Correction: Pannala et al. High-Throughput Transcriptomics Differentiates Toxic versus Non-Toxic Chemical Exposures Using a Rat Liver Model. Int. J. Mol. Sci. 2023, 24, 17425.

In the original publication [...].

Whole exome and transcript profiling of liver following aflatoxin B1 exposure in rats.

We recently developed a rat whole exome sequencing (WES) panel and used it to evaluate early somatic mutations in archival liver tissues from F344/N rats exposed to the hepatocarcinogen, Aflatoxin B1 (AFB1), a widely studied, potent mutagen and hepatocarcinogen associated with hepatocellular carcinoma (HCC). Rats were exposed to 1-ppm AFB1 in feed for 14, 90, and 90 days plus a recovery 60-day, non-exposure period (150-day) timepoint. Isolated liver DNA was exome sequenced. We identified 172 sequence variants across all timepoints, of which 101 were non-synonymous variants. Well-annotated genes carried a diverse set of 29 non-synonymous mutations at 14 days, increasing to 39 mutations at 90 days and then decreasing to 33 mutations following the 60-day recovery. Gene Set Enrichment Analysis conducted on previously reported, available RNA expression data of the same exome sequenced archival samples identified altered transcripts in pathways associated with malignant transformation. These included HALLMARK gene sets associated with cell proliferation (MYC Targets Version 1 and Version 2, E2F targets), cell cycle (G2M checkpoint, mitotic spindle), cell death (apoptosis), and DNA damage (DNA repair, UV response Up, Reactive oxygen species) pathways. DriverNet Impact analysis integrated exome-seq and expression data to reveal somatic mutations in Mcm8, Bdp1, and Cct6a that may drive cancer formation. Connectivity with transcript expression changes identified these genes as the top-ranked candidate driver genes associated with hepatocellular transformation. In conclusion, exome sequencing revealed early somatic mutations that may play a role in cancer cell transformation that are translatable to aflatoxin-induced HCC.

Single nucleotide polymorphism patterns associated with a cancer resistant phenotype.

BACKGROUND AND AIMS: In this study ten mouse strains representing ~90% of genetic diversity in laboratory mice (B6C3F1/J, C57BL/6J, C3H/HeJ, A/J, NOD.B1oSnH2/J, NZO/HILtJ, 129S1/SvImJ, WSB/EiJ, PWK/PhJ, CAST/EiJ) were examined to identify the mouse strain with the lowest incidence of cancer. The unique single polymorphisms (SNPs) associated with this low cancer incidence are reported. METHODS: Evaluations of cancer incidence in the 10 mouse strains were based on gross and microscopic diagnosis of tumors. Single nucleotide polymorphisms (SNPs) in the coding regions of the genome were derived from the respective mouse strains located in the Sanger mouse sequencing database and the B6C3F1/N genome from the National Toxicology Program (NTP). RESULTS: The WSB strain had an overall lower incidence of both benign and malignant tumors compared to the other mouse strains. At 2 years, the incidence of total malignant tumors (Poly-3 incidence rate) ranged from 2% (WSB) to 92% (C3H) in males, and 14% (WSB) to 93% (NZO) in females, and the total incidence of benign and malignant tumor incidence ranged from 13% (WSB) to 99% (C3H) in males and 25% (WSB) to 96% (NOD) in females. Single nucleotide polymorphism (SNP) patterns were examined in the following strains: B6C3F1/N, C57BL/6J, C3H/HeJ, 129S1/SvImJ, A/J, NZO/HILtJ, CAST/EiJ, PWK/PhJ, and WSB/EiJ. We identified 7519 SNPs (involving 5751 Ensembl transcripts of 3453 Ensembl Genes) that resulted in a unique amino acid change in the coding region of the WSB strain. CONCLUSIONS: The inherited genetic patterns in the WSB cancer-resistant mouse strain occurred in genes involved in multiple cell functions including mitochondria, metabolic, immune, and membrane-related cell functions. The unique SNP patterns in a cancer resistant mouse strain provides insights for understanding and developing strategies for cancer prevention.

