Search

Introduction: The vast majority of cancer patients continue to receive treatments that are minimally informed by omics data. In the case of breast cancer, only ER and HER2 are routinely used for treatment selection. There is a particular need for personalized treatment in individuals with primary and secondary drug resistance or aggressive breast cancers. Emerging bioinformatics and statistical methods have made a fundamental impact on cancer research. However, challenges remains with regard to patient-centric data analysis and providing genomic data guidance to oncologists. There exists a large number of FDA approved anti-neoplastic drugs used to treat cancers other than breast and the development of innovative informatics methods and algorithms to repurpose those drugs should benefit breast cancer patients. Methods and Results: We have developed precision care systems (such as PANOPLY and CORPUS) to identify personalized therapies for an individual patient and to deliver genomic reports in a standard, searchable format so that a researcher or an oncologist can quickly navigate through molecular data and obtain prioritized drugs and targets.The PANOPLY (Precision cancer genomics report: single sample inventory) algorithm applies machine learning and topology-based network analysis methods to integrate multi-omics profiles and clinical data; individual-specific molecular alterations are identified and compared with a set of matched-controls having similar clinical data. Since there is a lack of a “gold standard” dataset to test such algorithms, we simulated 500 case-control sets and evaluated drug predictions across multiple simulation scenarios. We applied the PANOPLY algorithm to The Cancer Genome Atlas (TCGA) breast cancer cohort, which consists of multi-omics data and clinical data. In addition, PANOPLY was also applied to an in-house neoadjuvant breast cancer study (BEAUTY) that consists of multi-omics data, clinical data, and patient-derived xenografts (PDXs). In the TCGA breast cancer study we obtained survival data to determine the cases and matched-controls; and in the BEAUTY, we used pathologic complete response (pCR) as an outcome to determine responders and non-responders. Recurrent targetable alterations were not enriched in patients without pCR in the BEAUTY study. We have applied the PANOPLY to non-responder patients to identify individual specific alterations, dysregulated networks, drug targets, and drugs for each patient and stored them as case reports in CORPUS (Computational Oncology Reports and Precision therapeUticS), a web-based repository that allows clinicians to review genomic reports. Using comprehensive “omic” data derived from a triple negative breast cancer patient who had pre and post-neoadjuvant chemotherapy PDXs, PANOPLY prioritized the PARP inhibitors as the top class of drug. Using the PDX models available from this patient, we tested olaparib and confirmed the in vivo antitumor activity (more effective than vehicle with a p-value < 0.05 in the PDXs). Further studies to confirm PANOPLY findings are currently underway. Conclusions: In summary, the PANOPLY and CORPUS systems incorporate molecular data together with clinical data to provide genomic reports with proposed drug targets to advance or enable precision breast cancer care. Citation Format: Kalari KR, Sinnwell JP, Thompson KJ, Tang X, Carlson EE, Alaparthi T, Yu J, Vedell PT, Kalmbach MT, Bockol MA, Hossain A, Weinshilboum RM, Boughey JC, Wang L, Suman VJ, Goetz MP. Multiscale modeling of omics data for precision breast cancer treatment [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P3-06-10.

Breast cancer is a heterogeneous disease, and unsupervised clustering approaches using gene expression data have identified 3-6 distinct subtypes of triple negative breast cancer (TNBC). A genomically and clinically distinct subtype of TNBC is referred to as LAR (Luminal Androgen Receptor). Tumors with this subtype typically express high levels of the AR and exhibit alterations within genes involved in the PI3K pathway (e.g. PIK3CA mutations). Prospective studies are underway using drugs that target the AR alone or in combination with PI3K and CDK 4/6 inhibitors. Given the importance of accurately identifying this subtype, we sought to develop an online tool that uses submitted gene expression data to confidently characterize LAR samples by corroborating the classification with previously published clustering approaches. Methods: We have investigated TNBC RNA-Seq data from The Cancer Genome Atlas (TCGA) breast cancer study (N=123 samples) by cluster analysis. Analysis of the average silhouette width in both biased and unbiased K-means clustering approaches demonstrated LAR and basal as two distinct and significant clusters. A shrunken centroid model of 426 differentially expressed genes, named as CABAL (Clustering Among BAsal and Luminal androgen receptor), was constructed by comparing LAR and basal subtypes. Results: We applied the CABAL model to classify the four TNBC microarray datasets that were previously used in clustering experiments as well as an independent RNA-Seq data cohort. Non-negative matrix factorization (NMF) and fuzzy clustering were applied to the samples (N=1046). Clustering similarity among the methods was assessed with the adjusted rand index, and CABAL demonstrated significant similarity with both fuzzy and NMF clustering methods. Similarly, hierarchical clustering analysis performed on the pooled cohort of 1046 samples recapitulated the CABAL classification with an area under the receiver operating curve of 0.91. Conclusions: Confident and robust identification of samples with the LAR phenotype is paramount in the assessment of clinical associations and therapeutic efficacy. To facilitate LAR identification, we have provided a web-based prediction tool of the CABAL classification, integrated with the NMF and fuzzy clustering results to identify candidate LAR samples. The end user is provided with the pair-wise adjusted rand indexes, thus reinforcing in the clustering characterizations. Further, our online LAR depiction tool provides a set of graphical and tabular summaries, which will be illustrated, while providing additional molecular characterizations of the PAM50 and Metabric classifications. The availability of this tool could advance the genomic research and treatment of TNBC patients. Citation Format: Thompson KJ, Alaparthi T, Sinnwell JP, Carlson EE, Tang X, Bockol M, Vedell PT, Ingle JN, Suman V, Weinshilboum RM, Wang L, Boughey JC, Kalari KR, Goetz MP. Molecular subtyping of androgen receptor-positive patients using gene expression profiles [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P1-03-04.

