Search

Supplementary Figure S4

Genomic studies of pediatric cancer have primarily focused on specific tumor types or high-risk disease. Here, we used a three-platform sequencing approach, including whole-genome sequencing (WGS), whole-exome sequencing (WES), and RNA sequencing (RNA-seq), to examine tumor and germline genomes from 309 prospectively identified children with newly diagnosed (85%) or relapsed/refractory (15%) cancers, unselected for tumor type. Eighty-six percent of patients harbored diagnostic (53%), prognostic (57%), therapeutically relevant (25%), and/or cancer-predisposing (18%) variants. Inclusion of WGS enabled detection of activating gene fusions and enhancer hijacks (36% and 8% of tumors, respectively), small intragenic deletions (15% of tumors), and mutational signatures revealing of pathogenic variant effects. Evaluation of paired tumor–normal data revealed relevance to tumor development for 55% of pathogenic germline variants. This study demonstrates the power of a three-platform approach that incorporates WGS to interrogate and interpret the full range of genomic variants across newly diagnosed as well as relapsed/refractory pediatric cancers.Significance:Pediatric cancers are driven by diverse genomic lesions, and sequencing has proven useful in evaluating high-risk and relapsed/refractory cases. We show that combined WGS, WES, and RNA-seq of tumor and paired normal tissues enables identification and characterization of genetic drivers across the full spectrum of pediatric cancers.This article is highlighted in the In This Issue feature, p. 2945

Effective data sharing is key to accelerating research to improve diagnostic precision, treatment efficacy, and long-term survival in pediatric cancer and other childhood catastrophic diseases. We present St. Jude Cloud (https://www.stjude.cloud), a cloud-based data-sharing ecosystem for accessing, analyzing, and visualizing genomic data from >10,000 pediatric patients with cancer and long-term survivors, and >800 pediatric sickle cell patients. Harmonized genomic data totaling 1.25 petabytes are freely available, including 12,104 whole genomes, 7,697 whole exomes, and 2,202 transcriptomes. The resource is expanding rapidly, with regular data uploads from St. Jude's prospective clinical genomics programs. Three interconnected apps within the ecosystem—Genomics Platform, Pediatric Cancer Knowledgebase, and Visualization Community—enable simultaneously performing advanced data analysis in the cloud and enhancing the Pediatric Cancer knowledgebase. We demonstrate the value of the ecosystem through use cases that classify 135 pediatric cancer subtypes by gene expression profiling and map mutational signatures across 35 pediatric cancer subtypes.Significance:To advance research and treatment of pediatric cancer, we developed St. Jude Cloud, a data-sharing ecosystem for accessing >1.2 petabytes of raw genomic data from >10,000 pediatric patients and survivors, innovative analysis workflows, integrative multiomics visualizations, and a knowledgebase of published data contributed by the global pediatric cancer community.This article is highlighted in the In This Issue feature, p. 995

Involves Supplementary Table Legends and Supplementary Figures and associated legends

We present a comprehensive and versatile toolkit, NetBID2, for integrating multi-omics data and exposing hidden drivers that are not genetically altered or differentially expressed at the RNA/protein level by reverse-engineering context-specific interactomes and inferring protein activity. We deploy NetBID2 Viewer, Runner, and Cloud apps with 145 context-specific gene regulatory and signalling networks across normal tissues and paediatric and adult cancers to facilitate real-time interactive visualization, analysis, end-to-end use, and sharing.

Effective data sharing is key to accelerating research to improve diagnostic precision, treatment efficacy, and long-term survival in pediatric cancer and other childhood catastrophic diseases. We present St. Jude Cloud (https://www.stjude.cloud), a cloud-based data-sharing ecosystem for accessing, analyzing, and visualizing genomic data from >10,000 pediatric patients with cancer and long-term survivors, and >800 pediatric sickle cell patients. Harmonized genomic data totaling 1.25 petabytes are freely available, including 12,104 whole genomes, 7,697 whole exomes, and 2,202 transcriptomes. The resource is expanding rapidly, with regular data uploads from St. Jude's prospective clinical genomics programs. Three interconnected apps within the ecosystem—Genomics Platform, Pediatric Cancer Knowledgebase, and Visualization Community—enable simultaneously performing advanced data analysis in the cloud and enhancing the Pediatric Cancer knowledgebase. We demonstrate the value of the ecosystem through use cases that classify 135 pediatric cancer subtypes by gene expression profiling and map mutational signatures across 35 pediatric cancer subtypes. Significance: To advance research and treatment of pediatric cancer, we developed St. Jude Cloud, a data-sharing ecosystem for accessing >1.2 petabytes of raw genomic data from >10,000 pediatric patients and survivors, innovative analysis workflows, integrative multiomics visualizations, and a knowledgebase of published data contributed by the global pediatric cancer community. This article is highlighted in the In This Issue feature, p. 995

To discover driver fusions beyond canonical exon-to-exon chimeric transcripts, we develop CICERO, a local assembly-based algorithm that integrates RNA-seq read support with extensive annotation for candidate ranking. CICERO outperforms commonly used methods, achieving a 95% detection rate for 184 independently validated driver fusions including internal tandem duplications and other non-canonical events in 170 pediatric cancer transcriptomes. Re-analysis of TCGA glioblastoma RNA-seq unveils previously unreported kinase fusions (KLHL7-BRAF) and a 13% prevalence of EGFR C-terminal truncation. Accessible via standard or cloud-based implementation, CICERO enhances driver fusion detection for research and precision oncology. The CICERO source code is available at https://github.com/stjude/Cicero.

