Cancer Cell Line Encyclopedia Dataset Research Articles

SVExpress: identifying gene features altered recurrently in expression with nearby structural variant breakpoints

BackgroundCombined whole-genome sequencing (WGS) and RNA sequencing of cancers offer the opportunity to identify genes with altered expression due to genomic rearrangements. Somatic structural variants (SVs), as identified by WGS, can involve altered gene cis-regulation, gene fusions, copy number alterations, or gene disruption. The absence of computational tools to streamline integrative analysis steps may represent a barrier in identifying genes recurrently altered by genomic rearrangement.ResultsHere, we introduce SVExpress, a set of tools for carrying out integrative analysis of SV and gene expression data. SVExpress enables systematic cataloging of genes that consistently show increased or decreased expression in conjunction with the presence of nearby SV breakpoints. SVExpress can evaluate breakpoints in proximity to genes for potential enhancer translocation events or disruption of topologically associated domains, two mechanisms by which SVs may deregulate genes. The output from any commonly used SV calling algorithm may be easily adapted for use with SVExpress. SVExpress can readily analyze genomic datasets involving hundreds of cancer sample profiles. Here, we used SVExpress to analyze SV and expression data across 327 cancer cell lines with combined SV and expression data in the Cancer Cell Line Encyclopedia (CCLE). In the CCLE dataset, hundreds of genes showed altered gene expression in relation to nearby SV breakpoints. Altered genes involved TAD disruption, enhancer hijacking, and gene fusions. When comparing the top set of SV-altered genes from cancer cell lines with the top SV-altered genes previously reported for human tumors from The Cancer Genome Atlas and the Pan-Cancer Analysis of Whole Genomes datasets, a significant number of genes overlapped in the same direction for both cell lines and tumors, while some genes were significant for cell lines but not for human tumors and vice versa.ConclusionOur SVExpress tools allow computational biologists with a working knowledge of R to integrate gene expression with SV breakpoint data to identify recurrently altered genes. SVExpress is freely available for academic or commercial use at https://github.com/chadcreighton/SVExpress. SVExpress is implemented as a set of Excel macros and R code. All source code (R and Visual Basic for Applications) is available.

DeepDSC: A Deep Learning Method to Predict Drug Sensitivity of Cancer Cell Lines.

High-throughput screening technologies have provided a large amount of drug sensitivity data for a panel of cancer cell lines and hundreds of compounds. Computational approaches to analyzing these data can benefit anticancer therapeutics by identifying molecular genomic determinants of drug sensitivity and developing new anticancer drugs. In this study, we have developed a deep learning architecture to improve the performance of drug sensitivity prediction based on these data. We integrated both genomic features of cell lines and chemical information of compounds to predict the half maximal inhibitory concentrations [Formula: see text] on the Cancer Cell Line Encyclopedia (CCLE) and the Genomics of Drug Sensitivity in Cancer (GDSC) datasets using a deep neural network, which we called DeepDSC. Specifically, we first applied a stacked deep autoencoder to extract genomic features of cell lines from gene expression data, and then combined the compounds' chemical features to these genomic features to produce final response data. We conducted 10-fold cross-validation to demonstrate the performance of our deep model in terms of root-mean-square error (RMSE) and coefficient of determination [Formula: see text]. We show that our model outperforms the previous approaches with RMSE of 0.23 and [Formula: see text] of 0.78 on CCLE dataset, and RMSE of 0.52 and [Formula: see text] of 0.78 on GDSC dataset, respectively. Moreover, to demonstrate the prediction ability of our models on novel cell lines or novel compounds, we left cell lines originating from the same tissue and each compound out as the test sets, respectively, and the rest as training sets. The performance was comparable to other methods.

Clinical Implications of Aberrant PD-1 and CTLA4 Expression for Cancer Immunity and Prognosis: A Pan-Cancer Study.