Transcriptional pathways linked to fetal and maternal hepatic dysfunction caused by gestational exposure to perfluorooctanoic acid (PFOA) or hexafluoropropylene oxide-dimer acid (HFPO-DA or GenX) in CD-1 mice.

Per- and polyfluoroalkyl substances (PFAS) comprise a diverse class of chemicals used in industrial processes, consumer products, and fire-fighting foams which have become environmental pollutants of concern due to their persistence, ubiquity, and associations with adverse human health outcomes, including in pregnant persons and their offspring. Multiple PFAS are associated with adverse liver outcomes in adult humans and toxicological models, but effects on the developing liver are not fully described. Here we performed transcriptomic analyses in the mouse to investigate the molecular mechanisms of hepatic toxicity in the dam and its fetus after exposure to two different PFAS, perfluorooctanoic acid (PFOA) and its replacement, hexafluoropropylene oxide-dimer acid (HFPO-DA, known as GenX). Pregnant CD-1 mice were exposed via oral gavage from embryonic day (E) 1.5-17.5 to PFOA (0, 1, or 5 mg/kg-d) or GenX (0, 2, or 10 mg/kg-d). Maternal and fetal liver RNA was isolated (N = 5 per dose/group) and the transcriptome analyzed by Affymetrix Array. Differentially expressed genes (DEG) and differentially enriched pathways (DEP) were obtained. DEG patterns were similar in maternal liver for 5 mg/kg PFOA, 2 mg/kg GenX, and 10 mg/kg GenX (R2: 0.46-0.66). DEG patterns were similar across all 4 dose groups in fetal liver (R2: 0.59-0.81). There were more DEGs in fetal liver compared to maternal liver at the low doses for both PFOA (fetal = 69, maternal = 8) and GenX (fetal = 154, maternal = 93). Upregulated DEPs identified across all groups included Fatty Acid Metabolism, Peroxisome, Oxidative Phosphorylation, Adipogenesis, and Bile Acid Metabolism. Transcriptome-phenotype correlation analyses demonstrated > 1000 maternal liver DEGs were significantly correlated with maternal relative liver weight (R2 >0.92). These findings show shared biological pathways of liver toxicity for PFOA and GenX in maternal and fetal livers in CD-1 mice. The limited overlap in specific DEGs between the dam and fetus suggests the developing liver responds differently than the adult liver to these chemical stressors. This work helps define mechanisms of hepatic toxicity of two structurally unique PFAS and may help predict latent consequences of developmental exposure.

BMDExpress 2: enhanced transcriptomic dose-response analysis workflow.

SUMMARY: A new version (version 2) of the genomic dose-response analysis software, BMDExpress, has been created. The software addresses the increasing use of transcriptomic dose-response data in toxicology, drug design, risk assessment and translational research. In this new version, we have implemented additional statistical filtering options (e.g. Williams' trend test), curve fitting models, Linux and Macintosh compatibility and support for additional transcriptomic platforms with up-to-date gene annotations. Furthermore, we have implemented extensive data visualizations, on-the-fly data filtering, and a batch-wise analysis workflow. We have also significantly re-engineered the code base to reflect contemporary software engineering practices and streamline future development. The first version of BMDExpress was developed in 2007 to meet an unmet demand for easy-to-use transcriptomic dose-response analysis software. Since its original release, however, transcriptomic platforms, technologies, pathway annotations and quantitative methods for data analysis have undergone a large change necessitating a significant re-development of BMDExpress. To that end, as of 2016, the National Toxicology Program assumed stewardship of BMDExpress. The result is a modernized and updated BMDExpress 2 that addresses the needs of the growing toxicogenomics user community. AVAILABILITY AND IMPLEMENTATION: BMDExpress 2 is available at https://github.com/auerbachs/BMDExpress-2/releases. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Clustered mutations in yeast and in human cancers can arise from damaged long single-strand DNA regions.

