection method for gene expression data. We tested the proposed method in order to diagnose and differentiate cancer types. It is a disease specific method as both the mutations and expressions are filtered according to the selected cancer types. Our experiment results show that the proposed gene selection method leads to similar or improved performance metrics compared to classical feature selection methods and curated gene sets.

Clinical, molecular, metabolic, and immune features associated with oxidative phosphorylation in melanoma brain metastases.

BackgroundRecently, we showed that melanoma brain metastases (MBMs) are characterized by increased utilization of the oxidative phosphorylation (OXPHOS) metabolic pathway compared to melanoma extracranial metastases (ECMs). MBM growth was inhibited by a potent direct OXPHOS inhibitor, but observed toxicities support the need to identify alternative therapeutic strategies. Thus, we explored the features associated with OXPHOS to improve our understanding of the pathogenesis and potential therapeutic vulnerabilities of MBMs.MethodsWe applied an OXPHOS gene signature to our cohort of surgically resected MBMs that had undergone RNA-sequencing (RNA-seq) (n = 88). Clustering by curated gene sets identified MBMs with significant enrichment (High-OXPHOS; n = 21) and depletion (Low-OXPHOS; n = 25) of OXPHOS genes. Clinical data, RNA-seq analysis, and immunohistochemistry were utilized to identify significant clinical, molecular, metabolic, and immune associations with OXPHOS in MBMs. Preclinical models were used to further compare melanomas with High- and Low-OXPHOS and for functional validation.ResultsHigh-OXPHOS MBMs were associated with shorter survival from craniotomy compared to Low-OXPHOS MBMs. High-OXPHOS MBMs exhibited an increase in glutamine metabolism, and treatment with the glutaminase inhibitor CB839 improved survival in mice with MAPKi-resistant, High-OXPHOS intracranial xenografts. High-OXPHOS MBMs also exhibited a transcriptional signature of deficient immune activation, which was reversed in B16-F10 intracranial tumors with metformin treatment, an OXPHOS inhibitor.ConclusionsOXPHOS is associated with distinct clinical, molecular, metabolic, and immune phenotypes in MBMs. These associations suggest rational therapeutic strategies for further testing to improve outcomes in MBM patients.

Dysregulation of the mevalonate pathway during SARS-CoV-2 infection: An in silico study.

SARS-CoV-2 triggers a dysregulated innate immune system activation. As the mevalonate pathway (MVP) prevents the activation of inflammasomes and cytokine release and regulates endosomal transport, compromised signaling could be associated with the pathobiology of COVID-19. Prior transcriptomic studies of host cells in response to SARS-CoV-2 infection have not reported to date the effects of SARS-CoV-2 on the MVP. In this study, we accessed public data sets to report in silico investigations into gene expression. In addition, we proposed candidate genes that are thought to have a direct association with the pathogenesis of COVID-19, and which may be dependent on signals derived from the MVP. Our results revealed dysregulation of genes involved in the MVP. These results were not found when investigating the gene expression data from host cells infected with H3N2 influenza virus, H1N1 influenza virus, or respiratory syncytial virus. Our manually curated gene set showed significant gene expression variability in A549 cells infected with SARS-CoV-2, as per Blanco-Melo et al. data set (GSE147507). In light of the present findings, SARS-CoV-2 could hijack the MVP, leading to hyperinflammatory responses. Prompt reconstitution of this pathway with available agents should be considered in future studies.

Single Cell Immune Profiling Reveals Distinct T Cell Clones and Functional States in in-Vitro Expanded Cord Blood Derived Gamma Delta T Cells