Detection of gene fusions is important for discovery of cancer drivers and clinical oncology testing, but existing software tools for fusion detection usually take hours to run and may fail to find lowly expressed fusions. To overcome these limitations, we developed the Fuzzion2 program, which uses pattern matching to detect known gene fusions in unmapped paired-read RNA-Seq data. Given a set of patterns representing fusion transcript breakpoints, Fuzzion2 finds every read pair matching any of the patterns. Both exact and inexact (fuzzy) matches are detected; the fuzzy matching tolerates variations caused by sequencing errors, SNVs, and indels. By employing a novel index of frequency minimizers, Fuzzion2 needs only minutes to process a sample. We have also developed pipelines to produce patterns for Fuzzion2, from fusion contig sequences, from genomic breakpoints in DNA and RNA, and from fusion protein sequences. To evaluate its applicability in clinical testing, we ran Fuzzion2 on ~2,000 RNA-seq samples profiled by the St. Jude clinical genomics program and confirmed its sensitivity in identifying lowly expressed fusions, such as KIAA1549-BRAF in low-grade glioma, which are frequently missed by commonly used fusion detection programs. Notably, Fuzzion2 detected a subclonal BCR-ABL1 fusion expressed at 1% and 6% of the wild-type BCR and ABL1 transcription level, respectively, in a B-lineage ALL sample that also has an IGH-CRLF fusion. Processing RNA-seq data from BCR-ABL1 cell lines, K562 with p210 fusion and OP1 with p190 fusion, diluted at 1:10, 1:100, and 1:1000 showed that Fuzzion2 can detect the fusion at 1:10-1:100 dilution, achieving a sensitivity 10 times greater than that of other fusion detection programs. We also evaluated the performance of Fuzzion2 for large-scale data mining in a study to compare the prevalence of gene fusions in pediatric versus adult cancers. We assembled a set of 15,474 patterns representing 5,480 fusions identified in the Pediatric Cancer Genome Project, NCI TARGET, clinical sequencing, and the COSMIC database. Fuzzion2 was deployed to the NCI Cancer Genomics Cloud and analyzed 9,464 TCGA RNA-seq samples from adult solid and brain tumors. Processing took an average of 6 minutes at a cost of only US$0.16 per sample. Among the 105 recurrent fusions identified in pediatric cancers, only 11 were also found in adult cancers. These shared fusions can be classified into two categories: 1) gene fusions present in cancers that occur in both children and young adults, e.g., synovial sarcoma, papillary thyroid cancer, and fibrolamellar hepatocellular carcinoma; and 2) kinase fusions involving ABL1, NTRK, and FGFR. Our experience with Fuzzion2 demonstrates that it is a powerful tool for time-critical clinical application and large-scale data mining. It is publicly available at https://github.com/stjude/fuzzion2. Citation Format: Stephen V. Rice, Michael N. Edmonson, Liqing Tian, Michael Rusch, David A. Wheeler, Jennifer L. Neary, Scott Newman, Lu Wang, Patrick R. Blackburn, Michael Macias, Andrew Thrasher, Jian Wang, Mark R. Wilkinson, Xin Zhou, Jinghui Zhang. Fuzzion2: Fast, sensitive detection of known gene fusions by fuzzy pattern matching for clinical testing and large-scale data mining [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4092.

Scientific visualization is important to cancer research as different graphs enable researchers to spot trends and detect outliers in large-scale, high-dimensional data. For example, mutation landscape maps integrate mutational and clinical data, revealing which somatic alterations are most closely associated with diagnoses, age, ethnicity, and outcome; t-SNE plots allow dissection of cancer subtypes based on gene expression or methylation profiling. These graphs are often created by researchers after extensive data curation and optimization of the layout using software tools designed for data visualization and are commonly presented as static figures in published literature. To permit seamless access to the genomic and epigenomic visualization tools created at St. Jude, we developed Visualization Community (https://viz.stjude.cloud/, VisComm), which allows researchers to explore published graphs, develop custom visualization for their private data, and also serve as a public repository for completed graphs, as part of the St. Jude Cloud data sharing ecosystem. Sixty-five (65) interactive graphs from 31 publications in 231 pediatric subtypes are currently available, including t-SNE plots of RNA-seq and methylation data, mutation landscape maps, variant visualization on protein or genomic axes, epigenome states, and chromatin contact interactions. The graphs are searchable by cancer diagnosis, molecular profiling, research interest, publication, and research organization with features designed to enable user customization. To date, 5,014 users, 25% of whom are recurrent, have accessed VisComm. On average, each public visualization has over 4,000 views while the most accessed graph has >13,000 views. We have developed the ability to work in teams with in VisComm to empower collaborative development of new visualizations and controlled release by a research team. This feature enables the professional quality visualizations serve as resources for focused scientific communities such as the Audacious Goals Initiative (AGI) - Retina), a multi-institutional cross-disciplinary research team whose goal is restoring vision to patients by regenerating retinas. Videos and textual tutorials are being prepared using data generated from programs such as the Genome for Kids (G4K), a clinical genomic profiling program, that will help guide users to apply the underlying visualization software tools and to integrate public and private data of all types. VisComm is the first platform dedicated to enabling the development and sharing of interactive high-quality data-rich visualizations, which we expect will enhance data exploration and hypothesis generation for cancer research. Citation Format: David B. Finkelstein, James Madson, Lucian Vacaroiu, Alexander M. Gout, Leonard Hirja, Stephanie Sandor, Andrew Thrasher, Colleen Reilly, Jian Wang, Xiaolong Chen, Xin Zhou, Jinghui Zhang, Clay McLeod. VisComm: A cloud-based platform for accessing and creating scientific visualization for the pediatric cancer research community [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1901.

