Evaluation of an integrated clinical workflow for targeted next-generation sequencing of low-quality tumor DNA using a 51-gene enrichment panel.

BACKGROUND: Improvements in both performance and cost for next-generation sequencing (NGS) have spurred its rapid adoption for clinical applications. We designed and optimized a pan-cancer target-enrichment panel for 51 well-established oncogenes and tumor suppressors, in conjunction with a bioinformatic pipeline informed by in-process controls and pre- and post-analytical quality control measures. METHODS: The evaluation of this workflow consisted of sequencing mixtures of intact DNA to establish analytical sensitivity and precision, utilization of heuristics to identify systematic artifacts, titration studies of intact and FFPE samples for input optimization, and incorporation of orthogonal sequencing strategies to increase both positive predictive value and variant detection. We also used 128 FFPE samples to assess clinical accuracy and incorporated the previously described quantitative functional index (QFI) for sample qualification as part of detailing complete system performance. RESULTS: We observed a concordance correlation coefficient of 0.99 between the observed versus expected percent variant at 250 ng input across 4 independent sequencing runs. A subset of the systematic variants were confirmed to be barely detectable on an independent sequencing platform (Wilcox signed-rank test p-value <10(-16)), and the incorporation of orthogonal sequencing strategies increased the harmonic mean of sensitivity and positive predictive value of mutation detection by 41%. In one cohort of FFPE tumor samples, coverage and inter-platform concordance were positively correlated with the QFI, emphasizing the need for pre-analytical sample quality control to reduce the risk of false positives and negatives. In a separate cohort of FFPE samples, the 51-gene panel achieved 78% sensitivity (95% CI = 56.3, 92.5) with 100% PPV (95% CI = 81.5, 100.0) based on known mutations at 7.9% median abundance. By sequencing specimens using an orthogonal NGS technology, sensitivity was improved to 87.0% (95% CI = 66.4,97.2) while maintaining PPV. CONCLUSIONS: The results highlight the value of process integration in a comprehensive targeted NGS system, enabling both discovery and diagnostic applications, particularly when sequencing low-quality cancer specimens.

A microRNA-based test improves endoscopic ultrasound-guided cytologic diagnosis of pancreatic cancer.

BACKGROUND & AIMS: Endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) in combination with cytopathology is the optimal method for diagnosis and staging of pancreatic ductal adenocarcinoma (PDAC) and other pancreatic lesions. Its clinical utility, however, can be limited by high rates of indeterminate or false-negative results. We aimed to develop and validate a microRNA (miRNA)-based test to improve preoperative detection of PDAC. METHODS: Levels of miRNAs were analyzed in a centralized clinical laboratory by relative quantitative polymerase chain reaction in 95 formalin-fixed paraffin-embedded specimens and 228 samples collected by EUS-FNA during routine evaluations of patients with solid pancreatic masses at 4 institutions in the United States, 1 in Canada, and 1 in Poland. RESULTS: We developed a 5-miRNA expression classifier, consisting of MIR24, MIR130B, MIR135B, MIR148A, and MIR196, that could identify PDAC in well-characterized, formalin-fixed, paraffin-embedded specimens. Detection of PDAC in EUS-FNA samples increased from 78.8% by cytology analysis alone (95% confidence interval, 72.2%-84.5%) to 90.8% when combined with miRNA analysis (95% confidence interval, 85.6%-94.5%). The miRNA classifier correctly identified 22 additional true PDAC cases among 39 samples initially classified as benign, indeterminate, or nondiagnostic by cytology. Cytology and miRNA test results each were associated significantly with PDAC (P < .001), with positive predictive values greater than 99% (95% confidence interval, 96%-100%). CONCLUSIONS: We developed and validated a 5-miRNA classifier that can accurately predict which preoperative pancreatic EUS-FNA specimens contain PDAC. This test might aid in the diagnosis of pancreatic cancer by reducing the number of FNAs without a definitive adenocarcinoma diagnosis, thereby reducing the number of repeat EUS-FNA procedures.

Targeted, high-depth, next-generation sequencing of cancer genes in formalin-fixed, paraffin-embedded and fine-needle aspiration tumor specimens.

