Search

Supplementary Table S3. Comprehensive drug sensitivity and resistance testing with 200 cancer drugs. Cell viability in GIST882 and GIST48 cell lines was measured in five different concentrations over a 10,000-fold concentration range.

Supplementary Figure S1. A box-whisker plot showing the relative KIT, ANO1, and PRKCQ (encodes for protein kinase C theta) mRNA expression in soft-tissue sarcoma samples in the MediSapiens database. Supplementary Figure S2. PDE3A (A) and KIT (B) expression in human intestinal smooth muscle tissue (magnification x400, scale bar 100 μm). Supplementary Figure S3 A, Effects of the combinations of PDE3 siRNAs and DMSO on the GIST882 and GIST48 cell lines. siPDE3B and siPDE3A/B reduced GIST882 cell viability, and none of the treatments influenced markedly GIST48 viability. SEM < 0.05 in all data points. B, Effects of PDE3 inhibitors cilostazol, milrinone, and amrinone given with DMSO on GIST882 and GIST48 cell line viability. SE < 0.05 in all data points. Supplementary Figure S4. A Sankey diagram of top SPIA pathways. A Sankey diagram showing differentially expressed miRNAs, their predicted target genes, and the biological pathways estimated to be significantly impacted by the expression change (SPIA pGFDR

Supplemental legend

Supplementary Materials and Methods

Supplementary Table S1. The outlier mRNA expression profiles of the gene tissue index (GTI) outlier analysis. The list of 2415 outlier genes in GIST is ranked by the gene tissue index.

Purpose:Gastrointestinal stromal tumor (GIST) is a common type of soft-tissue sarcoma. Imatinib, an inhibitor of KIT, platelet-derived growth factor receptor alpha (PDGFRA), and a few other tyrosine kinases, is highly effective for GIST, but advanced GISTs frequently progress on imatinib and other approved tyrosine kinase inhibitors. We investigated phosphodiesterase 3 (PDE3) as a potential therapeutic target in GIST cell lines and xenograft models.Experimental Design:The GIST gene expression profile was interrogated in the MediSapiens IST Online transcriptome database comprising human tissue and cancer samples, and PDE3A and PDE3B expression was studied using IHC on tissue microarrays (TMA) consisting of 630 formalin-fixed human tissue samples. GIST cell lines were screened for sensitivity to 217 anticancer compounds, and the efficacy of PDE inhibitors on GIST was further studied in GIST cell lines and patient-derived mouse xenograft models.Results:GISTs expressed PDE3A and PDE3B frequently compared with other human normal or cancerous tissues both in the in silico database and the TMAs. Anagrelide was identified as the most potent of the PDE3 modulators evaluated. It reduced cell viability, promoted cell death, and influenced cell signaling in GIST cell lines. Anagrelide inhibited tumor growth in GIST xenograft mouse models. Anagrelide was also effective in a GIST xenograft mouse model with KIT exon 9 mutation that may pose a therapeutic challenge, as these GISTs require a high daily dose of imatinib.Conclusions:PDE3A and PDE3B are frequently expressed in GIST. Anagrelide had anticancer efficacy in GIST xenograft models and warrants further testing in clinical trials.

Motivation Fusion genes are both useful cancer biomarkers and important drug targets. Finding relevant fusion genes is challenging due to genomic instability resulting in a high number of passenger events. To reveal and prioritize relevant gene fusion events we have developed FUsionN Gene Identification toolset (FUNGI) that uses an ensemble of fusion detection algorithms with prioritization and visualization modules. Results We applied FUNGI to an ovarian cancer dataset of 107 tumor samples from 36 patients. Ten out of 11 detected and prioritized fusion genes were validated. Many of detected fusion genes affect the PI3K-AKT pathway with potential role in treatment resistance. Availabilityand implementation FUNGI and its documentation are available at https://bitbucket.org/alejandra_cervera/fungi as standalone or from Anduril at https://www.anduril.org. Supplementary information Supplementary data are available at Bioinformatics online.