Individuals with monogenic disorders can experience variable phenotypes that are influenced by genetic variation. To investigate this in sickle cell disease (SCD), we performed whole-genome sequencing (WGS) of 722 individuals with hemoglobin HbSS or HbSβ0-thalassemia from Baylor College of Medicine and from the St. Jude Children’s Research Hospital Sickle Cell Clinical Research and Intervention Program (SCCRIP) longitudinal cohort study. We developed pipelines to identify genetic variants that modulate sickle hemoglobin polymerization in red blood cells and combined these with pain-associated variants to build a polygenic score (PGS) for acute vaso-occlusive pain (VOP). Overall, we interrogated the α-thalassemia deletion −α3.7 and 133 candidate single-nucleotide polymorphisms (SNPs) across 66 genes for associations with VOP in 327 SCCRIP participants followed longitudinally over 6 years. Twenty-one SNPs in 9 loci were associated with VOP, including 3 (BCL11A, MYB, and the β-like globin gene cluster) that regulate erythrocyte fetal hemoglobin (HbF) levels and 6 (COMT, TBC1D1, KCNJ6, FAAH, NR3C1, and IL1A) that were associated previously with various pain syndromes. An unweighted PGS integrating all 21 SNPs was associated with the VOP event rate (estimate, 0.35; standard error, 0.04; P = 5.9 × 10−14) and VOP event occurrence (estimate, 0.42; standard error, 0.06; P = 4.1 × 10−13). These associations were stronger than those of any single locus. Our findings provide insights into the genetic modulation of VOP in children with SCD. More generally, we demonstrate the utility of WGS for investigating genetic contributions to the variable expression of SCD-associated morbidities.

Genomic studies of pediatric cancer have primarily focused on specific tumor types or high-risk disease. Here, we used a three-platform sequencing approach, including whole-genome sequencing (WGS), whole-exome sequencing (WES), and RNA sequencing (RNA-seq), to examine tumor and germline genomes from 309 prospectively identified children with newly diagnosed (85%) or relapsed/refractory (15%) cancers, unselected for tumor type. Eighty-six percent of patients harbored diagnostic (53%), prognostic (57%), therapeutically relevant (25%), and/or cancer-predisposing (18%) variants. Inclusion of WGS enabled detection of activating gene fusions and enhancer hijacks (36% and 8% of tumors, respectively), small intragenic deletions (15% of tumors), and mutational signatures revealing of pathogenic variant effects. Evaluation of paired tumor–normal data revealed relevance to tumor development for 55% of pathogenic germline variants. This study demonstrates the power of a three-platform approach that incorporates WGS to interrogate and interpret the full range of genomic variants across newly diagnosed as well as relapsed/refractory pediatric cancers. Significance: Pediatric cancers are driven by diverse genomic lesions, and sequencing has proven useful in evaluating high-risk and relapsed/refractory cases. We show that combined WGS, WES, and RNA-seq of tumor and paired normal tissues enables identification and characterization of genetic drivers across the full spectrum of pediatric cancers. This article is highlighted in the In This Issue feature, p. 2945

Effective data sharing is key to accelerating research that will improve the precision of diagnoses, efficacy of treatments and long-term survival of pediatric cancer and other childhood catastrophic diseases. We present St. Jude Cloud (https://www.stjude.cloud), a cloud-based data sharing ecosystem developed via collaboration between St. Jude Children’s Research Hospital, DNAnexus, and Microsoft, for accessing, analyzing and visualizing genomic data from >10,000 pediatric cancer patients, long-term survivors of pediatric cancer and >800 pediatric sickle cell patients. Harmonized genomic data totaling 1.25 petabyes on St. Jude Cloud include 12,104 whole genomes, 7,697 whole exomes and 2,202 transcriptomes, which are freely available to researchers worldwide. The resource is expanding rapidly with regular data uploads from St. Jude’s prospective clinical genomics programs, providing public access as soon as possible rather than holding data back until publication. Three interconnected apps within the St. Jude Cloud ecosystem—Genomics Platform, Pediatric Cancer Knowledgebase (PeCan) and Visualization Community—provide a unique experience for simultaneously performing advanced data analysis in the cloud and enhancing the pediatric cancer knowledgebase. We demonstrate the value of the St. Jude Cloud ecosystem through use cases that classify 48 pediatric cancer subtypes by gene expression profiling and map mutational signatures across 35 subtypes of pediatric cancer.

Although genome-wide DNA methylomes have demonstrated their clinical value as reliable biomarkers for tumor detection, subtyping, and classification, their direct biological impacts at the individual gene level remain elusive. Here we present MethylationToActivity (M2A), a machine learning framework that uses convolutional neural networks to infer promoter activities based on H3K4me3 and H3K27ac enrichment, from DNA methylation patterns for individual genes. Using publicly available datasets in real-world test scenarios, we demonstrate that M2A is highly accurate and robust in revealing promoter activity landscapes in various pediatric and adult cancers, including both solid and hematologic malignant neoplasms. Supplementary Information The online version contains supplementary material available at 10.1186/s13059-020-02220-y.

To evaluate the potential of an integrated clinical test to detect diverse classes of somatic and germline mutations relevant to pediatric oncology, we performed three-platform whole-genome (WGS), whole exome (WES) and transcriptome (RNA-Seq) sequencing of tumors and normal tissue from 78 pediatric cancer patients in a CLIA-certified, CAP-accredited laboratory. Our analysis pipeline achieves high accuracy by cross-validating variants between sequencing types, thereby removing the need for confirmatory testing, and facilitates comprehensive reporting in a clinically-relevant timeframe. Three-platform sequencing has a positive predictive value of 97–99, 99, and 91% for somatic SNVs, indels and structural variations, respectively, based on independent experimental verification of 15,225 variants. We report 240 pathogenic variants across all cases, including 84 of 86 known from previous diagnostic testing (98% sensitivity). Combined WES and RNA-Seq, the current standard for precision oncology, achieved only 78% sensitivity. These results emphasize the critical need for incorporating WGS in pediatric oncology testing., Clinical oncology is rapidly adopting next-generation sequencing technology for nucleotide variant and indel detection. Here the authors present a three-platform approach (whole-genome, whole-exome, and whole-transcriptome) in pediatric patients for the detection of diverse types of germline and somatic variants.