Implementation of highly sophisticated technologies, such as next-generation sequencing (NGS), into routine clinical practice requires compatibility with common tumor biopsy types, such as formalin-fixed, paraffin-embedded (FFPE) and fine-needle aspiration specimens, and validation metrics for platforms, controls, and data analysis pipelines. In this study, a two-step PCR enrichment workflow was used to assess 540 known cancer-relevant variants in 16 oncogenes for high-depth sequencing in tumor samples on either mature (Illumina GAIIx) or emerging (Ion Torrent PGM) NGS platforms. The results revealed that the background noise of variant detection was elevated approximately twofold in FFPE compared with cell line DNA. Bioinformatic algorithms were optimized to accommodate this background. Variant calls from 38 residual clinical colorectal cancer FFPE specimens and 10 thyroid fine-needle aspiration specimens were compared across multiple cancer genes, resulting in an accuracy of 96.1% (95% CI, 96.1% to 99.3%) compared with Sanger sequencing, and 99.6% (95% CI, 97.9% to 99.9%) compared with an alternative method with an analytical sensitivity of 1% mutation detection. A total of 45 of 48 samples were concordant between NGS platforms across all matched regions, with the three discordant calls each represented at <10% of reads. Consequently, NGS of targeted oncogenes in real-life tumor specimens using distinct platforms addresses unmet needs for unbiased and highly sensitive mutation detection and can accelerate both basic and clinical cancer research.

miRNA biomarkers in cyst fluid augment the diagnosis and management of pancreatic cysts.

PURPOSE: The diagnosis of pancreatic cystic lesions has increased dramatically. Most are benign, whereas some, such as intraductal papillary mucinous neoplasms (IPMN), represent precursors of pancreatic adenocarcinoma. Therapeutic stratification of IPMNs is challenging without precise information on dysplasia grade and presence of invasion. We assessed the diagnostic benefit of using miRNAs as biomarkers in pancreatic cyst fluid, focusing on IPMNs because of their frequency and malignant potential. EXPERIMENTAL DESIGN: RNA was extracted from 55 microdissected formalin-fixed, paraffin-embedded (FFPE) IPMN specimens, and 65 cyst fluid specimens aspirated following surgical resection. Expression of 750 miRNAs was evaluated with TaqMan miRNA Arrays using 22 FFPE and 15 cyst fluid specimens. Differential expression of selected miRNA candidates was validated in 33 FFPE and 50 cyst fluid specimens using TaqMan miRNA Assays. RESULTS: We identified 26 and 37 candidate miRNAs that distinguish low-grade from high-grade IPMNs using FFPE and cyst fluid specimens, respectively. A subset of 18 miRNAs, selected from FFPE and cyst fluid data, separated high-grade IPMNs from low-grade IPMNs, serous cystadenomas (SCA) and uncommon cysts, such as solid pseudopapillary neoplasms (SPN) and cystic pancreatic neuroendocrine tumors (PanNET). A logistic regression model using nine miRNAs allowed prediction of cyst pathology implying resection (high-grade IPMNs, PanNETs, and SPNs) versus conservative management (low-grade IPMNs, SCAs), with a sensitivity of 89%, a specificity of 100%, and area under the curve of 1. CONCLUSIONS: We found candidate miRNAs that helped identify patients with high-grade IPMN and exclude nonmucinous cysts. These classifiers will require validation in a prospective setting to ultimately confirm their clinical usefulness.

Global microRNA expression profiling of microdissected tissues identifies miR-135b as a novel biomarker for pancreatic ductal adenocarcinoma.

Pancreatic ductal adenocarcinoma (PDAC) is known for its poor prognosis resulting from being diagnosed at an advanced stage. Accurate early diagnosis and new therapeutic modalities are therefore urgently needed. MicroRNAs (miRNAs), considered a new class of biomarkers and therapeutic targets, may be able to fulfill those needs. Combining tissue microdissection with global miRNA array analyses, cell type-specific miRNA expression profiles were generated for normal pancreatic ductal cells, acinar cells, PDAC cells derived from xenografts and also from macrodissected chronic pancreatitis (CP) tissues. We identified 78 miRNAs differentially expressed between ND and PDAC cells providing new insights into the miRNA-driven pathophysiological mechanisms involved in PDAC development. Having filtered miRNAs which are upregulated in the three pairwise comparisons of PDAC vs. ND, PDAC vs. AZ and PDAC vs. CP, we identified 15 miRNA biomarker candidates including miR-135b. Using relative qRT-PCR to measure miR-135b normalized to miR-24 in 75 FFPE specimens (42 PDAC and 33 CP) covering a broad range of tumor content, we discriminated CP from PDAC with a sensitivity and specificity of 92.9% [95% CI=(80.5, 98.5)] and 93.4% [95% CI=(79.8, 99.3)], respectively. Furthermore, the area under the curve (AUC) value reached of 0.97 was accompanied by positive and negative predictive values of 95% and 91%, respectively. In conclusion, we report pancreatic cell-specific global miRNA profiles, which offer new candidate miRNAs to be exploited for functional studies in PDAC. Furthermore, we provide evidence that miRNAs are well-suited analytes for development of sensitive and specific aid-in-diagnosis tests for PDAC.

A novel mean-centering method for normalizing microRNA expression from high-throughput RT-qPCR data.