Purpose: Gastrointestinal stromal tumor (GIST) is a common type of soft-tissue sarcoma. Imatinib, an inhibitor of KIT, platelet-derived growth factor receptor alpha (PDGFRA), and a few other tyrosine kinases, is highly effective for GIST, but advanced GISTs frequently progress on imatinib and other approved tyrosine kinase inhibitors. We investigated phosphodiesterase 3 (PDE3) as a potential therapeutic target in GIST cell lines and xenograft models. Experimental Design: The GIST gene expression profile was interrogated in the MediSapiens IST Online transcriptome database comprising human tissue and cancer samples, and PDE3A and PDE3B expression was studied using IHC on tissue microarrays (TMA) consisting of 630 formalin-fixed human tissue samples. GIST cell lines were screened for sensitivity to 217 anticancer compounds, and the efficacy of PDE inhibitors on GIST was further studied in GIST cell lines and patient-derived mouse xenograft models. Results: GISTs expressed PDE3A and PDE3B frequently compared with other human normal or cancerous tissues both in the in silico database and the TMAs. Anagrelide was identified as the most potent of the PDE3 modulators evaluated. It reduced cell viability, promoted cell death, and influenced cell signaling in GIST cell lines. Anagrelide inhibited tumor growth in GIST xenograft mouse models. Anagrelide was also effective in a GIST xenograft mouse model with KIT exon 9 mutation that may pose a therapeutic challenge, as these GISTs require a high daily dose of imatinib. Conclusions: PDE3A and PDE3B are frequently expressed in GIST. Anagrelide had anticancer efficacy in GIST xenograft models and warrants further testing in clinical trials.

Aims and objectives Our objective was to rapidly adapt and scale a registered nurse-driven Coordinated Transitional Care (C-TraC) programme to provide intensive home monitoring and optimise care for outpatient Veterans with COVID-19 in a large urban Unites States healthcare system. Background Our diffuse primary care network had no existing model of care by which to provide coordinated result tracking and monitoring of outpatients with COVID-19. Design Quality improvement implementation project. Methods We used the Replicating Effective Programs model to guide implementation, iterative Plan-Do-Study-Act cycles and SQUIRE reporting guidelines. Two transitional care registered nurses, and a geriatrician medical director developed a protocol that included detailed initial assessment, overnight delivery of monitoring equipment and phone-based follow-up tailored to risk level and symptom severity. We tripled programme capacity in time for the surge of cases by training Primary Care registered nurses. Results Between 23 March and 15 May 2020, 120 Veterans with COVID-19 were enrolled for outpatient monitoring; over one-third were aged 65 years or older, and 70% had medical conditions associated with poor COVID-19 outcomes. All Veterans received an initial call within a few hours of the laboratory reporting positive results. The mean length of follow-up was 8.1 days, with an average of 4.2 nurse and 1.3 physician or advanced practice clinician contacts per patient. The majority (85%) were managed entirely in the outpatient setting. After the surge, the model was disseminated to individual primary care teams through educational sessions. Conclusion A model based on experienced registered nurses can provide comprehensive, effective and sustainable outpatient monitoring to high-risk populations with COVID-19.

Cancer is the most complex genetic disease known, with mutations implicated in more than 250 genes. However, it is still elusive which specific mutations found in human patients lead to tumorigenesis. Here we show that a combination of oncogenes that is characteristic of liver cancer (CTNNB1, TERT, MYC) induces senescence in human fibroblasts and primary hepatocytes. However, reprogramming fibroblasts to a liver progenitor fate, induced hepatocytes (iHeps), makes them sensitive to transformation by the same oncogenes. The transformed iHeps are highly proliferative, tumorigenic in nude mice, and bear gene expression signatures of liver cancer. These results show that tumorigenesis is triggered by a combination of three elements: the set of driver mutations, the cellular lineage, and the state of differentiation of the cells along the lineage. Our results provide direct support for the role of cell identity as a key determinant in transformation and establish a paradigm for studying the dynamic role of oncogenic drivers in human tumorigenesis.