SummaryXenografts are important models for cancer research and the presence of mouse reads in xenograft next generation sequencing data can potentially confound interpretation of experimental results. We present an efficient, cloud-based BAM-to-BAM cleaning tool called XenoCP to remove mouse reads from xenograft BAM files. We show application of XenoCP in obtaining accurate gene expression quantification in RNA-seq and tumor heterogeneity in WGS of xenografts derived from brain and solid tumors.Availability and ImplementationSt. Jude Cloud (https://pecan.stjude.cloud/permalink/xenocp) and St. Jude Github (https://github.com/stjude/XenoCP)

Clinical genomic studies of pediatric cancer have primarily focused on specific tumor types or high-risk disease. In the Genomes for Kids study (NCT02530658) we used a three-platform sequencing approach, including whole genome (WGS), whole exome (WES) and RNA sequencing, to examine tumor and paired germline genomes from prospectively identified children with cancer. The goal of the study was to assess the potential of comprehensive next generation sequencing to elucidate the molecular mechanisms underlying tumor formation and investigate the potential of this information to influence clinical decision-making.The cohort, with a median age of 6 yrs, range 0 - 26 yrs, included 301 patients with newly diagnosed (85%) or relapsed/refractory (15%) cancers, unselected for tumor type or stage. Patients with hematologic malignancies accounted for 41% of cases, 31% had CNS tumors, and 28% had other non-CNS solid tumors. This cohort also included 18 patients with very rare tumor types, defined here as occurring in less than 2 cases per million person per year.Two hundred fifty three patients (84%) had sufficient tumor for three-platform sequencing and all 301 had adequate paired germline samples. Following analysis, 86% of patients harbored diagnostic (53%), prognostic (57%), therapeutically relevant (25%), and/or cancer predisposing (18%) variants. The inclusion of WGS enabled detection of oncogenic gene fusions, as well as 22 cases in which oncogenes were activated through enhancer hijacking, a particularly frequent occurrence in hematologic malignancies. In addition, WGS effectively detected clinically relevant small intragenic deletions (15% of tumors) and a variety of mutational signatures, which were not detectable through analysis of whole exome data. Evaluation of 56 pathogenic germline variants in the context of paired tumor sequence data helped establish the disease relevance of several genes that are not typically associated with the cancer type in question, providing critical insights on a case-by-case basis. Examples include a pathogenic germline variant in MUTYH in a patient with retinoblastoma whose tumor exhibited a mutation signature associated with reactive oxygen species indicative of loss of MUTYH function; and conversely, a likely pathogenic variant in PMS2 in a rare brain cancer, which did not exhibit a mutation signature associated with microsatellite instability. This study successfully demonstrated the power of this three-platform approach to interrogate and interpret the full range of genomic variants across newly diagnosed as well as relapsed/refractory pediatric cancers. As a result of these findings, we have incorporated this three-platform approach into our routine real-time clinical service at St. Jude Children's Hospital. Citation Format: David A. Wheeler, Scott Newman, Joy Nakitandwe, Chimene A. Kesserwan, Elizabeth M. Azzato, Michael C. Rusch, Sheila Shurtleff, Armita Bahrami, Brent Orr, Jeffery M. Klco, Dale J. Hedges, Kayla V. Hamilton, Scott G. Foy, Michael N. Edmonson, Andrew Thrasher, Jiali Gu, Lynn W. Harrison, Lu Wang, Roya Mostafavi, Manish Kubal, Jamie Maciaszek, Michael Clay, Annastasia Ouma, Antonina Silkov, Yanling Liu, Zhaojie Zhang, Yu Liu, Samuel W. Brady, Xin Zhou, Mark Wilkinson, Delaram Rahbarinia, Jay Knight, Jian Wang, Charles G. Mullighan, Rose B. McGee, Emily A. Quinn, Elsie L. Gerhardt, Leslie M. Taylor, Regina Nuccio, Jessica M. Valdez, Stacy J. Hines-Dowell, Alberto Pappo, Giles Robinson, Liza-Marie Johnson, Ching-Hon Pui, David W. Ellison, James R. Downing, Jinghui Zhang, Kim E. Nichols. Genomes for Kids: Comprehensive DNA and RNA sequencing defining the scope of actionable mutations in pediatric cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 642.

