CellMiner Cross-Database (CellMinerCDB) version 1.2: Exploration of patient-derived cancer cell line pharmacogenomics.

CellMiner Cross-Database (CellMinerCDB, discover.nci.nih.gov/cellminercdb) allows integration and analysis of molecular and pharmacological data within and across cancer cell line datasets from the National Cancer Institute (NCI), Broad Institute, Sanger/MGH and MD Anderson Cancer Center (MDACC). We present CellMinerCDB 1.2 with updates to datasets from NCI-60, Broad Cancer Cell Line Encyclopedia and Sanger/MGH, and the addition of new datasets, including NCI-ALMANAC drug combination, MDACC Cell Line Project proteomic, NCI-SCLC DNA copy number and methylation data, and Broad methylation, genetic dependency and metabolomic datasets. CellMinerCDB (v1.2) includes several improvements over the previously published version: (i) new and updated datasets; (ii) support for pattern comparisons and multivariate analyses across data sources; (iii) updated annotations with drug mechanism of action information and biologically relevant multigene signatures; (iv) analysis speedups via caching; (v) a new dataset download feature; (vi) improved visualization of subsets of multiple tissue types; (vii) breakdown of univariate associations by tissue type; and (viii) enhanced help information. The curation and common annotations (e.g. tissues of origin and identifiers) provided here across pharmacogenomic datasets increase the utility of the individual datasets to address multiple researcher question types, including data reproducibility, biomarker discovery and multivariate analysis of drug activity.

Oncogenic driver genes and the inflammatory microenvironment dictate liver tumor phenotype.

UNLABELLED: The majority of hepatocellular carcinoma develops in the background of chronic liver inflammation caused by viral hepatitis and alcoholic or nonalcoholic steatohepatitis. However, the impact of different types of chronic inflammatory microenvironments on the phenotypes of tumors generated by distinct oncogenes is largely unresolved. To address this issue, we generated murine liver tumors by constitutively active AKT-1 (AKT) and ß-catenin (CAT), followed by induction of chronic liver inflammation by 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC) and carbon tetrachloride. Also, the impact of DDC-induced chronic liver inflammation was compared between two liver tumor models using a combination of AKT-CAT or AKT-NRAS(G12V) . Treatment with DDC and carbon tetrachloride significantly facilitated the adenoma-to-carcinoma conversion and accelerated the growth of AKT-CAT tumors. Furthermore, DDC treatment altered the morphology of AKT-CAT tumors and caused loss of lipid droplets. Transcriptome analysis of AKT-CAT tumors revealed that cellular growth and proliferation were mainly affected by chronic inflammation and caused up-regulation of Cxcl16, Galectin-3, and Nedd9, among others. Integration with transcriptome profiles from human hepatocellular carcinomas further demonstrated that AKT-CAT tumors generated in the context of chronic liver inflammation showed enrichment of poor prognosis gene sets or decrease of good prognosis gene sets. In contrast, DDC had a more subtle effect on AKT-NRAS(G12V) tumors and primarily enhanced already existent tumor characteristics as supported by transcriptome analysis. However, it also reduced lipid droplets in AKT-NRAS(G12V) tumors. CONCLUSION: Our study suggests that liver tumor phenotype is defined by a combination of driving oncogenes but also the nature of chronic liver inflammation. (Hepatology 2016;63:1888-1899).

Sarcoma_CellminerCDB: A tool to interrogate the genomic and functional characteristics of a comprehensive collection of sarcoma cell lines.

Sarcomas are a diverse group of rare malignancies composed of multiple different clinical and molecular subtypes. Due to their rarity and heterogeneity, basic, translational, and clinical research in sarcoma has trailed behind that of other cancers. Outcomes for patients remain generally poor due to an incomplete understanding of disease biology and a lack of novel therapies. To address some of the limitations impeding preclinical sarcoma research, we have developed Sarcoma_CellMinerCDB, a publicly available interactive tool that merges publicly available sarcoma cell line data and newly generated omics data to create a comprehensive database of genomic, transcriptomic, methylomic, proteomic, metabolic, and pharmacologic data on 133 annotated sarcoma cell lines. The reproducibility, functionality, biological relevance, and therapeutic applications of Sarcoma_CellMinerCDB described herein are powerful tools to address and generate biological questions and test hypotheses for translational research. Sarcoma_CellMinerCDB (https://discover.nci.nih.gov/SarcomaCellMinerCDB) aims to contribute to advancing the preclinical study of sarcoma.