Combination therapy with inhibitors of cytotoxic T lymphocyte-associated protein (CTLA)4 and programmed death (PD)-1 has demonstrated efficacy in cancer patients. However, there is little information on CTLA4 and PD-1 expression levels and their clinical significance across diverse cancers. In this study, we addressed this question by analyzing PD-1 and CTLA4 levels in 33 different types of cancer along with their prognostic significance using The Cancer Genome Atlas (TCGA) and Cancer Cell Line Encyclopedia datasets. Liver hepatocellular carcinoma (LIHC) patients receiving cytokine-induced killer cell (CIK) immunotherapy at Sun Yat-sen University cancer center were enrolled for survival analysis. The correlation between PD-1/CTLA4 expression and cancer immunity was also analyzed. The results showed that PD-1 and CTLA4 transcript levels varied across cancer cell lines, with aberrant expression detected in certain cancer types; Kaplan–Meier analysis with the Cox proportional hazards model showed that this was closely related to overall survival in breast invasive carcinoma, glioblastoma multiforme, head and neck squamous cell carcinoma, acute myeloid leukemialymphoma, uterine corpus endometrial carcinoma, and uveal melanoma in TCGA. High serum PD-1 and CTLA4 levels predicted better survival in LIHC patients receiving CIK therapy. PD-1 and CTLA4 levels were found to be significantly correlated with the degree of tumor cell infiltration using Tumor Immune Estimation Resource, Estimating Relative Subsets of RNA Transcripts, and Estimation of Stromal and immune Cells in Malignant Tumor Tissues Using Expression Data as well as with tumor-infiltrating lymphocyte marker expression; they were also related to tumor mutation burden, microsatellite instability, mismatch repair, and the expression of DNA methyltransferases in some cancer types. Gene set enrichment analysis of 33 cancer types provided further evidence for associations between PD-1/CTLA4 levels and cancer development and immunocyte infiltration. Thus, PD-1 and CTLA4 play important roles in tumorigenesis and tumor immunity and can serve as prognostic biomarkers in different cancer types.

An Improved Anticancer Drug-Response Prediction Based on an Ensemble Method Integrating Matrix Completion and Ridge Regression

In this study, we proposed an ensemble learning method, simultaneously integrating a low-rank matrix completion model and a ridge regression model to predict anticancer drug response on cancer cell lines. The model was applied to two benchmark datasets, including the Cancer Cell Line Encyclopedia (CCLE) and the Genomics of Drug Sensitivity in Cancer (GDSC). As previous studies suggest, the dual-layer integrated cell line-drug network model was one of the best models by far and outperformed most state-of-the-art models. Thus, we performed a head-to-head comparison between the dual-layer integrated cell line-drug network model and our model by a 10-fold crossvalidation study. For the CCLE dataset, our model has a higher Pearson correlation coefficient between predicted and observed drug responses than that of the dual-layer integrated cell line-drug network model in 18 out of 23 drugs. For the GDSC dataset, our model is better in 26 out of 28 drugs in the phosphatidylinositol 3-kinase (PI3K) pathway and 26 out of 30 drugs in the extracellular signal-regulated kinase (ERK) signaling pathway, respectively. Based on the prediction results, we carried out two types of case studies, which further verified the effectiveness of the proposed model on the drug-response prediction. In addition, our model is more biologically interpretable than the compared method, since it explicitly outputs the genes involved in the prediction, which are enriched in functions, like transcription, Src homology 2/3 (SH2/3) domain, cell cycle, ATP binding, and zinc finger.

Abstract 33: Drug sensitivity prediction modeling from genomics, transcriptomics and inferred protein activity