Mutations are typically perceived as random, independent events. We describe here nonrandom clustered mutations in yeast and in human cancers. Genome sequencing of yeast grown under chronic alkylation damage identified mutation clusters that extend up to 200 kb. A predominance of "strand-coordinated" changes of either cytosines or guanines in the same strand, mutation patterns, and genetic controls indicated that simultaneous mutations were generated by base alkylation in abnormally long single-strand DNA (ssDNA) formed at double-strand breaks (DSBs) and replication forks. Significantly, we found mutation clusters with analogous features in sequenced human cancers. Strand-coordinated clusters of mutated cytosines or guanines often resided near chromosome rearrangement breakpoints and were highly enriched with a motif targeted by APOBEC family cytosine-deaminases, which strongly prefer ssDNA. These data indicate that hypermutation via multiple simultaneous changes in randomly formed ssDNA is a general phenomenon that may be an important mechanism producing rapid genetic variation.

Genome-wide promoter DNA methylation profiling of hepatocellular carcinomas arising either spontaneously or due to chronic exposure to Ginkgo biloba extract (GBE) in B6C3F1/N mice.

Epigenetic modifications, such as DNA methylation, play an important role in carcinogenesis. In a recent NTP study, chronic exposure of B6C3F1/N mice to Ginkgo biloba extract (GBE) resulted in a high incidence of hepatocellular carcinomas (HCC). Genome-wide promoter methylation profiling on GBE-exposed HCC (2000 mg/kg group), spontaneous HCC (vehicle-control group), and age-matched vehicle control liver was performed to identify differentially methylated genes in GBE-exposed HCC and spontaneous HCC. DNA methylation alterations were correlated to the corresponding global gene expression changes. Compared to control liver, 1296 gene promoters (719 hypermethylated, 577 hypomethylated) in GBE-exposed HCC and 738 (427 hypermethylated, 311 hypomethylated) gene promoters in spontaneous HCC were significantly differentially methylated, suggesting an impact of methylation on GBE-exposed HCC. Differential methylation of promoter regions in relevant cancer genes (cMyc, Spry2, Dusp5) and their corresponding differential gene expression was validated by quantitative pyrosequencing and qRT-PCR, respectively. In conclusion, we have identified differentially methylated promoter regions of relevant cancer genes altered in GBE-exposed HCC compared to spontaneous HCC. Further study of unique sets of differentially methylated genes in chemical-exposed mouse HCC could potentially be used to differentiate treatment-related tumors from spontaneous-tumors in cancer bioassays and provide additional understanding of the underlying epigenetic mechanisms of chemical carcinogenesis.

DNA Methylation Changes in Tbx3 in a Mouse Model Exposed to Polybrominated Diphenyl Ethers.

DE-71, a commercial mixture of polybrominated diphenyl ethers widely used in flame retardants, is a pervasive environmental contaminant due to its continuing release from waste material and its long half-life in humans. Although the genotoxic potential of DE-71 appears to be low based on bacterial mutagenicity, it remains a public health concern due to its reported involvement in tumor development. Molecular mechanisms by which DE-71 influences tumor incidence or progression remain understudied. We used liver carcinoma tissue from mice exposed to DE-71 to test the hypothesis that epigenetic alterations consistent with tumor development, specifically DNA methylation, result from long-term DE-71 exposure. We profiled DNA methylation status using the methylated-CpG island recovery assay coupled with microarray analysis of hepatocellular carcinoma DNA from animals exposed to DE-71. DE-71 exposure had little impact on global DNA methylation. However, we detected gene body-specific hypomethylation within the Tbx3 locus, a transcription factor important in liver tumorigenesis and in embryonic and cancer stem cell proliferation. This nonpromoter hypomethylation was accompanied by upregulation of Tbx3 mRNA and protein and by alterations in downstream cell cycle-associated marker expression. Thus, exposure to DE-71 may facilitate tumor development by inducing epigenetic programs that favor expansion of progenitor cell populations.

DNA methylation profiling in human B cells reveals immune regulatory elements and epigenetic plasticity at Alu elements during B-cell activation.