Gamma-delta (γδ) T cells represent a special class of unconventional T cells defined by their expression of the somatically rearranged T cell receptor (TCR) γ and δ chains. Unlike TCRαβ, it has been reported that different TCRγδ are also able to bind to their antigens in the context of non-classical MHC-like molecules or totally independent of the MHC complexes. Additionally, a variety of NK receptors are known to be expressed by γδ T cells, conferring their ability to sense alternative classes of cancer-associated antigens in a multimodal manner. Such a diverse mode of antigen recognition possibly endows γδ T cells with a wide spectrum of functional activation program. Our team had previously explored the potential of expanding cord blood (CB) derived γδ T cells (CB-gdT) as well as their corresponding ability to target primary acute myeloid leukemia (AML) cells. Using a feeder cell line-based in vitro expansion protocol, we achieved a clinically relevant scale expansion of γδ T cells over a period of 14 days. These cells exhibit variable degree of potency against a range of human AML cell lines and primary patient samples. In order to dissect the cellular and molecular programs governing the activation, differentiation and functional states of our in vitro expanded CB-gdT, we performed multiplex single cell sequencing analysis using the 10X Genomics Chromium System. After initial quality check and filtering, data from a total of 4,276 cells were retrieved. Among which, we identified 742 unique TCRγδ clonotypes, representing 18.6% of the starting 4,000 FACS purified γδ T cells seeded for expansion. The largest 10% of the clones was found to make up 60.9% of the total retrieved cells, demonstrating a significant extent of clonal focusing in our expansion cultures. Consistent to our FACS analysis, Vδ1 is the predominant TRD chain in the expanded cultures, accounting for 61.2% of all clones. Vγ4 is the most prevalent TRG chain making up to 24.9% of all clones regardless of the paired Vδ subtype. Notably, however, the largest γδ T cell clone did not utilize Vγ4, indicating that Vγ4 clones, although frequent, are not the most proliferative clone. These data are supportive of the adaptive characteristics of CB-gdTs, likely in a TCRγδ dependent manner. Based on uniform manifold approximation and projection for dimension reduction (UMAP), all cells were clustered into 11 subsets. Key cytotoxic genes including GZMB, GZMA and NKG7 were all highly expressed across all clusters, indicating that the expanded cells were indeed functionally cytotoxic. Comparing against multiple curated gene sets, we have identified 3 main subsets of γδ T cells: the Proliferative, Cytotoxic γδ T cells (P-CT), Differentiated Cytotoxic γδ T cells (D-CT) and Late Activated Cytotoxic γδ T cells (LA-CT). P-CT (~46% of all cells) shows an expression profile positively associated with cell proliferation as well as increased cell surface expression of memory T cell markers CD27, CCR7 and CD62L. Similar to cytotoxic genes, genes associated with TCR signaling and interferon response were found to be expressed across all cell clusters, yet with elevated levels in D-CT and LA-CT. Furthermore, cell surface expression of different NK receptors including NKG2D, DNAM1 and NKp30 are more enriched in LA-CT compared to the other 2 subsets, suggesting the acquisition of additional NK receptor related functions in this group of cells. Consistent with the concept of progressive γδ T cell differentiation and activation in culture, we found that in 85 (11.5%) of the γδ T cell clones bearing more than 10 cells each, all clones contain cells distributed across the 3 different γδ T cell subsets. Further analysis did not reveal any relationship between the relative proportion of the subsets within each clone with clone size nor any specific type of delta/gamma chain. Taken together, our high-resolution transcriptome analysis suggests that as CB-gdT expand and differentiate in culture, they are likely to adopt dynamic memory and signal -specific functional programs. More importantly, our data highlights the rich clonal and cellular composition of in vitro expanded CB-gdT. These unique characteristics of our CB-gdT can overcome the challenges of tumor heterogeneity and cell persistence, with the potential of improving outcomes in cell immunotherapy. Disclosures Tan: Tessa Therapeutics Ltd: Current Employment.

Effects of tesamorelin on hepatic transcriptomic signatures in HIV-associated NAFLD.

Nonalcoholic fatty liver disease (NAFLD) is a common comorbidity among people living with HIV that has a more aggressive course than NAFLD among the general population. In a recent randomized placebo-controlled trial, we demonstrated that the growth hormone–releasing hormone analog tesamorelin reduced liver fat and prevented fibrosis progression in HIV-associated NAFLD over 1 year. As such, tesamorelin is the first strategy that has shown to be effective against NAFLD among the population with HIV. The current study leveraged paired liver biopsy specimens from this trial to identify hepatic gene pathways that are differentially modulated by tesamorelin versus placebo. Using gene set enrichment analysis, we found that tesamorelin increased hepatic expression of hallmark gene sets involved in oxidative phosphorylation and decreased hepatic expression of gene sets contributing to inflammation, tissue repair, and cell division. Tesamorelin also reciprocally up- and downregulated curated gene sets associated with favorable and poor hepatocellular carcinoma prognosis, respectively. Notably, among tesamorelin-treated participants, these changes in hepatic expression correlated with improved fibrosis-related gene score. Our findings inform our knowledge of the biology of pulsatile growth hormone action and provide a mechanistic basis for the observed clinical effects of tesamorelin on the liver.

Abstract 4418: Integrated whole exome and transcriptome analyses of the tumor and microenvironment provide new opportunities for rational design of cancer therapy