Abstract Background: Machine learning models that rely on single omics data for drug sensitivity prediction are challenging and frequently fail within precision medicine scenarios. Proteomics, for example, reflects the system biology and the regulatory network better than genomics. However, it is less likely available for many preclinical and in vivo data, mainly due to the cost of data generation. Currently, there are massive amounts of genomics data available, such as CNV, mutation, and RNA-seq signatures. However, these data do not characterize the post-translational modifications in proteins, limiting their utility for biomarker discovery. Objective: The main objective of this study is to overcome the lack of proteomics data gap. Herein, we propose a novel modeling approach to improve the prediction accuracy of drug sensitivity, that is, combining genomics and proteomics signatures. In addition to genomics and transcriptomics data, the model infers proteomic activity from gene expression using VIPER from in vitro data, and this is extendable to in vivo settings. Material and Methods: Using PharmacoGX package developed in our lab, we downloaded, curated, and annotated the genomic and pharmacologic data of the Cancer Cell line Encyclopedia (CCLE), as well as the Cancer Therapeutics Response Portal (CTRPV2) dataset that is a continuation of the CTRP project and the largest pharmacologic screen conducted to date, containing several hundreds of thousands of drug dose-response curves. In this study, we extracted from CCLE dataset the following signatures: RPPA, RNASeq, Mutation, and CNV, then inferred the VIPER protein using RNASeq. Then, we built a model that a) checks the different omics combinations, b) applies random forest with ten-fold cross-validation for sensitivity prediction using CTRPV2, and c) evaluates the significance of each model using the concordance index (CI) package developed in our lab. Results: The proposed model was tested in vitro using CCLE and CTRPv2 common cell lines. We tested the model with drugs having known biomarkers in pharmacogenomics literature. ERBB2 biomarker for lapatinib showed the best CI=0.95 by combining CNV and VIPER models while CI equals 0.77 and 0.9 for each of them, respectively. Moreover, MET biomarker for crizotinib showed best CI=0.89 by integrating RNASeq, Mutation, RPPA, and VIPER, while each model obtained CI=(0.45, 0.5, 0.85, 0.5), respectively. Conclusion: In conclusion, omics integration boosted the drug sensitivity prediction compared to single models. The application of the proposed model in vivo will improve drug development and increase the prediction quality of precision medicine. Citation Format: Hassan Mahmoud, Benjamin Haibe-Kains. Drug sensitivity prediction modeling from genomics, transcriptomics and inferred protein activity [abstract]. In: Proceedings of the AACR Special Conference on Advancing Precision Medicine Drug Development: Incorporation of Real-World Data and Other Novel Strategies; Jan 9-12, 2020; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(12_Suppl_1):Abstract nr 33.

Synthetic lethality between VPS4A and VPS4B triggers an inflammatory response in colorectal cancer.

Somatic copy number alterations play a critical role in oncogenesis. Loss of chromosomal regions containing tumor suppressors can lead to collateral deletion of passenger genes. This can be exploited therapeutically if synthetic lethal partners of such passenger genes are known and represent druggable targets. Here, we report that VPS4B gene, encoding an ATPase involved in ESCRT‐dependent membrane remodeling, is such a passenger gene frequently deleted in many cancer types, notably in colorectal cancer (CRC). We observed downregulation of VPS4B mRNA and protein levels from CRC patient samples. We identified VPS4A paralog as a synthetic lethal interactor for VPS4B in vitro and in mouse xenografts. Depleting both proteins profoundly altered the cellular transcriptome and induced cell death accompanied by the release of immunomodulatory molecules that mediate inflammatory and anti‐tumor responses. Our results identify a pair of novel druggable targets for personalized oncology and provide a rationale to develop VPS4 inhibitors for precision therapy of VPS4B‐deficient cancers.

Molecular Profiling Establishes Genetic Features Predictive of the Efficacy of the p110β Inhibitor KIN-193.

Aberrant activation of the PI3K pathway is a common alteration in human cancers. Therapeutic intervention targeting the PI3K pathway has achieved limited success due to the intricate balance of its different components and isoforms. Here, we systematically investigated the genomic and transcriptomic signatures associated with response to KIN-193, a compound specifically targeting the p110β isoform. By integrating genomic, transcriptomic, and drug response profiles from the Genomics of Drug Sensitivity in Cancer database, we identified mutational and transcriptomic signatures associated with KIN-193 and further created statistical models to predict the treatment effect of KIN-193 in cell lines, which may eventually be clinically valuable. These predictions were validated by analysis of the external Cancer Cell Line Encyclopedia dataset. These results may assist precise therapeutic intervention targeting the PI3K pathway. SIGNIFICANCE: These findings provide new insights into molecular signatures associated with sensitivity of the p110β inhibitor KIN-193, which may provide a useful guide for developing precise treatment methods for cancer.