Adult-type granulosa cell tumors (AGCTs) are sex-cord derived neoplasms with a propensity for late relapse. Hormonal modulators have been used empirically in the treatment of recurrent AGCT, albeit with limited success. To provide a more rigorous foundation for hormonal therapy in AGCT, we used a multimodal approach to characterize the expressions of key hormone biomarkers in 175 tumor specimens and 51 serum samples using RNA sequencing, immunohistochemistry, RNA in situ hybridization, quantitative PCR, and circulating biomarker analysis, and correlated these results with clinical data. We show that FSH receptor and estrogen receptor beta (ERβ) are highly expressed in the majority of AGCTs, whereas the expressions of estrogen receptor alpha (ERα) and G-protein coupled estrogen receptor 1 are less prominent. ERβ protein expression is further increased in recurrent tumors. Aromatase expression levels show high variability between tumors. None of the markers examined served as prognostic biomarkers for progression-free or overall survival. In functional experiments, we assessed the effects of FSH, estradiol (E2), and the aromatase inhibitor letrozole on AGCT cell viability using 2 in vitro models: KGN cells and primary cultures of AGCT cells. FSH increased cell viability in a subset of primary AGCT cells, whereas E2 had no effect on cell viability at physiological concentrations. Letrozole suppressed E2 production in AGCTs; however, it did not impact cell viability. We did not find preclinical evidence to support the clinical use of aromatase inhibitors in AGCT treatment, and thus randomized, prospective clinical studies are needed to clarify the role of hormonal treatments in AGCTs.

Despite having a favorable response to platinum-based chemotherapies, ~15% of Testicular Germ-Cell Tumor (TGCT) patients are platinum-resistant. Mortality rates among Latin American countries have remained constant over time, which makes the study of this population of particular interest. To gain insight into this phenomenon, we conducted whole-exome sequencing, microarray-based comparative genomic hybridization, and copy number analysis of 32 tumors from a Mexican cohort, of which 18 were platinum-sensitive and 14 were platinum-resistant. We incorporated analyses of mutational burden, driver mutations, and SNV and CNV signatures. DNA breakpoints in genes were also investigated and might represent an interesting research opportunity. We observed that sensitivity to chemotherapy does not seem to be explained by any of the mutations detected. Instead, we uncovered CNVs, particularly amplifications on segment 2q11.1 as a novel variant with chemosensitivity biomarker potential. Our data shed light into understanding platinum resistance in a Latin-origin population.

Coding variants represent many of the strongest associations between genotype and phenotype; however, they exhibit interindividual differences in effect, termed 'variable penetrance'. Here, we study how cis-regulatory variation modifies the penetrance of coding variants. Using functional genomic and genetic data from the Genotype-Tissue Expression Project (GTEx), we observed that in the general population, purifying selection has depleted haplotype combinations predicted to increase pathogenic coding variant penetrance. Conversely, in cancer and autism patients, we observed an enrichment of penetrance increasing haplotype configurations for pathogenic variants in disease-implicated genes, providing evidence that regulatory haplotype configuration of coding variants affects disease risk. Finally, we experimentally validated this model by editing a Mendelian single-nucleotide polymorphism (SNP) using CRISPR/Cas9 on distinct expression haplotypes with the transcriptome as a phenotypic readout. Our results demonstrate that joint regulatory and coding variant effects are an important part of the genetic architecture of human traits and contribute to modified penetrance of disease-causing variants.