Memory is a hallmark of adaptive immunity, wherein lymphocytes mount a superior response to a previously encountered antigen. It has been speculated that epigenetic alterations in memory lymphocytes contribute to their functional distinction from their naive counterparts. However, the nature and extent of epigenetic alterations in memory compartments remain poorly characterized. Here we profile the DNA methylome and the transcriptome of B-lymphocyte subsets representing stages of the humoral immune response before and after antigen exposure in vivo from multiple humans. A significant percentage of activation-induced losses of DNA methylation mapped to transcription factor binding sites. An additional class of demethylated loci mapped to Alu elements across the genome and accompanied repression of DNA methyltransferase 3A. The activation-dependent DNA methylation changes were largely retained in the progeny of activated B cells, generating a similar epigenetic signature in downstream memory B cells and plasma cells with distinct transcriptional programs. These findings provide insights into the methylation dynamics of the genome during cellular differentiation in an immune response.

MBD3 localizes at promoters, gene bodies and enhancers of active genes.

The Mi-2/nucleosome remodeling and histone deacetylase (NuRD) complex is a multiprotein machine proposed to regulate chromatin structure by nucleosome remodeling and histone deacetylation activities. Recent reports describing localization of NuRD provide new insights that question previous models on NuRD action, but are not in complete agreement. Here, we provide location analysis of endogenous MBD3, a component of NuRD complex, in two human breast cancer cell lines (MCF-7 and MDA-MB-231) using two independent genomic techniques: DNA adenine methyltransferase identification (DamID) and ChIP-seq. We observed concordance of the resulting genomic localization, suggesting that these studies are converging on a robust map for NuRD in the cancer cell genome. MBD3 preferentially associated with CpG rich promoters marked by H3K4me3 and showed cell-type specific localization across gene bodies, peaking around the transcription start site. A subset of sites bound by MBD3 was enriched in H3K27ac and was in physical proximity to promoters in three-dimensional space, suggesting function as enhancers. MBD3 enrichment was also noted at promoters modified by H3K27me3. Functional analysis of chromatin indicated that MBD3 regulates nucleosome occupancy near promoters and in gene bodies. These data suggest that MBD3, and by extension the NuRD complex, may have multiple roles in fine tuning expression for both active and silent genes, representing an important step in defining regulatory mechanisms by which NuRD complex controls chromatin structure and modification status.

RNA-Seq-based toxicogenomic assessment of fresh frozen and formalin-fixed tissues yields similar mechanistic insights.

Formalin-fixed, paraffin-embedded (FFPE) pathology specimens represent a potentially vast resource for transcriptomic-based biomarker discovery. We present here a comparison of results from a whole transcriptome RNA-Seq analysis of RNA extracted from fresh frozen and FFPE livers. The samples were derived from rats exposed to aflatoxin B1 (AFB1 ) and a corresponding set of control animals. Principal components analysis indicated that samples were separated in the two groups representing presence or absence of chemical exposure, both in fresh frozen and FFPE sample types. Sixty-five percent of the differentially expressed transcripts (AFB1 vs. controls) in fresh frozen samples were also differentially expressed in FFPE samples (overlap significance: P < 0.0001). Genomic signature and gene set analysis of AFB1 differentially expressed transcript lists indicated highly similar results between fresh frozen and FFPE at the level of chemogenomic signatures (i.e., single chemical/dose/duration elicited transcriptomic signatures), mechanistic and pathology signatures, biological processes, canonical pathways and transcription factor networks. Overall, our results suggest that similar hypotheses about the biological mechanism of toxicity would be formulated from fresh frozen and FFPE samples. These results indicate that phenotypically anchored archival specimens represent a potentially informative resource for signature-based biomarker discovery and mechanistic characterization of toxicity.

Exposure to PFAS chemicals induces sex-dependent alterations in key rate-limiting steps of lipid metabolism in liver steatosis.