A Database Tool Integrating Genomic and Pharmacologic Data from Adrenocortical Carcinoma Cell Lines, PDX, and Patient Samples.

Adrenocortical carcinoma (ACC) is a rare and highly heterogeneous disease with a notably poor prognosis due to significant challenges in diagnosis and treatment. Emphasizing on the importance of precision medicine, there is an increasing need for comprehensive genomic resources alongside well-developed experimental models to devise personalized therapeutic strategies. We present ACC_CellMinerCDB, a substantive genomic and drug sensitivity database (available at https://discover.nci.nih.gov/acc_cellminercdb) comprising ACC cell lines, patient-derived xenografts, surgical samples, and responses to more than 2,400 drugs examined by the NCI and National Center for Advancing Translational Sciences. This database exposes shared genomic pathways among ACC cell lines and surgical samples, thus authenticating the cell lines as research models. It also allows exploration of pertinent treatment markers such as MDR-1, SOAT1, MGMT, MMR, and SLFN11 and introduces the potential to repurpose agents like temozolomide for ACC therapy. ACC_CellMinerCDB provides the foundation for exploring larger preclinical ACC models. SIGNIFICANCE: ACC_CellMinerCDB, a comprehensive database of cell lines, patient-derived xenografts, surgical samples, and drug responses, reveals shared genomic pathways and treatment-relevant markers in ACC. This resource offers insights into potential therapeutic targets and the opportunity to repurpose existing drugs for ACC therapy.

Activity of the Ubiquitin-activating Enzyme Inhibitor TAK-243 in Adrenocortical Carcinoma Cell Lines, Patient-derived Organoids, and Murine Xenografts.

Current treatment options for metastatic adrenocortical carcinoma (ACC) have limited efficacy, despite the common use of mitotane and cytotoxic agents. This study aimed to identify novel therapeutic options for ACC. An extensive drug screen was conducted to identify compounds with potential activity against ACC cell lines. We further investigated the mechanism of action of the identified compound, TAK-243, its synergistic effects with current ACC therapeutics, and its efficacy in ACC models including patient-derived organoids and mouse xenografts. TAK-243, a clinical ubiquitin-activating enzyme (UAE) inhibitor, showed potent activity in ACC cell lines. TAK-243 inhibited protein ubiquitination in ACC cells, leading to the accumulation of free ubiquitin, activation of the unfolded protein response, and induction of apoptosis. TAK-243 was found to be effluxed out of cells by MDR1, a drug efflux pump, and did not require Schlafen 11 (SLFN11) expression for its activity. Combination of TAK-243 with current ACC therapies (e.g., mitotane, etoposide, cisplatin) produced synergistic or additive effects. In addition, TAK-243 was highly synergistic with BCL2 inhibitors (Navitoclax and Venetoclax) in preclinical ACC models including patient-derived organoids. The tumor suppressive effects of TAK-243 and its synergistic effects with Venetoclax were further confirmed in a mouse xenograft model. These findings provide preclinical evidence to support the initiation of a clinical trial of TAK-243 in patients with advanced-stage ACC. TAK-243 is a promising potential treatment option for ACC, either as monotherapy or in combination with existing therapies or BCL2 inhibitors. SIGNIFICANCE: ACC is a rare endocrine cancer with poor prognosis and limited therapeutic options. We report that TAK-243 is active alone and in combination with currently used therapies and with BCL2 and mTOR inhibitors in ACC preclinical models. Our results suggest implementation of TAK-243 in clinical trials for patients with advanced and metastatic ACC.

CellMinerCDB: NCATS Is a Web-Based Portal Integrating Public Cancer Cell Line Databases for Pharmacogenomic Explorations.