Abstract Identifying the unique molecular signature of tumors is essential for rational design of personalized treatments tailored to specific genomic alterations. It is now well accepted that the non-malignant microenvironment plays a critical role in tumor behavior and response to therapy. However, to date, multi-compartmental comprehensive genomic and transcriptomic analyses of malignant and nonmalignant compartments remain rare, are highly unstructured, and difficult to translate for practical treatment decisions. Here, we present a new analytic platform that integrates whole exome sequence (WES) and whole transcriptome (RNASeq) data to provide comprehensive tumor portraits. The portrait is based on the analysis of 28 newly curated gene sets capturing selective and specific molecular and functional properties of nonmalignant components of tumor microenvironment (such as fibroblasts, macrophages, dendritic cells, lymphocytes, extracellular matrices and blood and lymphatic networks), as well as intrinsic properties and genomic profile of malignant cells, including mutational and neoantigenic load, oncogenic and clinically actionable alterations. Analysis of tumor RNASeq data from over 8500 patients using unsupervised louvain dense clustering analysis based on detection of tightly connected networks revealed four unique and consistent microenvironment types characterized by distinct immune, stromal and cancer cell composition. Type A (immune “inflamed” cancers) were characterized by increased fibrosis and low mutational burden; Type B was characterized by increased infiltrating cytotoxic T lymphocytes and high mutations. Type C and D were characterized by negligible immune infiltration and minimal non-malignant cellular populations, while type C had increased vascularization and fibrosis. Remarkably, the four microenvironment types have high prognostic significance and are differently associated with immunotherapy response. In over 400 patients with skin and bladder cancers, type B was associated with a response rate (CR/PR) of >38% following immunotherapy (anti-PD1 /CTLA4) vs <10% in patients with type C. Importantly, dynamic changes in the microenvironment types correlated with response to immunotherapy. By identifying the unique combination of driver genomic alterations, immune processes and stromal components present in each tumor, this novel multicompartmental analysis provides a robust tool for better prediction of response to immunotherapy and for future tailoring of personalized therapeutic combinations. This comprehensive model, summarized and visualized as a Molecular-Functional portrait of the tumor (MF-portrait), could ultimately lead to better clinical outcomes. Citation Format: Alexander Bagaev, Nikita Kotlov, Felix Frenkel, Viktor Svekolkin, Nava Almog, Ravshan Ataullakhanov, Nathan Fowler. Integrated whole exome and transcriptome analyses of the tumor and microenvironment provide new opportunities for rational design of cancer therapy [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4418.

Systems biology-based approaches to summarize and identify novel genes and pathways associated with acute and chronic postsurgical pain

Study objectiveTo employ systems biology–based machine learning to identify biologic processes over-represented with genetic variants (gene enrichment) implicated in post-surgical pain. DesignInformed systems biology based integrative computational analyses. SettingPediatric research and teaching institution. InterventionsPubmed search (01/01/2001–10/31/2017) was performed to identify “training” genes associated with postoperative pain in humans. Candidate genes were identified and prioritized using Toppgene suite, based on functional enrichment using several gene ontology annotations, and curated gene sets associated with mouse phenotype–knockout studies. MeasurementsComputationally top-ranked candidate genes and literature-curated genes were included in pathway enrichment analyses. Hierarchical clustering was used to visualize select functional enrichment results between the two phenotypes. Main resultsLiterature review identified 38 training genes associated with postoperative pain and 31 with CPSP. We identified 2610 prioritized novel candidate genes likely associated with acute and chronic postsurgical pain, the top 10th percentile jointly enriched (p 0.05; Benjamini-Hochberg correction) several pathways, topmost being cAMP response element-binding protein and ion channel pathways. Heat maps demonstrated enrichment of inflammatory/drug metabolism processes in acute postoperative pain and immune mechanisms in CPSP. ConclusionHigh interindividual variability in pain responses immediately after surgery and risk for CPSP suggests genetic susceptibility. Lack of large homogenous sample sizes have led to underpowered genetic association studies. Systems biology can be leveraged to integrate genetic-level data with biologic processes to generate prioritized candidate gene lists and understand novel biological pathways involved in acute postoperative pain and CPSP. Such data would be key to informing future polygenic studies with targeted genome wide profiling. This study demonstrates the utility of functional annotation - based prioritization and enrichment approaches and identifies novel genes and unique/shared biological processes involved in acute and chronic postoperative pain. Results provide framework for future targeted genetic profiling of CPSP risk, to enable preventive and therapeutic approaches.

A comparison of curated gene sets versus transcriptomics-derived gene signatures for detecting pathway activation in immune cells

BackgroundDespite the significant contribution of transcriptomics to the fields of biological and biomedical research, interpreting long lists of significantly differentially expressed genes remains a challenging step in the analysis process. Gene set enrichment analysis is a standard approach for summarizing differentially expressed genes into pathways or other gene groupings. Here, we explore an alternative approach to utilizing gene sets from curated databases. We examine the method of deriving custom gene sets which may be relevant to a given experiment using reference data sets from previous transcriptomics studies. We call these data-derived gene sets, “gene signatures” for the biological process tested in the previous study. We focus on the feasibility of this approach in analyzing immune-related processes, which are complicated in their nature but play an important role in the medical research.ResultsWe evaluate several statistical approaches to detecting the activity of a gene signature in a target data set. We compare the performance of the data-derived gene signature approach with comparable GO term gene sets across all of the statistical tests. A total of 61 differential expression comparisons generated from 26 transcriptome experiments were included in the analysis. These experiments covered eight immunological processes in eight types of leukocytes. The data-derived signatures were used to detect the presence of immunological processes in the test data with modest accuracy (AUC = 0.67). The performance for GO and literature based gene sets was worse (AUC = 0.59). Both approaches were plagued by poor specificity.ConclusionsWhen investigators seek to test specific hypotheses, the data-derived signature approach can perform as well, if not better than standard gene-set based approaches for immunological signatures. Furthermore, the data-derived signatures can be generated in the cases that well-defined gene sets are lacking from pathway databases and also offer the opportunity for defining signatures in a cell-type specific manner. However, neither the data-derived signatures nor standard gene-sets can be demonstrated to reliably provide negative predictions for negative cases. We conclude that the data-derived signature approach is a useful and sometimes necessary tool, but analysts should be weary of false positives.