Toxicants with the potential to bioaccumulate in humans and animals have long been a cause for concern, particularly due to their association with multiple diseases and organ injuries. Per- and polyfluoro alkyl substances (PFAS) and polycyclic aromatic hydrocarbons (PAH) are two such classes of chemicals that bioaccumulate and have been associated with steatosis in the liver. Although PFAS and PAH are classified as chemicals of concern, their molecular mechanisms of toxicity remain to be explored in detail. In this study, we aimed to identify potential mechanisms by which an acute exposure to PFAS and PAH chemicals can induce lipid accumulation and whether the responses depend on chemical class, dose, and sex. To this end, we analyzed mechanisms beginning with the binding of the chemical to a molecular initiating event (MIE) and the consequent transcriptomic alterations. We collated potential MIEs using predictions from our previously developed ToxProfiler tool and from published steatosis adverse outcome pathways. Most of the MIEs are transcription factors, and we collected their target genes by mining the TRRUST database. To analyze the effects of PFAS and PAH on the steatosis mechanisms, we performed a computational MIE-target gene analysis on high-throughput transcriptomic measurements of liver tissue from male and female rats exposed to either a PFAS or PAH. The results showed peroxisome proliferator-activated receptor (PPAR)-α targets to be the most dysregulated, with most of the genes being upregulated. Furthermore, PFAS exposure disrupted several lipid metabolism genes, including upregulation of fatty acid oxidation genes (Acadm, Acox1, Cpt2, Cyp4a1-3) and downregulation of lipid transport genes (Apoa1, Apoa5, Pltp). We also identified multiple genes with sex-specific behavior. Notably, the rate-limiting genes of gluconeogenesis (Pck1) and bile acid synthesis (Cyp7a1) were specifically downregulated in male rats compared to female rats, while the rate-limiting gene of lipid synthesis (Scd) showed a PFAS-specific upregulation. The results suggest that the PPAR signaling pathway plays a major role in PFAS-induced lipid accumulation in rats. Together, these results show that PFAS exposure induces a sex-specific multi-factorial mechanism involving rate-limiting genes of gluconeogenesis and bile acid synthesis that could lead to activation of an adverse outcome pathway for steatosis.

Characterization and optimization of variability in a human colonic epithelium culture model.

Animal models have historically been poor preclinical predictors of gastrointestinal (GI) directed therapeutic efficacy and drug-induced GI toxicity. Human stem and primary cell-derived culture systems are a major focus of efforts to create biologically relevant models that enhance preclinical predictive value of intestinal efficacy and toxicity. The inherent variability in stem cell-based cultures makes development of useful models a challenge; the stochastic nature of stem cell differentiation interferes with the ability to build and validate reproducible assays that query drug responses and pharmacokinetics. In this study, we aimed to characterize and reduce sources of variability in a complex stem cell-derived intestinal epithelium model, termed RepliGut® Planar, across cells from multiple human donors, cell lots, and passage numbers. Assessment criteria included barrier for­mation and integrity, gene expression, and cytokine responses. Gene expression and culture metric analyses revealed that controlling cell passage number reduces variability and maximizes physi­ological relevance of the model. In a case study where passage number was optimized, distinct cytokine responses were observed among four human donors, indicating that biological variability can be detected in cell cultures originating from diverse human sources. These findings highlight key considerations for designing assays that can be applied to additional primary cell-derived systems, as well as establish utility of the RepliGut® Planar platform for robust development of human-pre­dictive drug-response assays.

Animal models are frequently used as tools for studying gastrointestinal (GI) disease, but they inad­equately replicate the complexities of the human gut, making them poor predictors of how humans respond to new drugs. Models using human stem cells are closer to human GI physiology, but their responses are not uniform owing to variability in the stem cells. We looked for the sources of this variability in the primary stem-cell derived RepliGut® Planar model. We found that limiting how long the cells were kept in culture reduced their variability and improved the physiological relevance of the model. These findings highlight key assay design considerations that also can be applied to other primary cell-derived systems. Reliable and physiologically relevant cell-based models can reduce animal testing, improve research accuracy, and ensure new treatments are more relevant and effective for patients.

Characterization and optimization of variability in a human colonic epithelium culture model.