BACKGROUND: Normalization is critical for accurate gene expression analysis. A significant challenge in the quantitation of gene expression from biofluids samples is the inability to quantify RNA concentration prior to analysis, underscoring the need for robust normalization tools for this sample type. In this investigation, we evaluated various methods of normalization to determine the optimal approach for quantifying microRNA (miRNA) expression from biofluids and tissue samples when using the TaqMan® Megaplex™ high-throughput RT-qPCR platform with low RNA inputs. FINDINGS: We compared seven normalization methods in the analysis of variation of miRNA expression from biofluid and tissue samples. We developed a novel variant of the common mean-centering normalization strategy, herein referred to as mean-centering restricted (MCR) normalization, which is adapted to the TaqMan Megaplex RT-qPCR platform, but is likely applicable to other high-throughput RT-qPCR-based platforms. Our results indicate that MCR normalization performs comparable to or better than both standard mean-centering and other normalization methods. We also propose an extension of this method to be used when migrating biomarker signatures from Megaplex to singleplex RT-qPCR platforms, based on the identification of a small number of normalizer miRNAs that closely track the mean of expressed miRNAs. CONCLUSIONS: We developed the MCR method for normalizing miRNA expression from biofluids samples when using the TaqMan Megaplex RT-qPCR platform. Our results suggest that normalization based on the mean of all fully observed (fully detected) miRNAs minimizes technical variance in normalized expression values, and that a small number of normalizer miRNAs can be selected when migrating from Megaplex to singleplex assays. In our study, we find that normalization methods that focus on a restricted set of miRNAs tend to perform better than methods that focus on all miRNAs, including those with non-determined (missing) values. This methodology will likely be most relevant for studies in which a significant number of miRNAs are not detected.

miR-221/222 targeting of trichorhinophalangeal 1 (TRPS1) promotes epithelial-to-mesenchymal transition in breast cancer.

Compared with the luminal subtype, the basal-like subtype of breast cancer has an aggressive clinical behavior, but the reasons for this difference between the two subtypes are poorly understood. We identified microRNAs (miRNAs) miR-221 and miR-222 (miR-221/222) as basal-like subtype-specific miRNAs that decrease expression of epithelial-specific genes and increase expression of mesenchymal-specific genes. In addition, expression of these miRNAs increased cell migration and invasion, which collectively are characteristics of the epithelial-to-mesenchymal transition (EMT). The basal-like transcription factor FOSL1 (also known as Fra-1) directly stimulated the transcription of miR-221/222, and the abundance of these miRNAs decreased with inhibition of MEK (mitogen-activated or extracellular signal-regulated protein kinase kinase), placing miR-221/222 downstream of the RAS pathway. The miR-221/222-mediated reduction in E-cadherin abundance depended on their targeting of the 3' untranslated region (3'UTR) of TRPS1 (trichorhinophalangeal syndrome type 1), which is a member of the GATA family of transcriptional repressors. TRPS1 inhibited EMT by directly repressing expression of ZEB2 (Zinc finger E-box-binding homeobox 2). Therefore, miR-221/222 may contribute to the aggressive clinical behavior of basal-like breast cancers.

TRPS1 targeting by miR-221/222 promotes the epithelial-to-mesenchymal transition in breast cancer.

The basal-like subtype of breast cancer has an aggressive clinical behavior compared to that of the luminal subtype. We identified the microRNAs (miRNAs) miR-221 and miR-222 (miR-221/222) as basal-like subtype-specific miRNAs and showed that expression of miR-221/222 decreased expression of epithelial-specific genes and increased expression of mesenchymal-specific genes, and increased cell migration and invasion in a manner characteristic of the epithelial-to-mesenchymal transition (EMT). The transcription factor FOSL1 (also known as Fra-1), which is found in basal-like breast cancers but not in the luminal subtype, stimulated the transcription of miR-221/222, and the abundance of these miRNAs decreased with inhibition of the epidermal growth factor receptor (EGFR) or MEK (mitogen-activated or extracellular signal-regulated protein kinase kinase), placing miR-221/222 downstream of the RAS pathway. Furthermore, miR-221/222-mediated reduction in E-cadherin abundance depended on their targeting the 3' untranslated region of the GATA family transcriptional repressor TRPS1 (tricho-rhino-phalangeal syndrome type 1), which inhibited EMT by decreasing ZEB2 (zinc finger E-box-binding homeobox2) expression. We conclude that by promoting EMT, miR-221/222 may contribute to the more aggressive clinical behavior of basal-like breast cancers.

Development of a miRNA-based diagnostic assay for pancreatic ductal adenocarcinoma.