A Drug Recommendation System (Dr.S) for Cancer Cell Lines

Personalizing drug prescriptions in cancer care based on genomic information requires associating genomic markers with treatment effects. This is an unsolved challenge requiring genomic patient data in yet unavailable volumes as well as appropriate quantitative methods. We attempt to solve this challenge for an experimental proxy for which sufficient data is available: 42 drugs tested on 1018 cancer cell lines. Our goal is to develop a method to identify the drug that is most promising based on a cell line's genomic information. For this, we need to identify for each drug the machine learning method, choice of hyperparameters and genomic features for optimal predictive performance. We extensively compare combinations of gene sets (both curated and random), genetic features, and machine learning algorithms for all 42 drugs. For each drug, the best performing combination (considering only the curated gene sets) is selected. We use these top model parameters for each drug to build and demonstrate a Drug Recommendation System (Dr.S). Insights resulting from this analysis are formulated as best practices for developing drug recommendation systems. The complete software system, called the Cell Line Analyzer, is written in Python and available on github.

Proteomic Profiling of HTLV-1 Carriers and ATL Patients Reveal TNFR2 As a Novel Diagnostic and Chemosensitivity Biomarker for ATL

Adult T-cell leukemia/lymphoma (ATL) is a peripheral T-cell malignancy associated with the human T-cell leukemia virus type I (HTLV-1). Classification of ATL into clinical subtypes acute, lymphoma, chronic and smoldering types was proposed based on prognostic factors, clinical features and natural history of the disease. Although HTLV-1 infection alone is not sufficient to cause ATL and only about 5% of HTLV-1 carriers progress to ATL, the prognosis is generally poor especially for patients with aggressive ATL (i.e., acute, lymphoma or unfavorable chronic types), with a median survival time at around 1 year, even after chemotherapy. Currently, biomarkers to predict ATL onset and progression are limited, making early diagnosis and treatment for ATL challenging. To develop early diagnostic biomarkers for ATL, we performed, for the first time, an extensive proteomic profiling of HTLV-1 carriers and ATL patients as a foundation for establishing a blood-based biomarker panel for ATL. Expression levels of 1305 plasma proteins in HTLV-1 carriers (n=40), untreated ATL patients (n=40, 28 acute; 4 lymphoma; 5 chronic; 3 smoldering), and remission status (n=5) were measured by SOMAscan assay (SomaLogic Inc, Boulder, CO). ATL diagnosis was based on criteria proposed by the Japan Clinical Oncology Group (JCOG) and identification of monoclonal integration of HTLV-1 proviral DNA using Southern blot hybridization method. Deregulated proteins in HTLV-1 versus ATL versus remission states were ranked by significance (Welch's t-test) and discrimination capacity (area under the curve [AUC]). In addition, machine learning algorithms were used to set discrimination boundaries for HTLV-1, ATL, and remission states using some of the top deregulated proteins. Statistical analyses were performed using Python 3.6.2 software. To elucidate on ATL pathogenesis, we further analyzed our proteomic data using Gene Set Enrichment Analysis (GSEA 3.0 hallmarks, curated gene sets) and Gene Ontology (GO Panther Pathways) and determined pathway deregulation among disease states as well as among ATL subtypes. Overrepresented pathways in ATL versus HTLV-1 included inflammation mediated by cytokine and chemokine signaling, angiogenesis, notch signaling, and IL6/JAK/STAT3, among others. Among a total of 176 proteins which were categorized as extremely significant (p<0.00001) with AUC scores ranging from 0.90-0.99, we further confirmed plasma protein concentrations of CD223 (LAG3) (n=40, 2 healthy individuals; 8 HTLV-1 carriers; 10 acute; 8 lymphoma; 6 chronic; 6 smoldering), CD30 (TNFRSF8), TIM-3 (HAVCR2) (n=40, 2 healthy individuals; 8 HTLV-1 carriers; 12 acute; 6 lymphoma, 6 chronic; 6 smoldering), and TNFR2 (TNFRSF1B) (n=79, 5 healthy individuals; 16 HTLV-1 carriers; 26 acute; 9 lymphoma; 11 chronic; 9 smoldering), through ELISA. We discovered significantly higher CD30 and TIM-3 levels in acute ATL versus HTLV-1 carriers (p<0.05) and remarkably high TNFR2 levels among aggressive ATL patients, acute (p<0.001) and lymphoma types (p<0.01) versus HTLV-1 carriers. In addition, a significant decrease in TNFR2 levels among ATL patients who have achieved remission was also seen (p<0.001). These results indicate the potential value of TNFR2 not only as a diagnostic biomarker for ATL, but also for predicting response or failure to therapy. Furthermore, this study represents a novel proteomic approach in developing candidate biomarkers for ATL, by combining data from high-throughput SOMAmer technology, machine learning, proteomic pathway analysis, in addition to previously well-established techniques such as ELISA. Further investigation of TNFR2 and other potential biomarkers in this study need to be done in the near future. Disclosures Fukushima: Daiichi-Sankyo: Research Funding.