Major advances have been made in the field of precision medicine for treating cancer. However, many open questions remain that need to be answered to realize the goal of matching every patient with cancer to the most efficacious therapy. To facilitate these efforts, we have developed CellMinerCDB: National Center for Advancing Translational Sciences (NCATS; https://discover.nci.nih.gov/rsconnect/cellminercdb_ncats/), which makes available activity information for 2,675 drugs and compounds, including multiple nononcology drugs and 1,866 drugs and compounds unique to the NCATS. CellMinerCDB: NCATS comprises 183 cancer cell lines, with 72 unique to NCATS, including some from previously understudied tissues of origin. Multiple forms of data from different institutes are integrated, including single and combination drug activity, DNA copy number, methylation and mutation, transcriptome, protein levels, histone acetylation and methylation, metabolites, CRISPR, and miscellaneous signatures. Curation of cell lines and drug names enables cross-database (CDB) analyses. Comparison of the datasets is made possible by the overlap between cell lines and drugs across databases. Multiple univariate and multivariate analysis tools are built-in, including linear regression and LASSO. Examples have been presented here for the clinical topoisomerase I (TOP1) inhibitors topotecan and irinotecan/SN-38. This web application provides both substantial new data and significant pharmacogenomic integration, allowing exploration of interrelationships. SIGNIFICANCE: CellMinerCDB: NCATS provides activity information for 2,675 drugs in 183 cancer cell lines and analysis tools to facilitate pharmacogenomic research and to identify determinants of response.

UGDH promotes tumor-initiating cells and a fibroinflammatory tumor microenvironment in ovarian cancer.

BACKGROUND: Epithelial ovarian cancer (EOC) is a global health burden, with the poorest five-year survival rate of the gynecological malignancies due to diagnosis at advanced stage and high recurrence rate. Recurrence in EOC is driven by the survival of chemoresistant, stem-like tumor-initiating cells (TICs) that are supported by a complex extracellular matrix and immunosuppressive microenvironment. To target TICs to prevent recurrence, we identified genes critical for TIC viability from a whole genome siRNA screen. A top hit was the cancer-associated, proteoglycan subunit synthesis enzyme UDP-glucose dehydrogenase (UGDH). METHODS: Immunohistochemistry was used to characterize UGDH expression in histological and molecular subtypes of EOC. EOC cell lines were subtyped according to the molecular subtypes and the functional effects of modulating UGDH expression in vitro and in vivo in C1/Mesenchymal and C4/Differentiated subtype cell lines was examined. RESULTS: High UGDH expression was observed in high-grade serous ovarian cancers and a distinctive survival prognostic for UGDH expression was revealed when serous cancers were stratified by molecular subtype. High UGDH was associated with a poor prognosis in the C1/Mesenchymal subtype and low UGDH was associated with poor prognosis in the C4/Differentiated subtype. Knockdown of UGDH in the C1/mesenchymal molecular subtype reduced spheroid formation and viability and reduced the CD133 + /ALDH high TIC population. Conversely, overexpression of UGDH in the C4/Differentiated subtype reduced the TIC population. In co-culture models, UGDH expression in spheroids affected the gene expression of mesothelial cells causing changes to matrix remodeling proteins, and fibroblast collagen production. Inflammatory cytokine expression of spheroids was altered by UGDH expression. The effect of UGDH knockdown or overexpression in the C1/ Mesenchymal and C4/Differentiated subtypes respectively was tested on mouse intrabursal xenografts and showed dynamic changes to the tumor stroma. Knockdown of UGDH improved survival and reduced tumor burden in C1/Mesenchymal compared to controls. CONCLUSIONS: These data show that modulation of UGDH expression in ovarian cancer reveals distinct roles for UGDH in the C1/Mesenchymal and C4/Differentiated molecular subtypes of EOC, influencing the tumor microenvironmental composition. UGDH is a strong potential therapeutic target in TICs, for the treatment of EOC, particularly in patients with the mesenchymal molecular subtype.

New Insights on the Genetics of Pheochromocytoma and Paraganglioma and Its Clinical Implications.

Pheochromocytomas (PHEOs) and paragangliomas (PGLs) are rare neuroendocrine tumors that arise from chromaffin cells. PHEOs arise from the adrenal medulla, whereas PGLs arise from the neural crest localized outside the adrenal gland. Approximately 40% of all cases of PPGLs (pheochromocytomas/paragangliomas) are associated with germline mutations and 30-40% display somatic driver mutations. The mutations associated with PPGLs can be classified into three groups. The pseudohypoxic group or cluster I includes the following genes: SDHA, SDHB, SDHC, SDHD, SDHAF2, FH, VHL, IDH1/2, MHD2, EGLN1/2 and HIF2/EPAS; the kinase group or cluster II includes RET, NF1, TMEM127, MAX and HRAS; and the Wnt signaling group or cluster III includes CSDE1 and MAML3. Underlying mutations can help understand the clinical presentation, overall prognosis and surveillance follow-up. Here we are discussing the new genetic insights of PPGLs.