Diffuse large B-cell lymphoma (DLBCL) is a biologically and clinically heterogeneous disease whose personalized clinical management requires robust molecular stratification. Here, we show that somatic hypermutation (SHM) patterns constitute a marker for DLBCL molecular classification. The activity of SHM mutational processes delineated the cell of origin (COO) in DLBCL. Expression of the herein identified 36 SHM target genes stratified DLBCL into four novel SHM subtypes. In a meta-analysis of patients with DLBCL treated with immunochemotherapy, the SHM subtypes were significantly associated with overall survival (1642 patients) and progression-free survival (795 patients). Multivariate analysis of survival indicated that the prognostic impact of the SHM subtypes is independent from the COO classification and the International Prognostic Index. Furthermore, the SHM subtypes had a distinct clinical outcome within each of the COO subtypes, and strikingly, even within unclassified DLBCL. The genetic landscape of the four SHM subtypes indicated unique associations with driver alterations and oncogenic signaling in DLBCL, which suggests a possibility for therapeutic exploitation. These findings provide a biologically driven classification system in DLBCL with potential clinical applications.

Summary Anduril is an analysis and integration framework that facilitates the design, use, parallelization and reproducibility of bioinformatics workflows. Anduril has been upgraded to use Scala for pipeline construction, which simplifies software maintenance, and facilitates design of complex pipelines. Additionally, Anduril’s bioinformatics repository has been expanded with multiple components, and tutorial pipelines, for next-generation sequencing data analysis. Availabilityand implementation Freely available at http://anduril.org. Supplementary information Supplementary data are available at Bioinformatics online.

Cancer is the most complex genetic disease known, with mutations implicated in more than 250 genes. However, it is still elusive which specific mutations found in human patients lead to tumorigenesis. Here we show that a combination of oncogenes that is characteristic of liver cancer (CTNNB1, TERT, MYC) induces senescence in human fibroblasts and primary hepatocytes. However, reprogramming fibroblasts to a liver progenitor fate, induced hepatocytes (iHeps), makes them sensitive to transformation by the same oncogenes. The transformed iHeps are highly proliferative, tumorigenic in nude mice, and bear gene expression signatures of liver cancer. These results show that tumorigenesis is triggered by a combination of three elements: the set of driver mutations, the cellular lineage, and the state of differentiation of the cells along the lineage. Our results provide direct support for the role of cell identity as a key determinant in transformation, and establish a paradigm for studying the dynamic role of oncogenic drivers in human tumorigenesis.

Despite better therapeutic options and improved survival of diffuse large B-cell lymphoma (DLBCL), 30–40% of the patients experience relapse or have primary refractory disease with a dismal prognosis. To identify biological correlates for treatment resistance, we profiled microRNAs (miRNAs) of matched primary and relapsed DLBCL by next-generation sequencing. Altogether 492 miRNAs were expressed in the DLBCL samples. Thirteen miRNAs showed significant differential expression between primary and relapse specimen pairs. Integration of the differentially expressed miRNAs with matched mRNA expression profiles identified highly anti-correlated, putative targets, which were significantly enriched in cancer-associated pathways, including phosphatidylinositol (PI)), mitogen-activated protein kinase (MAPK), and B-cell receptor (BCR) signaling. Expression data suggested activation of these pathways during disease progression, and functional analyses validated that miR-370-3p, miR-381-3p, and miR-409-3p downregulate genes on the PI, MAPK, and BCR signaling pathways, and enhance chemosensitivity of DLBCL cells in vitro. High expression of selected target genes, that is, PIP5K1 and IMPA1, was found to be associated with poor survival in two independent cohorts of chemoimmunotherapy-treated patients (n = 92 and n = 233). Taken together, our results demonstrate that differentially expressed miRNAs contribute to disease progression by regulating key cell survival pathways and by mediating chemosensitivity, thus representing potential novel therapeutic targets.

SummaryCoding variants represent many of the strongest associations between genotype and phenotype, however they exhibit inter-individual differences in effect, known as variable penetrance. In this work, we study how cis-regulatory variation modifies the penetrance of coding variants in their target gene. Using functional genomic and genetic data from GTEx, we observed that in the general population, purifying selection has depleted haplotype combinations that lead to higher penetrance of pathogenic coding variants. Conversely, in cancer and autism patients, we observed an enrichment of haplotype combinations that lead to higher penetrance of pathogenic coding variants in disease implicated genes, which provides direct evidence that regulatory haplotype configuration of causal coding variants affects disease risk. Finally, we experimentally demonstrated that a regulatory variant can modify the penetrance of a coding variant by introducing a Mendelian SNP using CRISPR/Cas9 on distinct expression haplotypes and using the transcriptome as a phenotypic readout. Our results demonstrate that joint effects of regulatory and coding variants are an important part of the genetic architecture of human traits, and contribute to modified penetrance of disease-causing variants.