Next-generation sequencing-based genomic profiling is now a mainstay of pediatric oncology research and clinical testing. Correlating genomic features of patient cancer genomes with curated data extracted from large reference cohorts is critical for identifying molecular subtypes and underlying mutagenesis processes. To facilitate such investigation, we developed two user-friendly workflows on St. Jude Cloud, a data sharing ecosystem hosting genomic data for >10,000 pediatric cancer patients and survivors. These workflows leverage St. Jude Cloud comprehensive pediatric cancer genomic data, including 1,616 RNA-seq of 135 cancer subtypes and 958 whole genome sequencing (WGS) of 35 subtypes, to enable user analysis of their data in the context of St. Jude Cloud cohorts without a need to download large datasets. The RNA-Seq Expression Classification workflow enables a user to compare their patient RNA-Seq gene expression data with blood (832), brain (456), and solid tumor (319) pediatric cancer reference cohorts and PDX models (45), enabling subtype classification using t-Distributed Stochastic Neighbor Embedding (t-SNE). Reference cohorts include curated subtype-defining somatic alterations integrating genomic variant data with expression profile. Resulting interactive t-SNE plots can be explored and annotated - with options to highlight cancer subtypes or samples and display sample information (age of onset, clinical diagnosis, molecular driver). To demonstrate, we analyze PAWNXH, a Children's Oncology Group AML sample with a novel ZBTB7A-NUTM1 fusion and find it clusters with AML samples harboring KMT2A re-arrangements suggesting a potential mechanism of pathogenesis. Integrating PDX samples enables model selection for functional experiments by connecting patient subtypes with mouse models. The Mutational Signatures workflow identifies and quantifies COSMIC mutational signatures in user-uploaded somatic VCF files for comparison to reference pediatric cancer cohorts. The interactive interface enables rapid identification of signatures within the query cohort and facilitates comparison to the reference using a cohort-level summary view. Identified signatures may also be explored at the sample-level for both query and reference cohorts, enabling the user to identify samples with signatures of interest for further analysis. We show an example comparison of mutational signatures identified in pediatric and adult AML samples. These workflows enable users to leverage curated pediatric cancer data to make discoveries in their own samples. Enabling point-and-click analysis in St. Jude Cloud removes the barrier for non-computational researchers and eliminates the need to download large reference datasets for local analysis. Both workflows utilize post-processed rather than raw genomic data, reducing transfer costs for uploading user data to the cloud. Citation Format: Andrew Thrasher, Michael Macias, Alexander M. Gout, Delaram Rahbarinia, Xin Zhou, Samuel W. Brady, Clay McLeod, Michael C. Rusch, Xiaolong Chen, Soheil Meshinchi, Michael A. Dyer, Suzanne J. Baker, Martine F. Roussel, Jinghui Zhang. Empowering point-and-click genomic analysis with large pediatric genomic reference data on St. Jude Cloud [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2289.

Although genome-wide DNA methylomes have demonstrated their clinical value as reliable biomarkers for tumor detection, subtyping, and classification, their direct biological impacts at the individual gene level remain elusive. Here we present MethylationToActivity (M2A), a machine learning framework that uses convolutional neural networks to infer promoter activities based on H3K4me3 and H3K27ac enrichment, from DNA methylation patterns for individual genes. Using publicly available datasets in real-world test scenarios, we demonstrate that M2A is highly accurate and robust in revealing promoter activity landscapes in various pediatric and adult cancers, including both solid and hematologic malignant neoplasms. Citation Format: Justin Williams, Beisi Xu, Daniel Putnam, Andrew Thrasher, Chunliang Li, Jun Yang, Xiang Chen. MethylationToActivity: A deep-learning framework that reveals promoter activity landscapes from DNA methylomes in individual tumors [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-046.

Gene fusions are important biomarkers for cancer diagnosis, subtype classification and therapeutic decision-making. While fusion detection using RNA-seq data has become a standard practice, existing computational methods primarily focus on identifying canonical exon-to-exon fusions. However, more complex events such as multi-partner fusions, truncations, enhancer hijacking and internal tandem duplications (ITD) can also lead to abnormal function or aberrant transcription of cancer driver genes. To aid discovery of complex and diverse driver fusions, we developed CICERO (CICERO Is Clipping Extended for RNA Optimization), a local assembly-based algorithm that integrates RNA-seq reads bearing aberrant mapping signatures with extensive annotation for ranking candidate fusions. Our benchmark data set, designed to support the main application of RNA-seq fusion analysis, consists of 184 driver fusions from 170 pediatric leukemia, solid tumor and brain tumor detected by paired tumor-normal WGS and orthogonally validated by capture sequencing, RT-PCR and/or FISH. CICERO detected 95% of these fusions with an average ranking of 1.9, whereas ChimeraScan, deFuse, FusionCatcher and STAR-Fusion detected only 63%, 66%, 77% and 63% with an average ranking of 37.0, 9.0, 18.1 and 4.4, respectively. Notably, events such as ITD and rearrangements involving the highly repetitive IGH locus were detected almost exclusively by CICERO. Our re-analysis of 167 RNA-seq data from the TCGA Glioblastoma Multiforme (GBM) cohort unveiled 158 fusions of cancer genes that were not reported previously. These include kinase fusions (KLHL7-BRAF), ITD of EGFR kinase domain and a 13% prevalence of EGFR C-terminal truncation compared to the 6% reported by the TCGA Network. CICERO has greatly improved our ability to discover non-canonical fusions which are overlooked by existing fusion detection methods, and has been used to analyze >2,000 RNA-seq samples generated by the two largest pediatric cancer genomics initiatives: the St. Jude/Washington University Pediatric Cancer Genome Project (PCGP) and the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) project. We anticipate that CICERO will also improve fusion analysis for adult cancer RNA-seq data, as demonstrated through our re-analysis of TCGA-GBM and our recent discovery of MAP3K8 C-terminal truncation fusion in 2% of TCGA melanoma samples. CICERO is accessible via standard (https://github.com/stjude/Cicero) or cloud-based (https://platform.stjude.cloud/tools/rapid_rna-seq) implementation. To further improve accuracy, fusions predicted by CICERO can be curated by FusionEditor (https://proteinpaint.stjude.org/FusionEditor/), an interactive viewer allowing inspection of protein domains involved in the fusion and the gene expression status of fusion-positive samples. Citation Format: Liqing Tian, Yongjin Li, Michael N. Edmonson, Xin Zhou, Scott Newman, Clay McLeod, Yu Liu, Bo Tang, Michael C. Rusch, John Easton, Jing Ma, Austyn Trull, J. Robert Michael, Andrew Thrasher, Charles Mullighan, Suzanne J. Baker, James R. Downing, David W. Ellison, Jinghui Zhang. CICERO: An accurate method for detecting complex and diverse driver fusions using cancer transcriptome sequencing (RNA-seq) data [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5478.