Integrative epigenomic analyses of small cell lung cancer cells demonstrates the clinical translational relevance of gene body methylation.

DNA methylation is a key regulator of gene expression and a clinical therapeutic predictor. We examined global DNA methylation beyond the generally used promoter areas in human small cell lung cancer (SCLC) and find that gene body methylation is a robust positive predictor of gene expression. Combining promoter and gene body methylation better predicts gene expression than promoter methylation alone including genes involved in the neuroendocrine classification of SCLC and the expression of therapeutically relevant genes including MGMT, SLFN11, and DLL3. Importantly, for super-enhancer (SE) covered genes such as NEUROD1 or MYC, using H3K27ac and NEUROD1, ASCL1, and POU2F3 ChIP-seq data, we show that genic methylation is inversely proportional to expression, thus providing a new approach to identify potential SE regulated genes involved in SCLC pathogenesis. To advance SCLC transitional research, these data are integrated into our web portal (https://discover.nci.nih.gov/SclcCellMinerCDB/) for open and easy access to basic and clinical investigators.

Novel and Highly Potent ATR Inhibitor M4344 Kills Cancer Cells With Replication Stress, and Enhances the Chemotherapeutic Activity of Widely Used DNA Damaging Agents.

Although several ATR inhibitors are in development, there are unresolved questions regarding their differential potency, molecular signatures of patients with cancer for predicting activity, and most effective therapeutic combinations. Here, we elucidate how to improve ATR-based chemotherapy with the newly developed ATR inhibitor, M4344 using in vitro and in vivo models. The potency of M4344 was compared with the clinically developed ATR inhibitors BAY1895344, berzosertib, and ceralasertib. The anticancer activity of M4344 was investigated as monotherapy and combination with clinical DNA damaging agents in multiple cancer cell lines, patient-derived tumor organoids, and mouse xenograft models. We also elucidated the anticancer mechanisms and potential biomarkers for M4344. We demonstrate that M4344 is highly potent among the clinically developed ATR inhibitors. Replication stress (RepStress) and neuroendocrine (NE) gene expression signatures are significantly associated with a response to M4344 treatment. M4344 kills cancer cells by inducing cellular catastrophe and DNA damage. M4344 is highly synergistic with a broad range of DNA-targeting anticancer agents. It significantly synergizes with topotecan and irinotecan in patient-derived tumor organoids and xenograft models. Taken together, M4344 is a promising and highly potent ATR inhibitor. It enhances the activity of clinical DNA damaging agents commonly used in cancer treatment including topoisomerase inhibitors, gemcitabine, cisplatin, and talazoparib. RepStress and NE gene expression signatures can be exploited as predictive markers for M4344.

Epigenetic suppression of SLFN11 in germinal center B-cells during B-cell development.

BACKGROUND: SLFN11 has recently been reported to execute cancer cells harboring replicative stress induced by DNA damaging agents. However, the roles of SLFN11 under physiological conditions remain poorly understood. Germinal center B-cells (GCBs) undergo somatic hypermutations and class-switch recombination, which can cause physiological genotoxic stress. Hence, we tested whether SLFN11 expression needs to be suppressed in GCBs during B-cell development. OBJECTIVE: To clarify the expression profile of SLFN11 in different developmental stages of B-cells and B-cell-derived cancers. METHODS: We analyzed the expression of SLFN11 by mining cell line databases for different stages of normal B-cells and various types of B-cell-derived cancer cell lines. We performed dual immunohistochemical staining for SLFN11 and B-cell specific markers in normal human lymphatic tissues. We tested the effects of two epigenetic modifiers, an EZH2 inhibitor, tazemetostat (EPZ6438) and a histone deacetylase inhibitor, panobinostat (LBH589) on SLFN11 expression in GCB-derived lymphoma cell lines. We also examined the therapeutic efficacy of these drugs in combination with cytosine arabinoside and the effects of SLFN11 on the efficacy of cytosine arabinoside in SLFN11-overexpressing cells. RESULTS: SLFN11 mRNA level was found low in both normal GCBs and GCB-DLBCL (GCB like-diffuse large B-cell lymphoma). Immunohistochemical staining showed low SLFN11 expression in GCBs and high SLFN11 expression in plasmablasts and plasmacytes. The EZH2 and HDAC epigenetic modifiers upregulated SLFN11 expression in GCB-derived lymphoma cells and made them more susceptible to cytosine arabinoside. SLFN11 overexpression further sensitized GCB-derived lymphoma cells to cytosine arabinoside. CONCLUSIONS: The expression of SLFN11 is epigenetically suppressed in normal GCBs and GCB-derived lymphomas. GCB-derived lymphomas with low SLFN11 expression can be treated by the combination of epigenetic modifiers and cytosine arabinoside.