PD-L1 and PD-L2 expression correlated genes in non-small-cell lung cancer

BackgroundProgrammed cell death ligand-1 (PD-L1) and ligand-2 (PD-L2) interaction with programmed cell death protein-1 (PD-1) represent an immune-inhibiting checkpoint mediating immune evasion and is, accordingly, an important target for blockade-based immunotherapy in cancer. In non-small-cell lung cancer (NSCLC), improved understanding of PD-1 checkpoint blockade-responsive biology and identification of biomarkers for prediction of a clinical response to immunotherapy is warranted. Thus, in the present study, we systematically described PD-L1 and PD-L2 expression correlated genes in NSCLC.MethodsWe performed comparative retrospective analyses to identify PD-L1 and PD-L2 mRNA expression correlated genes in NSCLC. For this, we examined available datasets from the cancer cell line encyclopedia (CCLE) project lung non-small-cell (Lung_NSC) and the cancer genome atlas (TCGA) projects lung adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC).ResultsAnalysis of the CCLE dataset Lung_NSC identified expression correlation between PD-L1 and PD-L2. Moreover, we identified expression correlation between 489 genes and PD-L1, 191 genes and PD-L2, and 111 genes for both. PD-L1 and PD-L2 also expression correlated in TCGA datasets LUAD and LUSC. In LUAD, we identified expression correlation between 257 genes and PD-L1, 914 genes and PD-L2, and 211 genes for both. In LUSC, we identified expression correlation between 26 genes and PD-L1, 326 genes and PD-L2, and 13 genes for both. Only a few genes expression correlated with PD-L1 and PD-L2 across the CCLE and TCGA datasets. Expression of Interferon signaling-involved genes converged in particular with the expression correlated genes for PD-L1 in Lung_NSC, for PD-L2 in LUSC, and for both PD-L1 and PD-L2 in LUAD. In LUSC, PD-L1, and to a lesser extent PD-L2, expression correlated with chromosome 9p24 localized genes, indicating a chromosome 9p24 topologically associated domain as an important driver of in particular LUSC PD-L1 expression. Expression correlation analyses of the PD-L1 and PD-L2 receptors programmed cell death protein-1 (PD-1), Cluster of differentiation 80 (CD80), and Repulsive guidance molecule B (RGMB) showed that PD-1 and CD80 expression correlated with both PD-L1 and PD-L2 in LUAD. CD80 expression correlated with PD-L2 in LUSC.ConclusionsWe present gene signatures associated with PD-L1 and PD-L2 mRNA expression in NSCLC which could possess importance in relation to understand PD-1 checkpoint blockade-responsive biology and development of gene signature based biomarkers for predicting clinical responses to immunotherapy.

Abstract A037: C1orf116, a gene of unknown function, promotes tight junctions and epithelial cell phenotype in prostate cancer

Abstract While advances in the treatment of localized prostate cancer (PCa) have improved five-year survival to near 100%, metastatic disease remains incurable. Lethal metastases are the end-result of a cancer cell that escapes the primary tumor, travels through the vasculature, and eventually invades and colonizes a secondary site. Thus, in order for a cancer cell to metastasize from the primary tumor, it must move. In a critical early event of metastasis, a rare population of cells in the primary tumor undergoes a phenotype switch from a proliferating epithelial cell to gain mesenchymal cell characteristics, including high migratory capacity, during the process of epithelial-to-mesenchymal transition (EMT). Using a multistudy gene expression discovery analysis, we previously identified C1orf116 as a novel candidate driver of EMT (doi: 10.1186/s12885-017-3413-3). C1orf116 is an unnamed gene of unknown function that remains largely uncharacterized in any disease state, including cancer or EMT. To substantiate its potential role in EMT, we queried the Cancer Cell Line Encyclopedia dataset (GSE36133) and found that C1orf116 expression is associated with classical epithelial markers and inversely associated with mesenchymal markers. To specifically interrogate the role of C1orf116 in EMT, we overexpressed C1orf116 in otherwise C1orf116low PC3 cells (PC3-oe). PC3-oe cells demonstrated distinctive cobblestone epithelial cellular morphology. Correspondingly, PC3-oe cultures showed increased epithelial marker expression (including OVOL1, CDH1, ESRP), indicating that C1orf116 expression was sufficient to induce an epithelial transition. To determine whether C1orf116 influenced migration, we performed a scratch-closure assay. PC3-oe cultures showed reduced motility compared to control, suggesting that C1orf116 restricts motility. Cellular migration is influenced by many factors, including cell-ECM adhesion, “go vs. grow” switching, and cell-cell adhesion. We observed that PC-oe established cell-cell contacts more rapidly than cells in the control cultures (time-lapse photography immediately following culture seeding) and hypothesized that C1orf116 promoted cell-cell adhesion. We found that PC3-oe had increased expression of tight junction complexes (including ZO1, ZO2, ZO3). Likewise, siRNA-mediated knockdown of C1orf116 expression in C42b (C1orf116high) cells resulted in decreased expression of tight junction proteins. These findings demonstrate that C1orf116 is a novel driver of the epithelial phenotype. In particular, C1orf116 expression promotes cell-cell tight junctions, thereby restricting the active emigrant phenotype of potentially lethal metastatic cells. Citation Format: Sarah R. Amend, James Hernandez, Princy Parsana, Kenneth J. Pienta. C1orf116, a gene of unknown function, promotes tight junctions and epithelial cell phenotype 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 A037.