Purpose Childhood cancer survivors are at increased risk of subsequent neoplasms (SNs), but the germline genetic contribution is largely unknown. We assessed the contribution of pathogenic/likely pathogenic (P/LP) mutations in cancer predisposition genes to their SN risk. Patients and Methods Whole-genome sequencing (30-fold) was performed on samples from childhood cancer survivors who were ≥ 5 years since initial cancer diagnosis and participants in the St Jude Lifetime Cohort Study, a retrospective hospital-based study with prospective clinical follow-up. Germline mutations in 60 genes known to be associated with autosomal dominant cancer predisposition syndromes with moderate to high penetrance were classified by their pathogenicity according to the American College of Medical Genetics and Genomics guidelines. Relative rates (RRs) and 95% CIs of SN occurrence by mutation status were estimated using multivariable piecewise exponential regression stratified by radiation exposure. Results Participants were 3,006 survivors (53% male; median age, 35.8 years [range, 7.1 to 69.8 years]; 56% received radiotherapy), 1,120 SNs were diagnosed among 439 survivors (14.6%), and 175 P/LP mutations were identified in 5.8% (95% CI, 5.0% to 6.7%) of survivors. Mutations were associated with significantly increased rates of breast cancer (RR, 13.9; 95% CI, 6.0 to 32.2) and sarcoma (RR, 10.6; 95% CI, 4.3 to 26.3) among irradiated survivors and with increased rates of developing any SN (RR, 4.7; 95% CI, 2.4 to 9.3), breast cancer (RR, 7.7; 95% CI, 2.4 to 24.4), nonmelanoma skin cancer (RR, 11.0; 95% CI, 2.9 to 41.4), and two or more histologically distinct SNs (RR, 18.6; 95% CI, 3.5 to 99.3) among nonirradiated survivors. Conclusion The findings support referral of all survivors for genetic counseling for potential clinical genetic testing, which should be prioritized for nonirradiated survivors with any SN and for those with breast cancer or sarcoma in the field of prior irradiation.

The empirical aim of this paper is motivated by the anecdotal belief among the professional and non-professional investment community, that a “low” reading in the CBOE Volatility Index (VIX) or large decline alone are ample reasons to believe that volatility will spike in the near future. While the Volatility Index can be a useful tool for investors and traders, it is often misinterpreted and poorly used. This paper will demonstrate that the dispersion of the Volatility Index acts as a better predictor of its future VIX spikes.

Although sickle cell disease (SCD) is a monogenic disorder, the severity and specific organ dysfunction and failure are strongly influenced by genetic modifiers. Rapid identification of all modifiers in patients and well-phenotyped cohorts will better define the impact of relevant variants on clinical status, inform disease biology, and identify new therapeutic strategies. We created the Sickle Genome Project (SGP), a whole genome sequencing (WGS) strategy, to define genomic variation and modifiers of SCD. We performed WGS on 871 African American SCD patients from St. Jude Children's Research Hospital who participated in the Sickle Cell Clinical Research and Intervention Program (SCCRIP, Hankins et al. Pediatr Blood Cancer, 2018) and Texas Children's Hospital Hematology Center (TCHC). We developed robust pipelines for accurate detection of single nucleotide polymorphisms (SNPs), identification of structural variants and data retrieval/sharing via the St. Jude Cloud platform (to be described elsewhere). Notable findings include: 1) Confirmed associations of common genetic modifiers with SCD phenotypes, including levels of fetal hemoglobin (BCL11A, HBS1L-MYB, HBB), bilirubin (UGT1A1), and microalbuminuria (APOL1). Additional associations approaching genome-wide significance require further investigation, including replication in independent samples. 2) Improved determination of the SCD modifier α-thalassemia. The most common α-thalassemia mutations in SCD are 3.7 kb or 4.2 kb deletions (-α3.7 and -α4.2 alleles), which arose from recombination between homologous HBA1 and HBA2 genes and are difficult to map using standard WGS reads. Three independent crossover events are described for -α3.7 and one for -α4.2 in SCD cohorts. We developed a novel approach to identify α-globin gene deletions by local de novo assembly of WGS data and coverage depth analysis. We identified 5 -α3.7 alleles (frequencies 0.77-32.12%) and 7 -α4.2 alleles (frequencies 0.19-5.77%). Collectively, the frequency of all -α alleles was 57%, reflecting at least 12 distinct recombination events, greatly exceeding previously published counts. These findings better define the evolution of α-globin genes to allow improved understanding of their regulation and influence on SCD. 3) Characterization of β0-thalassemia alleles. Mutations in the extended β-globin locus influence SCD phenotypes. Five SGP patients had large β-globin (HBB) deletions associated with elevated fetal hemoglobin, which ameliorates symptoms of SCD. Twenty-three patients had HbSβ0-thalassemia, which reduces the severity of some SCD phenotypes. Overall, 48.6% (18/37) of patients clinically designated as HbSβ0 -thalassemia had no identified β-thalassemia mutation. Moreover, 4/680 patients (0.6%) designated HbSS were identified to be β0-thalassemia heterozygotes. The MCV, RBC and %HbA2 distributions overlapped substantially in correct vs. incorrect genotype assignments. Improved discrimination of HbSβ0 vs HbSS genotypes by WGS will better define associated phenotype differences to impact clinical care. 4) Determination of a genetic variant linked to vaso-occlusive crisis (VOC). Previously, a single GWAS study linked rs3115229, located 63.7 kb 5′ upstream of the KIAA1109 gene, with VOC at borderline significance (P = 5.63 × 10−8) (Chaturvedi et al, Blood 130, 2017). Using WGS data for 327 SGP participants (HbSS or HbSβ0-thalassemia) enrolled in the SCCRIPP study, we found strong association (p = 7 x 10-5) between the onset of VOC and a 4-SNP diplotype within an adjacent LD block of the KIAA1109-TENR-IL2-IL21 region (chr4: 122.8Mb - 123.8Mb) which has been previously associated with numerous inflammatory disorders. We validated this association using imputed genome-wide array data in an independent group of SCD patients (Sleep and Asthma Cohort, n= 181 patients, p = 0.05) (Cohen et al, Ann Am Thorac Soc, 2016). This works provides confirmation that the region surrounding KIAA1109 is associated with pain crisis in SCD. Our studies provide new information on the genomic architecture of SCD patients and delineate a consolidated approach for future applications of precision medicine. Disclosures Hankins: Novartis: Research Funding; Global Blood Therapeutics: Research Funding; NCQA: Consultancy; bluebird bio: Consultancy. Estepp:Global Blood Therapeutics: Consultancy, Research Funding; ASH Scholar: Research Funding; NHLBI: Research Funding; Daiichi Sankyo: Consultancy.