Candidate biomarker assessment for pharmacological response.

Using the information from our CellMiner (https://discover.nci.nih.gov/cellminer/) and CellMinerCDB (https://discover.nci.nih.gov/cellminercdb/) web-based applications, we identified 3978 molecular events with significant links to pharmacological response for genes that are either targets, biomarkers, or have established causal linkage to drugs. Molecular events included DNA copy number, methylation and mutation; and transcript; and whole or phospho-protein expression for the NCI-60 human cancer cell lines. While all forms of molecular data were informative in some (gene-drug) pairings, the type of significantly linked molecular events was found to vary widely by drug. Some forms of molecular data were found to have more frequent significant correlation than others. Leading were phosphoproteins as measured by antibody (31%), followed by transcript as measured by microarray (16%), and total protein levels as measured by mass spectrometry or antibody (14%). All other measurements ranged between 5 and 11%. Data reliability was underscored by concordant results when using differing drugs with the same targets, as well as different measurements of the same molecular parameter. The significance of correlations of the various molecular parameters to the pharmacological responses provides functional indication of those parameters that are biologically relevant for each gene-drug pairing, as well as comparisons between measurement types.

TREML4 Promotes Inflammatory Programs in Human and Murine Macrophages and Alters Atherosclerosis Lesion Composition in the Apolipoprotein E Deficient Mouse.

The Triggering Receptor Expressed on Myeloid cells-like 4 (TREML4) is a member of the TREM receptor family, known modulators of inflammatory responses. We have previously found that TREML4 expression positively correlates with human coronary arterial calcification (CAC). However, the role of TREML4 in the pathogenesis of cardiovascular disease remains incompletely defined. Since macrophages play a key role in inflammatory conditions, we investigated if activated macrophages selectively expressed TREML4 and found that carriage of either one of the eQTL SNP's previously associated with increased TREML4 expression conferred higher expression in human inflammatory macrophages (M1) compared to alternatively activated macrophages (M2). Furthermore, we found that TREML4 expression in human M1 dysregulated several inflammatory pathways related to leukocyte activation, apoptosis and extracellular matrix degradation. Similarly, murine M1 expressed substantial levels of Treml4, as did oxLDL treated macrophages. Transcriptome analysis confirmed that murine Treml4 controls the expression of genes related to inflammation and lipid regulation pathways, suggesting a possible role in atherosclerosis. Analysis of Apoe-/-/Treml4-/- mice showed reduced plaque burden and lesion complexity as indicated by decreased stage scores, macrophage content and collagen deposition. Finally, transcriptome analysis of oxLDL-loaded murine macrophages showed that Treml4 represses a specific set of genes related to carbohydrate, ion and amino acid membrane transport. Metabolomic analysis confirmed that Treml4 deficiency may promote a beneficial relationship between iron homeostasis and glucose metabolism. Together, our results suggest that Treml4 plays a role in the development of cardiovascular disease, as indicated by Treml4-dependent dysregulation of macrophage inflammatory pathways, macrophage metabolism and promotion of vulnerability features in advanced lesions.

SCLC-CellMiner: A Resource for Small Cell Lung Cancer Cell Line Genomics and Pharmacology Based on Genomic Signatures.

CellMiner-SCLC (https://discover.nci.nih.gov/SclcCellMinerCDB/) integrates drug sensitivity and genomic data, including high-resolution methylome and transcriptome from 118 patient-derived small cell lung cancer (SCLC) cell lines, providing a resource for research into this "recalcitrant cancer." We demonstrate the reproducibility and stability of data from multiple sources and validate the SCLC consensus nomenclature on the basis of expression of master transcription factors NEUROD1, ASCL1, POU2F3, and YAP1. Our analyses reveal transcription networks linking SCLC subtypes with MYC and its paralogs and the NOTCH and HIPPO pathways. SCLC subsets express specific surface markers, providing potential opportunities for antibody-based targeted therapies. YAP1-driven SCLCs are notable for differential expression of the NOTCH pathway, epithelial-mesenchymal transition (EMT), and antigen-presenting machinery (APM) genes and sensitivity to mTOR and AKT inhibitors. These analyses provide insights into SCLC biology and a framework for future investigations into subtype-specific SCLC vulnerabilities.