Animal models have historically been poor preclinical predictors of gastrointestinal (GI) directed therapeutic efficacy and drug-induced GI toxicity. Human stem and primary cell-derived culture systems are a major focus of efforts to create biologically relevant models that enhance preclinical predictive value of intestinal efficacy and toxicity. The inherent variability in stem-cell-based complex cultures makes development of useful models a challenge; the stochastic nature of stem-cell differentiation interferes with the ability to build and validate robust, reproducible assays that query drug responses and pharmacokinetics. In this study, we aimed to characterize and reduce potential sources of variability in a complex stem cell-derived intestinal epithelium model, termed RepliGut® Planar, across cells from multiple human donors, cell lots, and passage numbers. Assessment criteria included barrier formation and integrity, gene expression, and cytokine responses. Gene expression and culture metric analyses revealed that controlling for stem/progenitor-cell passage number reduces variability and maximizes physiological relevance of the model. After optimizing passage number, donor-specific differences in cytokine responses were observed in a case study, suggesting biologic variability is observable in cell cultures derived from multiple human sources. Our findings highlight key considerations for designing assays that can be applied to additional primary-cell derived systems, as well as establish utility of the RepliGut® Planar platform for robust development of human-predictive drug-response assays.

Impact of Aligner, Normalization Method, and Sequencing Depth on TempO-seq Accuracy.

High-throughput transcriptomics has advanced through the introduction of TempO-seq, a targeted alternative to traditional RNA-seq. TempO-seq platforms use 50 nucleotide probes, each specifically designed to target a known transcript, thus allowing for reduced sequencing depth per sample compared with RNA-seq without compromising the accuracy of results. Thus far, studies using the TempO-seq method have relied on existing tools for processing the resulting short read data. However, these tools were originally designed for other data types. While they have been used for processing of early TempO-seq data, they have not been systematically assessed for accuracy or compared to determine an optimal framework for processing and analyzing TempO-seq data. In this work, we re-analyze several publicly available TempO-seq data sets covering a range of experimental designs and use corresponding RNA-seq data sets as a gold standard to rigorously assess accuracy at multiple levels. We compare 6 aligners and 5 normalization methods across various accuracy and performance metrics. Our results demonstrate the overall robust accuracy of the TempO-seq platform, independent of data processing methods. Complex aligners and advanced normalization methods do not appear to have any general advantage over simpler methods when it comes to analyzing TempO-seq data. The reduced complexity of the sequencing space, and the fact that TempO-seq probes are all equal length, appears to reduce the need for elaborate bioinformatic or statistical methods used to address these factors in RNA-seq data.

Runx2 transcriptome of prostate cancer cells: insights into invasiveness and bone metastasis.

BACKGROUND: Prostate cancer (PCa) cells preferentially metastasize to bone at least in part by acquiring osteomimetic properties. Runx2, an osteoblast master transcription factor, is aberrantly expressed in PCa cells, and promotes their metastatic phenotype. The transcriptional programs regulated by Runx2 have been extensively studied during osteoblastogenesis, where it activates or represses target genes in a context-dependent manner. However, little is known about the gene regulatory networks influenced by Runx2 in PCa cells. We therefore investigated genome wide mRNA expression changes in PCa cells in response to Runx2. RESULTS: We engineered a C4-2B PCa sub-line called C4-2B/Rx2 dox, in which Doxycycline (Dox) treatment stimulates Runx2 expression from very low to levels observed in other PCa cells. Transcriptome profiling using whole genome expression array followed by in silico analysis indicated that Runx2 upregulated a multitude of genes with prominent cancer associated functions. They included secreted factors (CSF2, SDF-1), proteolytic enzymes (MMP9, CST7), cytoskeleton modulators (SDC2, Twinfilin, SH3PXD2A), intracellular signaling molecules (DUSP1, SPHK1, RASD1) and transcription factors (Sox9, SNAI2, SMAD3) functioning in epithelium to mesenchyme transition (EMT), tissue invasion, as well as homing and attachment to bone. Consistent with the gene expression data, induction of Runx2 in C4-2B cells enhanced their invasiveness. It also promoted cellular quiescence by blocking the G1/S phase transition during cell cycle progression. Furthermore, the cell cycle block was reversed as Runx2 levels declined after Dox withdrawal. CONCLUSIONS: The effects of Runx2 in C4-2B/Rx2 dox cells, as well as similar observations made by employing LNCaP, 22RV1 and PC3 cells, highlight multiple mechanisms by which Runx2 promotes the metastatic phenotype of PCa cells, including tissue invasion, homing to bone and induction of high bone turnover. Runx2 is therefore an attractive target for the development of novel diagnostic, prognostic and therapeutic approaches to PCa management. Targeting Runx2 may prove more effective than focusing on its individual downstream genes and pathways.