Bioinformatics researchers need efficient means to process large collections of genomic sequence data. One application of interest, genome assembly, has great potential for parallelization; however, most previous attempts at parallelization require uncommon high-end hardware. This paper introduces the Scalable Assembler at Notre Dame (SAND) framework that can achieve significant speedup using large numbers of commodity machines harnessed from clusters, clouds, and grids. SAND interfaces with the Celera open-source assembly toolkit, replacing two independent sequential modules with scalable parallel alternatives: the candidate selector exploits distributed memory capacity, and the sequence aligner exploits distributed computing capacity. For large problems, these modules provide robust task and data management while also achieving speedup with high efficiency. We show results for several data sets ranging from 738 thousand to over 320 million alignments using resources ranging from a small cluster to more than a thousand nodes spanning three institutions.

Weaver is a high-level distributed computing framework that enables researchers to construct scalable scientific data-processing workflows. Instead of developing a new workflow language, we introduce a domain-specific language built on top of Python called Weaver, which takes advantage of users' familiarity with the programming language, minimizes barriers to adoption, and allows for integration with a rich ecosystem of existing software. In this paper, we provide an overview of Weaver's programming model, which allows users to organize and specify scientific workflows by using a collection of datasets, functions, and abstractions. We also explain how these workflow specifications are compiled into a directed acyclic graph that is used by the Makeflow workflow manager to dispatch work to a variety of distributed execution platforms. To demonstrate the power and benefits of using the framework in constructing scientific research applications, the paper examines four distinct real-world applications scripted using Weaver and analyzes the performance, scalability, and impact of the distributed generated scientific workflows. Copyright © 2011 John Wiley & Sons, Ltd.

Childhood cancer survivors are at increased risk of subsequent neoplasms (SN), largely considered to be therapy-related. Studies of cancer predisposition genes (CPGs) and risk of SN among long-term survivors are lacking. We characterized germline mutations in CPGs in childhood cancer survivors to determine their contribution to SN risk. Whole genome (30x) and exome (100x) sequencing was performed for 2988 5+ year survivors of childhood cancer (1629 leukemia/lymphoma, 332 CNS, 1027 other solid tumors, 53% male, median follow-up 28 [range 6-55] years). Survivors underwent a comprehensive clinical assessment, treatment exposures were abstracted from medical records, and SN were validated by pathology reports. Germline mutations in 63 CPGs were classified using the American College of Medical Genetics and Genomics guidelines as previously described (Zhang et al. NEJM 2015). Logistic regression, adjusting for age, sex and race, was used to evaluate associations between mutation status, cancer therapy and the SN risk. 1062 SNs were diagnosed in 437 survivors, of whom 98 developed ≥2 histologically distinct SNs. Median age at SN and time to first SN was 38.2 (range 3.3-67.4) and 29.2 (0.9-48.4) years, respectively. Common SNs were basal cell carcinoma (542 in 153 survivors), meningioma (201 in 100), thyroid (64 in 64), and breast cancer (58 in 50). Cumulative incidence of SN at age 45 was 25.5% (95% CI: 22.9-27.9). 169 survivors (5.7%) had a pathogenic/likely pathogenic (P/LP) mutation in a CPG, consisting of 97 single nucleotide variations, 63 insertion/deletions and 9 copy number alterations (49% of mutations not in ClinVar). Frequently mutated genes were: RB1 (n=41), NF1 (n=22), BRCA2 (n=13), BRCA1 (n=12) and TP53 (n=10). Our data confirmed known associations between CPG mutations and specific primary diagnoses including RB1 mutations in 32 of 41 (78%) of bilateral and 7 of 57 (12%) of unilateral retinoblastoma survivors, 22 NF1 (20 of 332 CNS survivors), 4 SUFU (all in medulloblastoma survivors) and 5 WT1 mutations (all in Wilms’ tumor survivors). Analyses revealed novel associations between CPG mutations and SN risk. Among 1326 survivors not exposed to radiation therapy (non-RT), 62 SNs developed in 54 survivors, of which 15 (24.2%) occurred in P/LP mutation carriers. Non-RT exposed survivors with a P/LP mutation had an increased risk of SN (OR=5.6, 95% CI=2.6-12.0, P Citation Format: Zhaoming Wang, Carmen L. Wilson, John Easton, Dale Hedges, Qi Liu, Gang Wu, Michael Rusch, Michael Edmonson, Shawn Levy, Jennifer Q. Lanctot, Eric Caron, Kyla Shelton, Kelsey Currie, Matthew Lear, Heather L. Mulder, Donald Yergeau, Celeste Rosencrance, Bhavin Vadodaria, Yadav Sapkota, Russell J. Brooke, Wonjong Moon, Evadnie Rampersaud, Xiaotu Ma, Shuoguo Wang, Ti-Cheng Chang, Stephen Rice, Andrew Thrasher, Aman Patel, Cynthia Pepper, Xin Zhou, Xiang Chen, Wenan Chen, Angela Jones, Braden Boone, Deo Kumar Srivastava, Chimene A. Kesserwan, Kim E. Nichols, James R. Downing, Melissa M. Hudson, Yutaka Yasui, Leslie L. Robison, Jinghui Zhang. Germline mutations in cancer predisposition genes and risk for subsequent neoplasms among long-term survivors of childhood cancer in the St. Jude Lifetime Cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3001. doi:10.1158/1538-7445.AM2017-3001