Diagnosis of pancreatic cancer remains a clinical challenge. Both chronic pancreatitis and pancreatic cancer may present with similar symptoms and similar imaging features, often leading to incorrect interpretation. Thus, the use of an objective molecular test that can discriminate between chronic pancreatitis and pancreatic cancer will be a valuable asset in obtaining a definitive diagnosis of pancreatic cancer. Following Clinical Laboratory Improvement Amendments and College of American Pathologists guidelines, Asuragen Clinical Services Laboratory has developed and validated a laboratory-developed test, miRInform(®) Pancreas, to aid in the identification of pancreatic ductal adenocarcinoma. This molecular diagnostic tool uses reverse-transcription quantitative PCR to measure the expression difference between two miRNAs, miR-196a and miR-217, in fixed tissue specimens. This article describes the test validation process as well as determination of performance parameters of miRInform Pancreas.

CSRDB: a small RNA integrated database and browser resource for cereals.

Plant small RNAs (smRNAs), which include microRNAs (miRNAs), short interfering RNAs (siRNAs) and trans-acting siRNAs (ta-siRNAs), are emerging as significant components of epigenetic processes and of gene networks involved in development and in homeostasis. Here we present a bioinformatics resource for cereal crops, the Cereal Small RNA Database (CSRDB), consisting of large-scale datasets of maize and rice smRNA sequences generated by high-throughput pyrosequencing. The smRNA sequences have been mapped to the rice genome and to the available maize genome sequence and these results are presented in two genome browser datasets using the Generic Genome Browser. Potential RNA targets for the smRNAs have been predicted and access to the resulting smRNA/RNA target pair dataset has been made available through a MySQL based relational database. Various ways to access the data are provided including links from the genome browser to the target database. Data linking and integration are the main focus for this interface, and internal as well as external links are present. The resource is available at http://sundarlab.ucdavis.edu/smrnas/ and will be updated as more sequences become available.

Comparative whole genome transcriptome analysis of three Plasmodium falciparum strains.

Gene expression patterns have been demonstrated to be highly variable between similar cell types, for example lab strains and wild strains of Saccharomyces cerevisiae cultured under identical growth conditions exhibit a wide range of expression differences. We have used a genome-wide approach to characterize transcriptional differences between strains of Plasmodium falciparum by characterizing the transcriptome of the 48 h intraerythrocytic developmental cycle (IDC) for two strains, 3D7 and Dd2 and compared these results to our prior work using the HB3 strain. These three strains originate from geographically diverse locations and possess distinct drug sensitivity phenotypes. Our goal was to identify transcriptional differences related to phenotypic properties of these strains including immune evasion and drug sensitivity. We find that the highly streamlined transcriptome is remarkably well conserved among all three strains, and differences in gene expression occur mainly in genes coding for surface antigens involved in parasite-host interactions. Our analysis also detects several transcripts that are unique to individual strains as well as identifying large chromosomal deletions and highly polymorphic regions across strains. The majority of these genes are uncharacterized and have no homology to other species. These tractable transcriptional differences provide important phenotypes for these otherwise highly related strains of Plasmodium.

A probabilistic functional network of yeast genes.

A conceptual framework for integrating diverse functional genomics data was developed by reinterpreting experiments to provide numerical likelihoods that genes are functionally linked. This allows direct comparison and integration of different classes of data. The resulting probabilistic gene network estimates the functional coupling between genes. Within this framework, we reconstructed an extensive, high-quality functional gene network for Saccharomyces cerevisiae, consisting of 4681 (approximately 81%) of the known yeast genes linked by approximately 34,000 probabilistic linkages comparable in accuracy to small-scale interaction assays. The integrated linkages distinguish true from false-positive interactions in earlier data sets; new interactions emerge from genes' network contexts, as shown for genes in chromatin modification and ribosome biogenesis.

LGL: creating a map of protein function with an algorithm for visualizing very large biological networks.

Networks are proving to be central to the study of gene function, protein-protein interaction, and biochemical pathway data. Visualization of networks is important for their study, but visualization tools are often inadequate for working with very large biological networks. Here, we present an algorithm, called large graph layout (LGL), which can be used to dynamically visualize large networks on the order of hundreds of thousands of vertices and millions of edges. LGL applies a force-directed iterative layout guided by a minimal spanning tree of the network in order to generate coordinates for the vertices in two or three dimensions, which are subsequently visualized and interactively navigated with companion programs. We demonstrate the use of LGL in visualizing an extensive protein map summarizing the results of approximately 21 billion sequence comparisons between 145579 proteins from 50 genomes. Proteins are positioned in the map according to sequence homology and gene fusions, with the map ultimately serving as a theoretical framework that integrates inferences about gene function derived from sequence homology, remote homology, gene fusions, and higher-order fusions. We confirm that protein neighbors in the resulting map are functionally related, and that distinct map regions correspond to distinct cellular systems, enabling a computational strategy for discovering proteins' functions on the basis of the proteins' map positions. Using the map produced by LGL, we infer general functions for 23 uncharacterized protein families.