Predicting the hepatocarcinogenic potential of alkenylbenzene flavoring agents using toxicogenomics and machine learning.

Identification of carcinogenic activity is the primary goal of the 2-year bioassay. The expense of these studies limits the number of chemicals that can be studied and therefore chemicals need to be prioritized based on a variety of parameters. We have developed an ensemble of support vector machine classification models based on male F344 rat liver gene expression following 2, 14 or 90 days of exposure to a collection of hepatocarcinogens (aflatoxin B1, 1-amino-2,4-dibromoanthraquinone, N-nitrosodimethylamine, methyleugenol) and non-hepatocarcinogens (acetaminophen, ascorbic acid, tryptophan). Seven models were generated based on individual exposure durations (2, 14 or 90 days) or a combination of exposures (2+14, 2+90, 14+90 and 2+14+90 days). All sets of data, with the exception of one yielded models with 0% cross-validation error. Independent validation of the models was performed using expression data from the liver of rats exposed at 2 dose levels to a collection of alkenylbenzene flavoring agents. Depending on the model used and the exposure duration of the test data, independent validation error rates ranged from 47% to 10%. The variable with the most notable effect on independent validation accuracy was exposure duration of the alkenylbenzene test data. All models generally exhibited improved performance as the exposure duration of the alkenylbenzene data increased. The models differentiated between hepatocarcinogenic (estragole and safrole) and non-hepatocarcinogenic (anethole, eugenol and isoeugenol) alkenylbenzenes previously studied in a carcinogenicity bioassay. In the case of safrole the models correctly differentiated between carcinogenic and non-carcinogenic dose levels. The models predict that two alkenylbenzenes not previously assessed in a carcinogenicity bioassay, myristicin and isosafrole, would be weakly hepatocarcinogenic if studied at a dose level of 2 mmol/kg bw/day for 2 years in male F344 rats; therefore suggesting that these chemicals should be a higher priority relative to other untested alkenylbenzenes for evaluation in the carcinogenicity bioassay. The results of the study indicate that gene expression-based predictive models are an effective tool for identifying hepatocarcinogens. Furthermore, we find that exposure duration is a critical variable in the success or failure of such an approach, particularly when evaluating chemicals with unknown carcinogenic potency.

SWIFT-Active Screener: Accelerated document screening through active learning and integrated recall estimation.

BACKGROUND: In the screening phase of systematic review, researchers use detailed inclusion/exclusion criteria to decide whether each article in a set of candidate articles is relevant to the research question under consideration. A typical review may require screening thousands or tens of thousands of articles in and can utilize hundreds of person-hours of labor. METHODS: Here we introduce SWIFT-Active Screener, a web-based, collaborative systematic review software application, designed to reduce the overall screening burden required during this resource-intensive phase of the review process. To prioritize articles for review, SWIFT-Active Screener uses active learning, a type of machine learning that incorporates user feedback during screening. Meanwhile, a negative binomial model is employed to estimate the number of relevant articles remaining in the unscreened document list. Using a simulation involving 26 diverse systematic review datasets that were previously screened by reviewers, we evaluated both the document prioritization and recall estimation methods. RESULTS: On average, 95% of the relevant articles were identified after screening only 40% of the total reference list. In the 5 document sets with 5,000 or more references, 95% recall was achieved after screening only 34% of the available references, on average. Furthermore, the recall estimator we have proposed provides a useful, conservative estimate of the percentage of relevant documents identified during the screening process. CONCLUSION: SWIFT-Active Screener can result in significant time savings compared to traditional screening and the savings are increased for larger project sizes. Moreover, the integration of explicit recall estimation during screening solves an important challenge faced by all machine learning systems for document screening: when to stop screening a prioritized reference list. The software is currently available in the form of a multi-user, collaborative, online web application.