A precision therapeutic strategy for hexokinase 1-null, hexokinase 2-positive cancers

BackgroundPrecision medicine therapies require identification of unique molecular cancer characteristics. Hexokinase (HK) activity has been proposed as a therapeutic target; however, different hexokinase isoforms have not been well characterized as alternative targets. While HK2 is highly expressed in the majority of cancers, cancer subtypes with differential HK1 and HK2 expression have not been characterized for their sensitivities to HK2 silencing.MethodsHK1 and HK2 expression in the Cancer Cell Line Encyclopedia dataset was analyzed. A doxycycline-inducible shRNA silencing system was used to examine the effect of HK2 knockdown in cultured cells and in xenograft models of HK1−HK2+ and HK1+HK2+ cancers. Glucose consumption and lactate production rates were measured to monitor HK activity in cell culture, and 18F-FDG PET/CT was used to monitor HK activity in xenograft tumors. A high-throughput screen was performed to search for synthetically lethal compounds in combination with HK2 inhibition in HK1−HK2+ liver cancer cells, and a combination therapy for liver cancers with this phenotype was developed. A metabolomic analysis was performed to examine changes in cellular energy levels and key metabolites in HK1−HK2+ cells treated with this combination therapy. The CRISPR Cas9 method was used to establish isogenic HK1+HK2+ and HK1−HK2+ cell lines to evaluate HK1−HK2+ cancer cell sensitivity to the combination therapy.ResultsMost tumors express both HK1 and HK2, and subsets of cancers from a wide variety of tissues of origin express only HK2. Unlike HK1+HK2+ cancers, HK1−HK2+ cancers are sensitive to HK2 silencing-induced cytostasis. Synthetic lethality was achieved in HK1−HK2+ liver cancer cells, by the combination of DPI, a mitochondrial complex I inhibitor, and HK2 inhibition, in HK1−HK2+ liver cancer cells. Perhexiline, a fatty acid oxidation inhibitor, further sensitizes HK1−HK2+ liver cancer cells to the complex I/HK2-targeted therapeutic combination. Although HK1+HK2+ lung cancer H460 cells are resistant to this therapeutic combination, isogenic HK1KOHK2+ cells are sensitive to this therapy.ConclusionsThe HK1−HK2+ cancer subsets exist among a wide variety of cancer types. Selective inhibition of the HK1−HK2+ cancer cell-specific energy production pathways (HK2-driven glycolysis, oxidative phosphorylation and fatty acid oxidation), due to the unique presence of only the HK2 isoform, appears promising to treat HK1−HK2+ cancers. This therapeutic strategy will likely be tolerated by most normal tissues, where only HK1 is expressed.

Discordancy Partitioning for Validating Potentially Inconsistent Pharmacogenomic Studies

The Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Cell Line Encyclopedia (CCLE) are two major studies that can be used to mine for therapeutic biomarkers for cancers of a large variety. Model validation using the two datasets however has proved challenging. Both predictions and signatures do not consistently validate well for models built on one dataset and tested on the other. While the genomic profiling seems consistent, the drug response data is not. Some efforts at harmonizing experimental designs has helped but not entirely removed model validation difficulties. In this paper, we present a partitioning strategy based on a data sharing concept which directly acknowledges a potential lack of concordance between datasets and in doing so, also allows for extraction of reproducible novel gene-drug interaction signatures as well as accurate test set predictions. We demonstrate these properties in a re-analysis of the GDSC and CCLE datasets.