RNA Sequencing of the NCI-60: Integration into CellMiner and CellMiner CDB.

CellMiner (http://discover.nci.nih.gov/cellminer) and CellMinerCDB (https://discover.nci.nih.gov/cellminercdb/) are web-based applications for mining publicly available genomic, molecular, and pharmacologic datasets of human cancer cell lines including the NCI-60, Cancer Cell Line Encyclopedia, Genomics of Drug Sensitivity in Cancer, Cancer Therapeutics Response Portal, NCI/DTP small cell lung cancer, and NCI Almanac cell line sets. Here, we introduce our RNA sequencing (RNA-seq) data for the NCI-60 and their access and integration with the other databases. Correlation to transcript microarray expression levels for identical genes and identical cell lines across CellMinerCDB demonstrates the high quality of these new RNA-seq data. We provide composite and isoform transcript expression data and demonstrate diversity in isoform composition for individual cancer- and pharmacologically relevant genes, including HRAS, PTEN, EGFR, RAD51, ALKBH2, BRCA1, ERBB2, TP53, FGFR2, and CTNND1. We reveal cell-specific differences in the overall levels of isoforms and show their linkage to expression of RNA processing and splicing genes as well as resultant alterations in cancer and pharmacologic gene sets. Gene-drug pairings linked by pathways or functions show specific correlations to isoforms compared with composite gene expression, including ALKBH2-benzaldehyde, AKT3-vandetanib, BCR-imatinib, CDK1 and 20-palbociclib, CASP1-imexon, and FGFR3-pazopanib. Loss of MUC1 20 amino acid variable number tandem repeats, which is used to elicit immune response, and the presence of the androgen receptor AR-V4 and -V7 isoforms in all NCI-60 tissue of origin types demonstrate translational relevance. In summary, we introduce RNA-seq data to our CellMiner and CellMinerCDB web applications, allowing their exploration for both research and translational purposes. SIGNIFICANCE: The current study provides RNA sequencing data for the NCI-60 cell lines made accessible through both CellMiner and CellMinerCDB and is an important pharmacogenomics resource for the field.

Quantitative Proteome Landscape of the NCI-60 Cancer Cell Lines.

Here we describe a proteomic data resource for the NCI-60 cell lines generated by pressure cycling technology and SWATH mass spectrometry. We developed the DIA-expert software to curate and visualize the SWATH data, leading to reproducible detection of over 3,100 SwissProt proteotypic proteins and systematic quantification of pathway activities. Stoichiometric relationships of interacting proteins for DNA replication, repair, the chromatin remodeling NuRD complex, ß-catenin, RNA metabolism, and prefoldins are more evident than that at the mRNA level. The data are available in CellMiner (discover.nci.nih.gov/cellminercdb and discover.nci.nih.gov/cellminer), allowing casual users to test hypotheses and perform integrative, cross-database analyses of multi-omic drug response correlations for over 20,000 drugs. We demonstrate the value of proteome data in predicting drug response for over 240 clinically relevant chemotherapeutic and targeted therapies. In summary, we present a novel proteome resource for the NCI-60, together with relevant software tools, and demonstrate the benefit of proteome analyses.

A comparison of machine learning algorithms for chemical toxicity classification using a simulated multi-scale data model.