Utility of Extrapolating Human S1500+ Genes to the Whole Transcriptome: Tunicamycin Case Study.

The TempO-Seq S1500+ platform(s), now available for human, mouse, rat, and zebrafish, measures a discrete number of genes that are representative of biological and pathway co-regulation across the entire genome in a given species. While measurement of these genes alone provides a direct assessment of gene expression activity, extrapolating expression values to the whole transcriptome (~26 000 genes in humans) can estimate measurements of non-measured genes of interest and increases the power of pathway analysis algorithms by using a larger background gene expression space. Here, we use data from primary hepatocytes of 54 donors that were treated with the endoplasmic reticulum (ER) stress inducer tunicamycin and then measured on the human S1500+ platform containing ~3000 representative genes. Measurements for the S1500+ genes were then used to extrapolate expression values for the remaining human transcriptome. As a case study of the improved downstream analysis achieved by extrapolation, the "measured only" and "whole transcriptome" (measured + extrapolated) gene sets were compared. Extrapolation increased the number of significant genes by 49%, bringing to the forefront many that are known to be associated with tunicamycin exposure. The extrapolation procedure also correctly identified established tunicamycin-related functional pathways reflected by coordinated changes in interrelated genes while maintaining the sample variability observed from the "measured only" genes. Extrapolation improved the gene- and pathway-level biological interpretations for a variety of downstream applications, including differential expression analysis, gene set enrichment pathway analysis, DAVID keyword analysis, Ingenuity Pathway Analysis, and NextBio correlated compound analysis. The extrapolated data highlight the role of metabolism/metabolic pathways, the ER, immune response, and the unfolded protein response, each of which are key activities associated with tunicamycin exposure that were unrepresented or underrepresented in one or more of the analyses of the original "measured only" dataset. Furthermore, the inclusion of the extrapolated genes raised "tunicamycin" from third to first upstream regulator in Ingenuity Pathway Analysis and from sixth to second most correlated compound in NextBio analysis. Therefore, our case study suggests an approach to extend and enhance data from the S1500+ platform for improved insight into biological mechanisms and functional outcomes of diseases, drugs, and other perturbations.

Evaluation of 5-day In Vivo Rat Liver and Kidney With High-throughput Transcriptomics for Estimating Benchmark Doses of Apical Outcomes.

A 5-day in vivo rat model was evaluated as an approach to estimate chemical exposures that may pose minimal risk by comparing benchmark dose (BMD) values for transcriptional changes in the liver and kidney to BMD values for toxicological endpoints from traditional toxicity studies. Eighteen chemicals, most having been tested by the National Toxicology Program in 2-year bioassays, were evaluated. Some of these chemicals are potent hepatotoxicants (eg, DE71, PFOA, and furan) in rodents, some exhibit toxicity but have minimal hepatic effects (eg, acrylamide and α,ß-thujone), and some exhibit little overt toxicity (eg, ginseng and milk thistle extract) based on traditional toxicological evaluations. Male Sprague Dawley rats were exposed once daily for 5 consecutive days by oral gavage to 8-10 dose levels for each chemical. Liver and kidney were collected 24 h after the final exposure and total RNA was assayed using high-throughput transcriptomics (HTT) with the rat S1500+ platform. HTT data were analyzed using BMD Express 2 to determine transcriptional gene set BMD values. BMDS was used to determine BMD values for histopathological effects from chronic or subchronic toxicity studies. For many of the chemicals, the lowest transcriptional BMDs from the 5-day assays were within a factor of 5 of the lowest histopathological BMDs from the toxicity studies. These data suggest that using HTT in a 5-day in vivo rat model provides reasonable estimates of BMD values for traditional apical endpoints. This approach may be useful to prioritize chemicals for further testing while providing actionable data in a timely and cost-effective manner.