Association of long-chain acyl-coenzyme A synthetase5 expression in human breast cancer by estrogen receptor status and its clinical significance.

The lipid metabolic enzymes are considered candidate therapeutic targets for breast cancer. Long-chain acyl-coenzyme A (CoA) synthase (ACSL) is one of lipid metabolic enzymes and converts free-fatty acid to fatty acid-CoA. Five ACSL isoforms including ACSL1, ACSL3, ACSL4, ACSL5 and ACSL6 are identified in human. High ACSL4 expression has been observed in aggressive breast cancer phenotype. However, the role of other isoforms is still little-known. We therefore, analyzed the expression of ACSL isoforms in each subtype of breast cancer within METABRIC dataset and cancer cell line encyclopedia dataset. The expression levels of ACSL1, ACSL4 and ACSL5 in estrogen receptor (ER)-negative group were higher than that in ER-positive group. Similar expression pattern was detected among breast cancer cell lines MCF-7 (ER-positive) and MDA-MB-231 (ER-negative). Treatment of ACSL inhibitor triacsinC which inhibited enzyme activity of ACSL 1, 3, 4 and 5 suppressed cell growth of MCF-7 and MDA-MB-231. Our results further showed that high ACSL5 expression was associated with good prognosis in patients with both ER-positive and ER-negative breast cancer through KM plotter analysis. These results suggest that ACSL1, ACSL4 and ACSL5 expression is regulated by ER signaling pathways and ACSL5 is a potential novel biomarker for predicting prognosis of breast cancer patients.

Abstract 4506: Identification of druggable transcriptional outliers in urothelial carcinoma

Abstract Background: Urothelial carcinoma (UC) is a major public health problem affecting new 75,000 patients and resulting in 15,000 deaths annually. Despite a variety of genetic and epigenetic alterations described in UC, few therapeutically exploitable drivers have been identified and validated. We hypothesized that transcriptional outliers from primary tumor biopsies could reveal such therapeutically exploitable weaknesses. Identification of these druggable outliers would be critical for individualized therapy of patients with UC. Methods: In order to identify transcriptional outliers we calculated Z-scores for a set of 72 druggable cancer-related genes from RNA sequencing data from our Institute of Precision Medicine (IPM) cohort of advanced platinum-resistant UC. We performed similar outlier analyses on The Cancer Genome Atlas (TCGA) cohort of UC and the Cancer Cell Line Encyclopedia (CCLE) database of urothelial cell lines. Common outliers between IPM and CCLE datasets or between TCGA and CCLE datasets were nominated for further in-vitro testing. Sensitivity scores of cell lines for 130 drugs were extracted from the publicly available cell line browser database and the variance these sensitivity scores across (640) cell lines was calculated. The drugs targeting the outlier genes (real sensitivity z-score) were compared to the mean z-scores of randomly chosen drugs associated with those genes (control sensitive z-score). To investigate the mechanisms underlying comparison of outliers methylation patterns of the promoter regions of outlier genes were examined. Results: 6 druggable outlier genes were common between IPM and CCLE cohorts, and 15 outlier genes between CCLE and TCGA cohorts. Common outliers included FGFR3, KIT, BCL2, ABL1, and SMO. The presence of outlier genes in cell lines correlated with higher sensitivity to drugs predicted to target these genes in comparison to randomly selected drugs (p<0.05). In vitro cell viability assays confirmed sensitivity of RT112 and RT4 cell lines to Dovitinib, an FGFR3 inhibitor as predicted by our algorithm. Outlier genes were frequently involved in gene fusions, amplification or DNA hypomethylation (p<0.02). Conclusions: Identification of druggable transcriptional outliers can elucidate novel oncogenic dependencies and aid in selecting precision therapies for UC patients. Citation Format: Bishoy S. M. Faltas, Rohan Bareja, Ethan Shelkey, Rebecca Mayer, Andrea Sboner, Mark A. Rubin, Olivier Elemento. Identification of druggable transcriptional outliers in urothelial carcinoma. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 4506.