A novel gene selection method for gene expression data for the task of cancer type classification

Cancer is a poligenetic disease with each cancer type having a different mutation profile. Genomic data can be utilized to detect these profiles and to diagnose and differentiate cancer types. Variant calling provide mutation information. Gene expression data reveal the altered cell behaviour. The combination of the mutation and expression information can lead to accurate discrimination of different cancer types. In this study, we utilized and transferred the information of existing mutations for a novel gene selection method for gene expression data. We tested the proposed method in order to diagnose and differentiate cancer types. It is a disease specific method as both the mutations and expressions are filtered according to the selected cancer types. Our experiment results show that the proposed gene selection method leads to similar or improved performance metrics compared to classical feature selection methods and curated gene sets.

Uncovering biomarker genes with enriched classification potential from Hallmark gene sets

Given the complex relationship between gene expression and phenotypic outcomes, computationally efficient approaches are needed to sift through large high-dimensional datasets in order to identify biologically relevant biomarkers. In this report, we describe a method of identifying the most salient biomarker genes in a dataset, which we call “candidate genes”, by evaluating the ability of gene combinations to classify samples from a dataset, which we call “classification potential”. Our algorithm, Gene Oracle, uses a neural network to test user defined gene sets for polygenic classification potential and then uses a combinatorial approach to further decompose selected gene sets into candidate and non-candidate biomarker genes. We tested this algorithm on curated gene sets from the Molecular Signatures Database (MSigDB) quantified in RNAseq gene expression matrices obtained from The Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx) data repositories. First, we identified which MSigDB Hallmark subsets have significant classification potential for both the TCGA and GTEx datasets. Then, we identified the most discriminatory candidate biomarker genes in each Hallmark gene set and provide evidence that the improved biomarker potential of these genes may be due to reduced functional complexity.

Curation and bioinformatic analysis of strabismus genes supports functional heterogeneity and proposes candidate genes with connections to RASopathies

Strabismus refers to the misalignment of the eyes and occurs in 2–4% of individuals. The low-resolution label “strabismus” covers a range of heterogeneous defects, which makes it challenging to unravel this condition. Consequently a coherent understanding of the causes is lacking. Here, we attempt to gain a better understanding of the underlying genetics by combining gene curation, diverse bioinformatic analyses (including gene ontology, pathway mapping, expression and network-based methods) and literature review. Through a phenotype-based curation process, we identify high-confidence and permissive sets of 54 and 233 genes potentially involved in strabismus. These genes can be grouped into 10 modules that together span a heterogeneous set of biological and molecular functions, and can be linked to clinical sub-phenotypes. Multiple lines of evidence associate retina and cerebellum biology with the strabismus genes. We further highlight a potential role of the Ras-MAPK pathway. Independently, sets of 11 genes and 15 loci tied to strabismus with definitive genetic basis have been compiled from the literature. We identify strabismus candidate genes for 5 of the 15 reported loci (CHD7; SLC9A6; COL18A1, COL6A2; FRY, BRCA2, SPG20; PARK2). Finally, we synthesize a Strabismus Candidate Gene Collection, which together with our curated gene sets will serve as a resource for future research. The results of this informatics study support the heterogeneity and complexity of strabismus and point to specific biological pathways and brain regions for future focus.

Abstract P1-09-06: Gene-set enrichment analysis (GSEA) of breast tissue from healthy women with less than six months history of breastfeeding shows enrichment in Hedgehog signaling, notch signaling and luminal progenitor gene signatures