Diffuse large B-cell lymphoma (DLBCL) is a group of clinically aggressive and heterogeneous malignancies, of which approximately 60% can be cured with anthracycline-based chemoimmunotherapy. Based on gene expression profiling, DLBCL can be classified into two molecularly distinct subgroups showing

Background Large-scale sequencing experiments are complex and require a wide spectrum of computational tools to extract and interpret relevant biological information. This is especially true in projects where individual processing and integrated analysis of both small RNA and complementary RNA data is needed. Such studies would benefit from a computational workflow that is easy to implement and standardizes the processing and analysis of both sequenced data types. Results We developed SePIA (Sequence Processing, Integration, and Analysis), a comprehensive small RNA and RNA workflow. It provides ready execution for over 20 commonly known RNA-seq tools on top of an established workflow engine and provides dynamic pipeline architecture to manage, individually analyze, and integrate both small RNA and RNA data. Implementation with Docker makes SePIA portable and easy to run. We demonstrate the workflow’s extensive utility with two case studies involving three breast cancer datasets. SePIA is straightforward to configure and organizes results into a perusable HTML report. Furthermore, the underlying pipeline engine supports computational resource management for optimal performance. Conclusion SePIA is an open-source workflow introducing standardized processing and analysis of RNA and small RNA data. SePIA’s modular design enables robust customization to a given experiment while maintaining overall workflow structure. It is available at http://anduril.org/sepia. Electronic supplementary material The online version of this article (doi:10.1186/s13040-016-0099-z) contains supplementary material, which is available to authorized users.

Pre-eclampsia is a common and complex pregnancy disorder that often involves impaired placental development. In order to identify altered gene expression in pre-eclamptic placenta, we sequenced placental transcriptomes of nine pre-eclamptic and nine healthy pregnant women in pools of three. The differential gene expression was tested both by including all the pools in the analysis and by excluding some of the pools based on phenotypic characteristics. From these analyses, we identified altogether 53 differently expressed genes, a subset of which was validated by qPCR in 20 cases and 19 controls. Furthermore, we conducted pathway and functional analyses which revealed disturbed vascular function and immunological balance in pre-eclamptic placenta. Some of the genes identified in our study have been reported by numerous microarray studies (BHLHE40, FSTL3, HK2, HTRA4, LEP, PVRL4, SASH1, SIGLEC6), but many have been implicated in only few studies or have not previously been linked to pre-eclampsia (ARMS2, BTNL9, CCSAP, DIO2, FER1L4, HPSE, LOC100129345, LYN, MYO7B, NCMAP, NDRG1, NRIP1, PLIN2, SBSPON, SERPINB9, SH3BP5, TET3, TPBG, ZNF175). Several of the molecules produced by these genes may have a role in the pathogenesis of pre-eclampsia, and some could qualify as biomarkers for prediction or detection of this pregnancy complication.

[EN] RNA-sequencing (RNA-seq) has a wide variety of applications, but no single analysis pipeline can be used in all cases. We review all of the major steps in RNA-seq data analysis, including experimental design, quality control, read alignment, quantification of gene and transcript levels, visualization, differential gene expression, alternative splicing, functional analysis, gene fusion detection and eQTL mapping. We highlight the challenges associated with each step. We discuss the analysis of small RNAs and the integration of RNA-seq with other functional genomics techniques. Finally, we discuss the outlook for novel technologies that are changing the state of the art in transcriptomics., The authors would like to thank Michael Love and Harold Pimentel for helpful suggestions on the initial draft of the manuscript. AC, ST, AM, DGC were supported by the FP7 STATegra project (grant 36000). Research in AC’s laboratory was supported by MINECO grant BIO2012-40244 and co-funded with European Regional Development Funds (ERDF). Research in PM’s laboratory is supported by ERC starting grant Relieve-IMDs and by a core support grant from the Wellcome Trust and MRC to the Wellcome TrustMedical Research Council Cambridge Stem Cell Institute. XZ was supported by the National Basic Research Program of China (2012CB316504). LLE was supported by JDRF (grant number 2-2013-32) and by the Sigrid Juselius Foundation. ACe was supported by the Academy of Finland (Center of Excellence in Cancer Genetics Research).