Background Use of the Genome Analysis Toolkit (GATK) continues to be the standard practice in genomic variant calling in both research and the clinic. Recently the toolkit has been rapidly evolving. Significant computational performance improvements have been introduced in GATK3.8 through collaboration with Intel in 2017. The first release of GATK4 in early 2018 revealed rewrites in the code base, as the stepping stone toward a Spark implementation. As the software continues to be a moving target for optimal deployment in highly productive environments, we present a detailed analysis of these improvements, to help the community stay abreast with changes in performance. Results We re-evaluated multiple options, such as threading, parallel garbage collection, I/O options and data-level parallelization. Additionally, we considered the trade-offs of using GATK3.8 and GATK4. We found optimized parameter values that reduce the time of executing the best practices variant calling procedure by 29.3% for GATK3.8 and 16.9% for GATK4. Further speedups can be accomplished by splitting data for parallel analysis, resulting in run time of only a few hours on whole human genome sequenced to the depth of 20X, for both versions of GATK. Nonetheless, GATK4 is already much more cost-effective than GATK3.8. Thanks to significant rewrites of the algorithms, the same analysis can be run largely in a single-threaded fashion, allowing users to process multiple samples on the same CPU. Conclusions In time-sensitive situations, when a patient has a critical or rapidly developing condition, it is useful to minimize the time to process a single sample. In such cases we recommend using GATK3.8 by splitting the sample into chunks and computing across multiple nodes. The resultant walltime will be nnn.4 hours at the cost of $41.60 on 4 c5.18xlarge instances of Amazon Cloud. For cost-effectiveness of routine analyses or for large population studies, it is useful to maximize the number of samples processed per unit time. Thus we recommend GATK4, running multiple samples on one node. The total walltime will be ∼34.1 hours on 40 samples, with 1.18 samples processed per hour at the cost of $2.60 per sample on c5.18xlarge instance of Amazon Cloud.

As reliable, efficient genome sequencing becomes more ubiquitous, the need for similarly reliable and efficient variant calling becomes increasingly important. The Genome Analysis Toolkit (GATK), maintained by the Broad Institute, is currently the widely accepted standard for variant calling software. However, alternative solutions may provide faster variant calling without sacrificing accuracy. One such alternative is Sentieon DNASeq, a toolkit analogous to GATK but built on a highly optimized backend. We evaluated the DNASeq single-sample variant calling pipeline in comparison to that of GATK. Our results confirm the near-identical accuracy of the two software packages, showcase perfect scalability and great speed from Sentieon, and describe computational performance considerations for the deployment of Sentieon DNASeq.

Use of the Genome Analysis Toolkit (GATK) continues to be the standard practice in genomic variant calling in both research and the clinic. Recently the toolkit has been rapidly evolving. Significant computational performance improvements have been introduced in GATK3.8 through collaboration with Intel in 2017. The first release of GATK4 in early 2018 revealed significant rewrites in the code base, as the stepping stone toward a Spark implementation. As the software continues to be a moving target for optimal deployment in highly productive environments, we present a detailed analysis of these improvements, to help the community stay abreast with changes in performance. We re-evaluated the options previously identified as advantageous, such as threading, parallel garbage collection, I/O options and data-level parallelization. Based on our results, we consider the performance and cost trade-offs of using GATK3.8 and GATK4 for different types of analyses.