Abstract 1520: Identifying differential dependency networks accounting for response to NEDD8-inhibitor in large-scale cancer cell line data

Abstract The Cancer Cell Line Encyclopedia (CCLE) houses molecular profiles of ∼800 human long term cell lines spanning several different histological types of cancer. In addition, Cancer Therapeutics Response Portal (CTRP) provides drug response measurements for 481 small molecules. Integration of these data enables investigation of the molecular correlates of drug response (sensitivity and resistance). In this current effort, we studied the NEDDylation small molecule inhibitor, MLN4924, in the context of genomic data to uncover novel mechanistic correlates of drug response across the panel of cell lines. We recently reported (Jung and Kim 2016 NAR) development of a robust computational method that shows promise to identify novel insights when applied to multi-dimensional data sets as outlined above. The Evaluation of Differential Dependency (EDDY) employs Bayesian networks to represent statistically distinct differences in relationships between genes within a specific biological pathway as queried between two conditions, in this instance, cell lines that are sensitive and those that are non-sensitive to MLN4924. While EDDY has been successfully employed in the analysis of specific diseases such as TCGA adrenocortical carcinoma, its statistical rigor incurred a prohibitive computational load to assess conditional differences across larger datasets. Recent computational enhancements to EDDY enable processing of larger datasets in reasonable time while maintaining sensitivity. The capability of analyzing broader pan-cancer datasets such as CCLE has enabled EDDY to become more capable in identifying general trends across disease subtypes. Specifically, we demonstrate the enhanced EDDY in analysis of MLN4924 response across the CCLE data set combined with CTRP data set. Initial outcomes from EDDY point to both anticipated and unanticipated biological determinants of response. For example, it is noted that specific oncogenic pathways, such as those centered on PIK3CA, appear to show differential dependencies in the sensitive and non-sensitive cell lines. We also observe genes and candidate pathways related to apoptotic mechanisms that may reveal mechanistic insights to predicting drug response. Specifically, genes and pathways associated with certain apoptotic mechanisms around mitochondrial proteins and glutathione peroxidase may serve as unique determinants of drug response. Multidimensional data analyzed by EDDY uncovers candidate mechanisms of vulnerability to specific small molecule inhibitors, which may guide development of predictive models for treatment planning when using agents with highly context-dependent efficacies. Supported by NIH U01CA168397 Citation Format: Gil Speyer, Harshil Dhruv, Jeff Kiefer, Stuart Schreiber, Paul Clemons, Michael E. Berens, Seungchan Kim. Identifying differential dependency networks accounting for response to NEDD8-inhibitor in large-scale cancer cell line data. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 1520.

MicroRNA-375/SEC23A as biomarkers of the in vitro efficacy of vandetanib.

In this study, we performed microRNA (miRNA) expression profiling on a large series of sporadic and hereditary forms of medullary thyroid carcinomas (MTC). More than 60 miRNAs were significantly deregulated in tumor vs adjacent non-tumor tissues, partially overlapping with results of previous studies. We focused our attention on the strongest up-regulated miRNA in MTC samples, miR-375, the deregulation of which has been previously observed in a variety of human malignancies including MTC. We identified miR-375 targets by combining gene expression signatures from human MTC (TT) and normal follicular (Nthy-ori 3-1) cell lines transfected with an antagomiR-375 inhibitor or a miR-375 mimic, respectively, and from an in silico analysis of thyroid cell lines of Cancer Cell Line Encyclopedia datasets. This approach identified SEC23A as a bona fide miR-375 target, which we validated by immunoblotting and immunohistochemistry of non-tumor and pathological thyroid tissue. Furthermore, we observed that miR-375 overexpression was associated with decreased cell proliferation and synergistically increased sensitivity to vandetanib, the clinically relevant treatment of metastatic MTC. We found that miR-375 increased PARP cleavage and decreased AKT phosphorylation, affecting both cell proliferation and viability. We confirmed these results through SEC23A direct silencing in combination with vandetanib, highlighting the importance of SEC23A in the miR-375-associated increased sensitivity to vandetanib.Since the combination of increased expression of miR-375 and decreased expression of SEC23A point to sensitivity to vandetanib, we question if the expression levels of miR-375 and SEC23A should be evaluated as an indicator of eligibility for treatment of MTC patients with vandetanib.