Pre-eclampsia is a common and complex pregnancy disorder that often involves impaired placental development. In order to identify altered gene expression in pre-eclamptic placenta, we sequenced placental transcriptomes of nine pre-eclamptic and nine healthy pregnant women in pools of three. The differential gene expression was tested both by including all the pools in the analysis and by excluding some of the pools based on phenotypic characteristics. From these analyses, we identified altogether 53 differently expressed genes, a subset of which was validated by qPCR in 20 cases and 19 controls. Furthermore, we conducted pathway and functional analyses which revealed disturbed vascular function and immunological balance in pre-eclamptic placenta. Some of the genes identified in our study have been reported by numerous microarray studies (BHLHE40, FSTL3, HK2, HTRA4, LEP, PVRL4, SASH1, SIGLEC6), but many have been implicated in only few studies or have not previously been linked to pre-eclampsia (ARMS2, BTNL9, CCSAP, DIO2, FER1L4, HPSE, LOC100129345, LYN, MYO7B, NCMAP, NDRG1, NRIP1, PLIN2, SBSPON, SERPINB9, SH3BP5, TET3, TPBG, ZNF175). Several of the molecules produced by these genes may have a role in the pathogenesis of pre-eclampsia and some could qualify as biomarkers for prediction or detection of this pregnancy complication.

Effect of alternative splicing (AS) on diffuse large B-cell lymphoma (DLBCL) pathogenesis and survival has not been systematically addressed. Here, we compared differentially expressed genes and exons in association with survival after chemoimmunotherapy, and between germinal center B-cell like (GCB) and activated B-cell like (ABC) DLBCLs. Genome-wide exon array-based screen was performed from samples of 38 clinically high-risk patients who were treated in a Nordic phase II study with dose-dense chemoimmunotherapy and central nervous system prophylaxis. The exon expression profile separated the patients according to molecular subgroups and survival better than the gene expression profile. Pathway analyses revealed enrichment of AS genes in inflammation and adhesion-related processes, and in signal transduction, such as phosphatidylinositol signaling system and adenosine triphosphate binding cassette transporters. Altogether, 49% of AS-related exons were protein coding, and domain prediction showed 28% of such exons to include a functional domain, such as transmembrane helix domain or phosphorylation sites. Validation in an independent cohort of 92 DLBCL samples subjected to RNA-sequencing confirmed differential exon usage of selected genes and association of AS with molecular subtypes and survival. The results indicate that AS events are able to discriminate GCB and ABC DLBCLs and have prognostic impact in DLBCL.