Abstract Introduction: Multiple epidemiological studies have shown that prolonged breastfeeding is associated with a reduced risk of developing triple negative/basal-like breast cancer (TN/BLBC). We have modeled abrupt involution (AI) due to lack of breastfeeding and gradual involution (GI) of the mammary gland that occurs over time upon prolonged breastfeeding in wild-type FVB/N mice and discovered prominent histological and molecular changes in the AI glands over time. Our studies revealed for the first time a clear and persistent expansion of mammary luminal progenitor (LP) epithelial cells in AI glands (AACR abstract#2242, 2018). Here, we corroborate animal studies using normal human breast tissue obtained from a reduction mammoplasty tissue collection study (OSU-2011C0094). Methods: Breast tissue obtained from parous premenopausal women with no history of breast cancer who breastfed for ≥6 months (GI, n=16) versus those who breastfed for <6 months (AI, n=16) (OSU-2011C0094) was used for gene expression analysis. RNA isolated from these normal mammary tissues was analyzed using Affymatrix Gene ChIP Human Transcriptome array 2.0; Gene Set Enrichment Analysis (GSEA) was used to analyze the microarray data. Molecular Signatures Database was used in GSEA querying C2 curated gene sets, Hallmark gene sets, and Lim-Mammary-Luminal-Progenitor gene sets. H&E sections of the breast tissue were used to assess lobular type by counting number of ductules per terminal ductal lobular unit (TDLU). False discovery rate (FDR) q-values and p-values were used for multiple comparison adjustment. Results: GSEA revealed that breast tissue obtained from women in the AI cohort exhibited strong positive enrichment for Notch and Hedgehog Signaling (Hhg) pathways (FDR q-value= 0.20 and 0.12, respectively). In GI women, GSEA showed an overall trend towards enrichment in metabolic pathways and immune system functions. Moreover, there was non-significant trend towards positive enrichment of mouse LP gene signature in AI women only (FDR q-value= 0.30). Age and BMI were not statistically different between AI and GI cohorts. Analysis of TDLU, the primary anatomical source of most breast cancers, revealed that breast tissue from AI women had proportionally higher lobular type 1 only epithelium than GI women who exhibited more differentiated lobular epithelium (p-value= 0.049). Conclusion: We report here for the first time that mammary glands from women who breastfed <6 months were enriched for stem-cell signaling pathways and LP gene signature. This reflects some similarity to BRCA1 mutation carriers, who demonstrate expanded luminal progenitor population. In addition, higher Type 1 TDLU's are seen in breast tissue from parous women who breastfed <6 months. Together, these data demonstrate features for TN/BLBC precursors enriched in patients who breastfed for <6 months. Understanding this mechanistic link will help in developing prevention strategies, particularly for African-American women who have lower prevalence of breastfeeding and higher incidence of TN/BLBC. Citation Format: Basree MM, Shinde N, Palettas M, Weng D, Stover DG, Sizemore GM, Shields P, Majumder S, Ramaswamy B. Gene-set enrichment analysis (GSEA) of breast tissue from healthy women with less than six months history of breastfeeding shows enrichment in Hedgehog signaling, notch signaling and luminal progenitor gene signatures [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P1-09-06.

NutriGenomeDB: a nutrigenomics exploratory and analytical platform.

Habitual consumption of certain foods has shown beneficial and protective effects against multiple chronic diseases. However, it is not clear by which molecular mechanisms they may exert their beneficial effects. Multiple -omic experiments available in public databases have generated gene expression data following the treatment of human cells with different food nutrients and bioactive compounds. Exploration of such data in an integrative manner offers excellent possibilities for gaining insights into the molecular effects of food compounds and bioactive molecules at the cellular level. Here we present NutriGenomeDB, a web-based application that hosts manually curated gene sets defined from gene expression signatures, after differential expression analysis of nutrigenomics experiments performed on human cells available in the Gene Expression Omnibus (GEO) repository. Through its web interface, users can explore gene expression data with interactive visualizations. In addition, external gene signatures can be connected with nutrigenomics gene sets using a gene pattern-matching algorithm. We further demonstrate how the application can capture the primary molecular mechanisms of a drug used to treat hypertension and thus connect its mode of action with hosted food compounds.

SU81 - COMPREHENSIVE EVALUATION OF ENRICHMENT FOR CIRCADIAN CLOCK GENE SETS IN PSYCHIATRIC TRAITS: SPECIFIC ENRICHMENT IN CLINICAL RESPONSE TO LITHIUM

Background A disrupted circadian clock has been linked to the risk of several neuropsychiatric disorders. Likewise, circadian rhythms have been suggested as mediators of the mechanism of action of lithium in bipolar disorder. Nonetheless the relationship between the ‘clock genes’ that regulate circadian rhythms and lithium treatment response is not completely understood. To our knowledge there has not been a systematic pathway / enrichment analysis of clock genes in the context of psychiatric traits, in general, and of clinical response to lithium, in particular. The objective of this study was to perform formal gene set enrichment analyses for circadian clock genes, using publicly available GWAS summary statistics from several psychiatric disorders, as well as the results from The International Consortium on Lithium Genetics (ConLiGen) GWAS (Hou et al., 2016). Methods Based on previous literature (Pizarro et al., 2013; Chen et al., 2016) and available resources (e.g. http://circadb.hogeneschlab.org/) curated gene sets related to circadian control were generated. These sets can be divided in three categories: clock modulator (upstream) genes, core clock genes and clock controlled (downstream) genes. GWAS summary statistics from schizophrenia, bipolar disorder, major depressive disorder, ADHD, autism, Alzheimer's disease and continuous/dichotomous lithium response were used as reference. Gene set enrichment analyses were carried out using both INRICH and MAGMA. Results Gene set enrichment analyses using INRICH and MAGMA reported a significant enrichment of a set of 19 genes that constitute the ‘core clock’ in the dichotomous lithium response phenotype (INRICH: empirical P=0.001; corrected P=0.008; MAGMA: competitive P=0.005; corrected P=0.0336). None of the circadian gene sets showed a significant enrichment in any of the other psychiatric traits included in this study (all corrected P>0.05). Discussion Our results suggest the involvement of those genes that constitute the core clock machinery in the determination of the clinical response to lithium in bipolar disorder patients. The specificity of these results suggests that the participation of circadian rhythms is especially relevant in the modulation of lithium response rather than in the overall risk of mental illness. A better understanding of potential links between circadian mechanisms, genetic risk factors, and the lithium treatment response may open new avenues into the clinical management of bipolar disorder.