Following publication of the original article [1], the author explained that Table 2 is displayed incorrectly. The correct Table 2 is given below. The original article has been corrected.

The majority of colorectal cancer (CRC) arises from precursor lesions known as polyps. The molecular determinants that distinguish benign from malignant polyps remain unclear. To molecularly characterize polyps, we utilized Cancer Adjacent Polyp (CAP) and Cancer Free Polyp (CFP) patients. CAPs had tissues from the residual polyp of origin and contiguous cancer; CFPs had polyp tissues matched to CAPs based on polyp size, histology and dysplasia. To determine whether molecular features distinguish CAPs and CFPs, we conducted Whole Genome Sequencing, RNA-seq, and RRBS on over 90 tissues from 31 patients. CAPs had significantly more mutations, altered expression and hypermethylation compared to CFPs. APC was significantly mutated in both polyp groups, but mutations in TP53, FBXW7, PIK3CA, KIAA1804 and SMAD2 were exclusive to CAPs. We found significant expression changes between CAPs and CFPs in GREM1, IGF2, CTGF, and PLAU, and both expression and methylation alterations in FES and HES1. Integrative analyses revealed 124 genes with alterations in at least two platforms, and ERBB3 and E2F8 showed aberrations specific to CAPs across all platforms. These findings provide a resource of molecular distinctions between polyps with and without cancer, which have the potential to enhance the diagnosis, risk assessment and management of polyps.

Blood-brain barrier (BBB) is a monolayer of endothelial cells that line brain capillaries. The BBB protects brain by blocking the entry of harmful substances from blood and shielding the brain from peripheral fluctuations in hormones, fatty acids, and electrolytes. In addition, the BBB effectively clears brain metabolites and serves as a major conduit for the delivery of crucial nutrients and growth factors needed for proper brain function. Owing to these critical responsibilities, any functional and structural impairment of the BBB may result in severe pathophysiological consequences in the brain. BBB dysfunction is implicated in several neurodegenerative disorders including Alzheimer's disease (Carmeliet and De Strooper, 2012), Parkinson's disease (Kortekaas et al., 2005), and cerebrovascular diseases (Yang and Rosenberg, 2011) such as cerebral amyloid angiopathy, stroke, and vascular dementia. Hence, the research community has been actively investigating the cerebrovascular contributions to neurological diseases with major emphasis on the BBB. The success of these efforts is heavily dependent upon the availability of reliable in vitro as well as in vivo BBB models. Polarized monolayers of human cerebrovascular endothelial cells (hCMEC/D3) described in the current work serves as one such in vitro model that can be easily cultured and manipulated in the lab (Poller et al., 2008; Vu et al., 2009; Weksler et al., 2013). The barrier properties and the expression of several classes of receptors, transporters, and enzymes in hCMEC/D3 cells have been previously investigated (Urich et al., 2012; Lopez-Ramirez et al., 2013, 2014; Bamji-Mirza et al., 2014; Ilina et al., 2015; Naik et al., 2015; Sajja and Cucullo, 2015). Thus, far, the genomic data for hCMEC/D3 cell lines have been generated using array-based approaches (Lopez-Ramirez et al., 2013). However, a comprehensive transcriptomic landscape of hCMEC/D3 cells, which is required for investigating molecular mechanisms using sophisticated computational biology approaches, is not currently available. Next-generation sequencing technology unveils the full potential of systems biology approaches to resolve cellular and molecular interaction networks that regulate the functional integrity of the BBB. Such a panoramic view of the interaction networks could enable us to isolate key players regulating a physiological process and investigate how they are affected in various diseases. To our knowledge, we are the first group to generate deep RNA sequencing and microRNA sequencing of a human BBB cell line. This data report describes BBBomics hub as a comprehensive portal for BBB transcriptomics data, obtained by sequencing mRNA (mRNA-seq) and microRNA (miRNA-seq) of polarized hCMEC/D3 cell monolayers. This data encompasses coding (gene expression, alternate splice forms, expressed single nucleotide variants -eSNVs) and non-coding (microRNA, LincRNA, circular RNA) counts that are easily accessible through BBBomics hub database. We also superimposed the RNA-seq coding data on 285 Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, which include canonical, non-canonical, and/or atypical pathways retrievable using BBBomics hub. The data is easily accessible and freely available at http://bioinformaticstools.mayo.edu/bbbomics/.