BACKGROUND: Bioactivity profiling using high-throughput in vitro assays can reduce the cost and time required for toxicological screening of environmental chemicals and can also reduce the need for animal testing. Several public efforts are aimed at discovering patterns or classifiers in high-dimensional bioactivity space that predict tissue, organ or whole animal toxicological endpoints. Supervised machine learning is a powerful approach to discover combinatorial relationships in complex in vitro/in vivo datasets. We present a novel model to simulate complex chemical-toxicology data sets and use this model to evaluate the relative performance of different machine learning (ML) methods. RESULTS: The classification performance of Artificial Neural Networks (ANN), K-Nearest Neighbors (KNN), Linear Discriminant Analysis (LDA), Naïve Bayes (NB), Recursive Partitioning and Regression Trees (RPART), and Support Vector Machines (SVM) in the presence and absence of filter-based feature selection was analyzed using K-way cross-validation testing and independent validation on simulated in vitro assay data sets with varying levels of model complexity, number of irrelevant features and measurement noise. While the prediction accuracy of all ML methods decreased as non-causal (irrelevant) features were added, some ML methods performed better than others. In the limit of using a large number of features, ANN and SVM were always in the top performing set of methods while RPART and KNN (k = 5) were always in the poorest performing set. The addition of measurement noise and irrelevant features decreased the classification accuracy of all ML methods, with LDA suffering the greatest performance degradation. LDA performance is especially sensitive to the use of feature selection. Filter-based feature selection generally improved performance, most strikingly for LDA. CONCLUSION: We have developed a novel simulation model to evaluate machine learning methods for the analysis of data sets in which in vitro bioassay data is being used to predict in vivo chemical toxicology. From our analysis, we can recommend that several ML methods, most notably SVM and ANN, are good candidates for use in real world applications in this area.

ACToR--Aggregated Computational Toxicology Resource.

ACToR (Aggregated Computational Toxicology Resource) is a database and set of software applications that bring into one central location many types and sources of data on environmental chemicals. Currently, the ACToR chemical database contains information on chemical structure, in vitro bioassays and in vivo toxicology assays derived from more than 150 sources including the U.S. Environmental Protection Agency (EPA), Centers for Disease Control (CDC), U.S. Food and Drug Administration (FDA), National Institutes of Health (NIH), state agencies, corresponding government agencies in Canada, Europe and Japan, universities, the World Health Organization (WHO) and non-governmental organizations (NGOs). At the EPA National Center for Computational Toxicology, ACToR helps manage large data sets being used in a high-throughput environmental chemical screening and prioritization program called ToxCast.

CellMinerCDB for Integrative Cross-Database Genomics and Pharmacogenomics Analyses of Cancer Cell Lines.

CellMinerCDB provides a web-based resource (https://discover.nci.nih.gov/cellminercdb/) for integrating multiple forms of pharmacological and genomic analyses, and unifying the richest cancer cell line datasets (the NCI-60, NCI-SCLC, Sanger/MGH GDSC, and Broad CCLE/CTRP). CellMinerCDB enables data queries for genomics and gene regulatory network analyses, and exploration of pharmacogenomic determinants and drug signatures. It leverages overlaps of cell lines and drugs across databases to examine reproducibility and expand pathway analyses. We illustrate the value of CellMinerCDB for elucidating gene expression determinants, such as DNA methylation and copy number variations, and highlight complexities in assessing mutational burden. We demonstrate the value of CellMinerCDB in selecting drugs with reproducible activity, expand on the dominant role of SLFN11 for drug response, and present novel response determinants and genomic signatures for topoisomerase inhibitors and schweinfurthins. We also introduce LIX1L as a gene associated with mesenchymal signature and regulation of cellular migration and invasiveness.

SEARCHPATTOOL: a new method for mining the most specific frequent patterns for binding sites with application to prokaryotic DNA sequences.

BACKGROUND: Computational methods to predict transcription factor binding sites (TFBS) based on exhaustive algorithms are guaranteed to find the best patterns but are often limited to short ones or impose some constraints on the pattern type. Many patterns for binding sites in prokaryotic species are not well characterized but are known to be large, between 16-30 base pairs (bp) and contain at least 2 conserved bases. The length of prokaryotic species promoters (about 400 bp) and our interest in studying a small set of genes that could be a cluster of co-regulated genes from microarray experiments led to the development of a new exhaustive algorithm targeting these large patterns. RESULTS: We present Searchpattool, a new method to search for and select the most specific (conservative) frequent patterns. This method does not impose restrictions on the density or the structure of the pattern. The best patterns (motifs) are selected using several statistics, including a new application of a z-score based on the number of matching sequences. We compared Searchpattool against other well known algorithms on a Bacillus subtilis group of 14 input sequences and found that in our experiments Searchpattool always performed the best based on performance scores. CONCLUSION: Searchpattool is a new method for pattern discovery relative to transcription factor binding sites for species or genes with short promoters. It outputs the most specific significant patterns and helps the biologist to choose the best candidates.