Machine learning approach to literature mining for the genetics of complex diseases.

To generate a parsimonious gene set for understanding the mechanisms underlying complex diseases, we reasoned it was necessary to combine the curation of public literature, review of experimental databases and interpolation of pathway-associated genes. Using this strategy, we previously built the following two databases for reproductive disorders: The Database for Preterm Birth (dbPTB) and The Database for Preeclampsia (dbPEC). The completeness and accuracy of these databases is essential for supporting our understanding of these complex conditions. Given the exponential increase in biomedical literature, it is becoming increasingly difficult to manually maintain these databases. Using our curated databases as reference data sets, we implemented a machine learning-based approach to optimize article selection for manual curation. We used logistic regression, random forests and neural networks as our machine learning algorithms to classify articles. We examined features derived from abstract text, annotations and metadata that we hypothesized would best classify articles with genetically relevant content associated to the disorder of interest. Combinations of these features were used build the classifiers and the performance of these feature sets were compared to a standard ‘Bag-of-Words’. Several combinations of these genetic based feature sets outperformed ‘Bag-of-Words’ at a threshold such that 95% of the curated gene set obtained from the original manual curation of all articles were extracted from the articles classified by machine learning as ‘considered’. The performance was superior in terms of the reduction of required manual curation and two measures of the harmonic mean of precision and recall. The reduction in workload ranged from 0.814 to 0.846 for the dbPTB and 0.301 to 0.371 for the dbPEC. Additionally, a database of metadata and annotations is generated which allows for rapid query of individual features. Our results demonstrate that machine learning algorithms can identify articles with relevant data for databases of genes associated with complex diseases.

Mechanism of Synergistic Effects on T Cell Activation By Avadomide (CC-122) and Nivolumab

Background: Avadomide (CC-122) is a cereblon modulator that promotes ubiquitination and degradation of the hematopoietic transcription factors Ikaros and Aiolos, leading to immunomodulation, such as T cell activation and increased interleukin-2 (IL-2) production in primary peripheral blood mononuclear cells (PBMCs). The immune checkpoint inhibitor nivolumab (nivo), an anti-PD-1 antibody, induces immune activation and can enhance immune response against various solid tumors. Previously, we have shown that the combination of avadomide and nivo synergistically enhance IL-2 production, T cell proliferation, and immune-mediated cytotoxicity, relative to single agent activity. To understand molecular mechanisms underlying these synergistic effects, we compared the effects of avadomide, nivo, or the combination on gene expression in primary human T cells using whole transcriptome RNA sequencing and differential pathway analysis.Methods: PBMCs were isolated from healthy donors (N=6), treated with DMSO/IgG, avadomide 50 nM, nivo 10 µg/mL, or avadomide and nivo for 1 hour, then stimulated with 0.5 ng/mL staphylococcus enterotoxin B for 48 hours. Culture supernatants were collected for cytokine analysis; T cells were isolated by magnetic cell separation for RNA extraction. RNA was sequenced by Illumina HiSeq v4; data was filtered to transcripts ≥10 counts across all samples and processed by DESeq2. Significantly differentially expressed genes (FDR-adj. P values <0.05) were computed for each treatment group relative to DMSO/IgG controls. Pathway analysis was performed with the GSEA Molecular Signatures Database (MSigDB) using the Hallmark and C2 curated gene sets, which comprise a diverse set of biological pathways, including KEGG and Reactome, to provide unbiased enrichment analysis. T cell-related pathways from the MSigDB C5 (Gene Ontology) collection were used to investigate specific effects on immune function. Synergy was defined by the fractional product method.Results: Avadomide, nivo, and the combination enhanced IL-2 production in SEB-stimulated PBMCs by 282%, 47%, and 586% respectively, compared with DMSO/IgG controls, confirming the synergistic effects on cytokine production. The top pathways upregulated in each treatment group included the following T cell related pathways: for avadomide - T cell receptor (TCR) signaling, JAK/STAT, cytokine/receptor interaction; for nivo - CD8 TCR pathway and HIF1A targets; and for the combination - TCR signaling, JAK/STAT, and HDAC3 targets which are required for T cell maturation and cytokine production. Combination treatment uniquely upregulated pathways including the Biocarta cytotoxic T cell pathway and calcium signaling in CD4+ T cells. Among the 7732 genes modified by any treatment, 1949 were uniquely differentially expressed by the combination. A targeted enrichment analysis using only T cell-related pathways from the C5 collection revealed that these uniquely differentially expressed genes represent processes such as T cell differentiation, proliferation, and activation.Conclusions: At the gene expression level, many T cell-related pathways were upregulated by one or more single agent and/or the combination. However, the presence of uniquely differentially regulated genes suggests that combination treatment induced broader effects in pathways involved in T cell differentiation, proliferation, and activation than either avadomide or nivo alone. Interestingly, the cytotoxic T cell pathway was uniquely upregulated by the combination, consistent with our previous finding of a significant increase in cytotoxicity with avadomide/nivo in combination. Calcium signaling in CD4+ T cell genes were also uniquely upregulated, suggesting that avadomide/nivo combination effects may involve nuclear factor of activated T cells (NFAT)-dependent T cell immune regulation. These data provide molecular support for the in vitro phenotypic activity of avadomide and checkpoint blockade on enhancing T cell activity. Avadomide is now under investigation in combination with checkpoint blockade in solid tumors (NCT02859324) and with CAR T therapy in lymphoma (NCT03310619). DisclosuresNakayama:Celgene Corporation: Employment, Equity Ownership. Stokes:Celgene Corporation: Employment, Equity Ownership. Waldman:Celgene Corporation: Employment, Equity Ownership. Hagner:Celgene Corporation: Employment, Equity Ownership. Gandhi:Celgene Corporation: Employment, Equity Ownership.