Clusters, clouds, and grids offer access to large scale computational resources at low cost. This is especially appealing to scientific applications that require a very large scale to compete in the research space. However, the resources available across these platforms differ significantly in their availability, hardware, environment, performance, cost of use, and more. This requires the use of elastic applications that can adapt to the resources available at run-time, transparently handling heterogeneity and failures. In this paper, we present case studies of several elastic applications built using the Work Queue programming framework. From this experience, we offer six general guidelines for the design and implementation of elastic applications that run on thousands of processors.

Next-generation sequencing (NGS) of the whole genome, whole exome, and transcriptome has enabled characterization of genetic landscapes of multiple cancers. By analyzing over 2,000 pediatric cancer patients, we have developed a comprehensive database for recurrent somatic alterations and pathogenic germline mutations as part of the St. Jude/Washington University Pediatric Cancer Genome Project. However, there is no systematic evaluation on whether NGS is able to identify germline and somatic lesions reported by existing molecular diagnostic assays and what combination of NGS platforms is best suited for clinical sequencing. Here we report the first comprehensive study that employs whole-genome sequencing at 30-45X coverage, whole-exome sequencing at 100X coverage and transcriptome sequencing using matched tumor/normal samples from cancer patients. A pilot study was carried out to perform NGS analysis on 78 children of leukemia, solid tumor or brain tumor with a total of 112 diagnostic or prognostic biomarkers previously characterized by multiple molecular diagnostic assays. We implemented an analysis pipeline that integrates the genetic lesions detected by all three NGS platforms to characterize somatic and germline single nucleotide variations (SNVs), short insertions and deletions (indels), structural variations including fusions, karyotypes, copy number alterations, loss of heterozygosity, tumor purity and tumor-in-normal contamination. The turn-around time for data analysis is 2 weeks with an overall sensitivity of 99% on detecting known biomarkers. Extensive validation of >3,000 somatic sequence mutations or structural variations from 38 cases shows that the specificity for somatic SNV, indel and structural variation is at 98%, 95% and 84% across the genome. We demonstrate that in addition to providing cross-validation, multi-platform NGS is required for detecting all genetic lesions of pathological significance including complex re-arrangements such as chromothripsis. In addition to known pathogenic or likely pathogenic mutations, our analysis has also unveiled novel pathogenic mutations (e.g. a germline deletion in TP53 in one patient with medulloblastoma) and identified multiple variants of unknown significance that may be worth further exploration (e.g. an in-frame deletion of exons 3-9 of DNMT3A in one neuroblastoma). Our study demonstrates that NGS is able to detect a wide range of genetic lesions currently characterized by multiple molecular diagnostic assays, providing critical insight into the design of clinical sequencing for ongoing studies. Citation Format: Jinghui Zhang, Michael Rusch, Joy Nakitandwe, Zhaojie Zhang, Michael N. Edmonson, Matthew Parker, Xiaotu Ma, Jared Becksfort, Andrew Thrasher, Jiali Gu, Yongjin Li, Erin Hedlund, Aman Patel, John Easton, Donald Yergeau, Bhavin Vadodaria, Xiang Chen, Tanja A. Gruber, Rose McGee, David Ellison, Sheila Shurtleff, James R. Downing. Molecular diagnosis for pediatric cancer through integrative analysis of whole-genome, whole-exome and transcriptome sequencing. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 2628.

Next generation sequencing technologies have enabled various entities, ranging from large sequencing centers to individual laboratories, to sequence organisms of choice and analyze them on demand. Sequencing and analysis, however, is only part of the equation: to learn about a certain organism, scientists need to annotate it. Each of these problems is highly parallel at a basic level of computation; however, only a few applications support single parallelization frameworks such as MPI. Because of the overall increasing demand for computational analysis and the inherent parallelism available in these problems, applications should easily run on clusters, clouds, and/or grids (even simultaneously); this would enable labs of various sizes to harness the computing power available to them without forcing them to invest in a particular type of batch system. Here we describe modifications made to one particular tool, MAKER. MAKER is a tool for genome annotation that is provided as both a serial application and as an MPI application. We make modifications to enable it to run without MPI and to utilize a wide variety of distributed computing platforms. Further, our proposed parallel framework allows for easy explicit data transfer, which helps overcome a major limitation of bioinformatics tools that generally rely on a shared filesystem. The distributed computing framework we chose to utilize can be used, even during early stages of development, to run bioinformatics tools on clusters, grids, and clouds. We present an evaluation of our modifications using the Caenorhabditis japonica genome comprising 180 megabases of data and achieve a speedup of 45× using 50 workers.

In this paper we discuss challenges of common bioinformatics applications when deployed outside their initial development environments. We propose a three-tiered approach to mitigate some of these issues by leveraging an encapsulation tool, a high-level workflow language, and a portable intermediary. As a case study, we apply this approach to refactor a custom EST analysis pipeline. The Starch tool encapsulates program dependencies to simplify task specification and deployment. The Weaver language provides abstractions for distributed computing and naturally encourages code modularity. The Makeflow workflow engine provides a batch system agnostic engine to execute compiled Weaver code. To illustrate the benefits of our framework, we compare implementations, show their performance, and discuss benefits derived from our new workflow approach relative to traditional bioinformatics development.