Impact of centralization on aCGH-based genomic profiles for precision medicine in oncology

Comparative genomic hybridization (CGH) arrays are increasingly used in personalized medicine programs to identify gene copy number aberrations (CNAs) that may be used to guide clinical decisions made during molecular tumor boards. However, analytical processes such as the centralization step may profoundly affect CGH array results and therefore may adversely affect outcomes in the precision medicine context. The effect of three different centralization methods: median, maximum peak, alternative peak, were evaluated on three datasets: (i) the NCI60 cell lines panel, (ii) the Cancer Cell Line Encyclopedia (CCLE) panel, and (iii) the patients enrolled in prospective molecular screening trials (SAFIR-01 n = 283, MOSCATO-01 n = 309), and compared with karyotyping, drug sensitivity, and patient-drug matching, respectively. Using the NCI60 cell lines panel, the profiles generated by the alternative peak method were significantly closer to the cell karyotypes than those generated by the other centralization strategies (P < 0.05). Using the CCLE dataset, selected genes (ERBB2, EGFR) were better or equally correlated to the IC50 of their companion drug (lapatinib, erlotinib), when applying the alternative centralization. Finally, focusing on 24 actionable genes, we observed as many as 7.1% (SAFIR-01) and 6.8% (MOSCATO-01) of patients originally not oriented to a specific treatment, but who could have been proposed a treatment based on the alternative peak centralization method. The centralization method substantially affects the call detection of CGH profiles and may thus impact precision medicine approaches. Among the three methods described, the alternative peak method addresses limitations associated with existing approaches.

Design of Probabilistic Random Forests with Applications to Anticancer Drug Sensitivity Prediction.

Random forests consisting of an ensemble of regression trees with equal weights are frequently used for design of predictive models. In this article, we consider an extension of the methodology by representing the regression trees in the form of probabilistic trees and analyzing the nature of heteroscedasticity. The probabilistic tree representation allows for analytical computation of confidence intervals (CIs), and the tree weight optimization is expected to provide stricter CIs with comparable performance in mean error. We approached the ensemble of probabilistic trees’ prediction from the perspectives of a mixture distribution and as a weighted sum of correlated random variables. We applied our methodology to the drug sensitivity prediction problem on synthetic and cancer cell line encyclopedia dataset and illustrated that tree weights can be selected to reduce the average length of the CI without increase in mean error.

Pattern discovery and cancer gene identification in integrated cancer genomic data

Large-scale integrated cancer genome characterization efforts including the cancer genome atlas and the cancer cell line encyclopedia have created unprecedented opportunities to study cancer biology in the context of knowing the entire catalog of genetic alterations. A clinically important challenge is to discover cancer subtypes and their molecular drivers in a comprehensive genetic context. Curtis et al. [Nature (2012) 486(7403):346-352] has recently shown that integrative clustering of copy number and gene expression in 2,000 breast tumors reveals novel subgroups beyond the classic expression subtypes that show distinct clinical outcomes. To extend the scope of integrative analysis for the inclusion of somatic mutation data by massively parallel sequencing, we propose a framework for joint modeling of discrete and continuous variables that arise from integrated genomic, epigenomic, and transcriptomic profiling. The core idea is motivated by the hypothesis that diverse molecular phenotypes can be predicted by a set of orthogonal latent variables that represent distinct molecular drivers, and thus can reveal tumor subgroups of biological and clinical importance. Using the cancer cell line encyclopedia dataset, we demonstrate our method can accurately group cell lines by their cell-of-origin for several cancer types, and precisely pinpoint their known and potential cancer driver genes. Our integrative analysis also demonstrates the power for revealing subgroups that are not lineage-dependent, but consist of different cancer types driven by a common genetic alteration. Application of the cancer genome atlas colorectal cancer data reveals distinct integrated tumor subtypes, suggesting different genetic pathways in colon cancer progression.