Candidate gene and pathway analyses identifying genetic variations associated with prasugrel pharmacokinetics and pharmacodynamics

AimWe aimed to investigate the genetic polymorphisms and pharmacogenetic variability associated with the pharmacodynamics (PD) and pharmacokinetics (PK) of prasugrel, in healthy Han Chinese subjects. Patients & methodsHealthy, native, Han Chinese subjects (n = 36) aged 18 to 45 years with unknown genotypes were included. All subjects received a loading dose (LD) on day 1 and a maintenance dose (MD) from day 2 until day 11. Candidate gene association and gene-set analysis of biological pathways related to prasugrel and platelet activity were analyzed. Results28 SNPs of 17 candidate genes previously associated with prasugrel or platelet activity were selected after a literature search. In the 30 mg LD groups (n = 24), ITGA2-rs28095 was found to be significantly associated with the P2Y12 reaction unit (PRU) value at 24 h after the LD (p = 0.015). 165 study genes related to platelet activation-related processes and prasugrel activity were selected from the MSigDB database, including curated gene sets from KEGG, Bio Carta, and Gene Cards. 14 SNPs of 9 genes were found to be significantly correlated both at 24 h and 12 days after LD: ADAMTSL1, PRKCA, ITPR2, P2RY12, P2RY14, PLCB4, PRKG1, ADCY1, and LYN. Seven SNPs of 6 protein-coding genes associated with area under the concentration-time curve (AUC0-tlast) were significantly identified among the 47 selected genes, including ADAMTSL1, CD36, P2RY1, PCSK9, PON1, and SCD. ConclusionThese results show that genetic variation affects the PK and PD of prasugrel in normal individuals. Further studies with larger sample sizes are required to explore whether the SNPs are associated only with prasugrel activity or also with cardiovascular events and all-cause mortality.

Wnt5a/FZD4 Mediates the Mechanical Stretch-Induced Osteogenic Differentiation of Bone Mesenchymal Stem Cells

Background/Aims: Mechanical stimulation and WNT signalling have essential roles in regulating the osteogenic differentiation of bone marrow stromal cells (BMSCs) and bone formation. However, little is known regarding the regulation of WNT signalling molecule expression and therefore the osteogenic differentiation of BMSCs during osteogenesis. Methods: Microarrays of BMSCs from elderly individuals or patients with osteoporosis (GSE35959) from the GEO database were analysed using GeneSight-Lite 4.1.6 (BioDiscovery) and C2 curated gene sets downloaded from Molecular Signatures Database (MSigDB). Realtime PCR and western blotting were used to measure the expression of the indicated genes. ALP and Alizarin red staining were used to evaluate the osteogenesis of BMSCs. Results: In this study, we investigated whether mechanical loading directly regulates the expression of WNT signalling molecules and examined the role of WNT signalling in mechanical loading-triggered osteogenic differentiation and bone formation. We first studied the microarrays of samples from patients with osteoporosis and found downregulation of the GPCR ligand binding gene set in the BMSCs of patients with osteoporosis. Then, we demonstrated that mechanical stimuli can regulate osteogenesis and bone formation both in vivo and in vitro. FZD4 was upregulated during cyclic mechanical stretch (CMS)-induced osteogenic differentiation, and the JNK signalling pathway was activated. FZD4 knockdown inhibited the mechanical stimuli-induced osteogenesis and JNK activity. More importantly, we found an activating effect of WNT5A and FZD4 that regulated bone formation in response to hindlimb unloading in mice, and pretreatment with WNT5A or activation of the expression of FZD4 partly rescued the osteoporosis caused by mechanical unloading. Conclusions: Our results demonstrate, for the first time, that mechanical stimulation alters the expression of genes involved in the osteogenic differentiation of BMSCs via the direct regulation of FZD4 and that therapeutic WNT5A and FZD saRNA may be an efficient strategy for enhancing bone formation under mechanical stimulation.