Background Despite better therapeutic options and improved survival of diffuse large B-cell lymphoma (DLBCL), 30-40% of the patients still experience relapse with a dismal prognosis. To identify biological correlates for poor survival, we compared differentially expressed genes between high-risk DLBCL patients, who had relapsed or remained in remission after dose-dense chemoimmunotherapy. Design and methods We performed genome-wide exon array analysis (Affymetrix Human Exon 1.0 ST) with 38 de novo high-risk (aaIPI>1) DLBCL patients less than 65 years old. The patients were treated in a Nordic phase II protocol with six courses of R-CHOEP-14 followed by systemic central nervous system prophylaxis with one course of high dose methotrexate and high dose cytarabine. We quantified the expression data by MEAP (Multiple Exon Array Preprocessing) algorithm and identified differentially expressed genes between the patients, who relapsed (n=9; poor prognosis group) or remained in long-term remission (n=29; good prognosis group). RNA sequencing and oligonucleotide-based microarray data from 92 (Cancer Genome Characterization Initiative, CGCI) and 233 (Lymphoma/Leukemia Molecular Profiling Project, LLMPP) DLBCL patients were utilized for validation. In both validation cohorts, the patients were treated with chemoimmunotherapy. Results Gene expression profiling (GEP) identified 215 differentially expressed genes (DEGs) between the good and poor prognosis groups. Of these, 41% were upregulated and 59% suppressed in the patients with poor prognosis. Pathway analyses indicated that the DEGs were enriched in many immune response related processes, such as the antigen processing and presenting pathway. Of note was the observation that major histocompatibility complex class II transactivator (CIITA) and selected human leucocyte antigen- (HLA) genes were significantly suppressed in the poor prognosis group. A correlation between CIITA and HLA gene expression levels was also observed. In Cox univariate analysis with continuous variables, low CIITA expression predicted poor overall survival (OS; p = 0.035). Five-year OS rates were 89% and 50% for the patients with high and low CIITA mRNA levels, respectively (p = 0.007). The prognostic impact of CIITA expression on survival was confirmed in independent CGCI (OS; p = 0.020, PFS; p = 0.002) and LLMPP validation cohorts (OS; p = 0.002, PFS data not available). In the CGCI cohort, the association of CIITA expression with survival was independent of IPI. Consistent with previous data, suppression of certain HLA-genes was also associated with poor outcome. Conclusions The results demonstrate that low CIITA expression is a novel independent predictor of poor survival in patients with DLBCL. The loss of CIITA offers, through the loss of MHC class II expression, a mechanism of immune escape and treatment resistance. Disclosures Leppä: Amgen: Research Funding; CTI Life Sciences: Honoraria; Janssen: Research Funding; Takeda: Honoraria, Other: Travel expenses; Roche: Honoraria, Other: Travel Expenses, Research Funding; Mundipharma: Research Funding; Bayer: Honoraria, Research Funding.

Learning how to count in different bases has been seen as a trivial task in almost all introductory mathematics courses. However, the low performance shown by many students, while performing this task, is appalling. This situation has motivated serious research in this matter. In order to study a model of count learning, we analyze the performance of a multilayer perceptron that learns to count in several bases (5, 10, 13, 20, 60). We give evidence that it is not equivalent for the model to learn to count in all base as the errors are not equivalent, biased toward a low error when the task is to learn to count in base 20. When the task is to learn to count in all bases following a given sequence, the model shows non-equivalent errors for some bases. This may shed some light on education planning that can result in better introductory courses.

Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous disease both biologically and clinically. Although a number of genetic alterations have been associated with the pathogenesis of DLBCL, survival association of most of the identified mutations remains to be shown. Here we have used exome and RNA sequencing (RNAseq) data to identify novel genetic events associated with treatment outcome after chemoimmunotherapy. Initial screen was performed with eight primary tumor samples and matched normal DNA using exome sequencing. The patients had clinically high risk DLBCL, and were treated in a Nordic phase II protocol with dose dense chemoimmunotherapy and CNS prophylaxis. After stringent filtering (VarScan2 p < 0.05), and removal of known germline variants, 967 somatic point mutations and indels in 561 different genes were identified. Of these, 436 were non-synonymous, and 27 were detected in more than one patient. These included both novel and several previously described DLBCL associated mutations, such as MYD88, CD79B, B2M and PTEN. To correlate the identified mutations with survival, we used RNAseq and survival data of 92 DLBCL samples from the Cancer Genome Characterization Initiative (CGCI) repository. The mutational status of the 436 genes identified in our screen was assessed with Bambino, which confirmed 405 of these genes also to be mutated in the CGCI set. Kaplan-Meier survival estimates per gene with at least one single nucleotide variant (SNV; with criteria of minimum 3 reads and 10% coverage and known polymorphisms filtered out) revealed 17 genes that were associated with overall survival (OS, p We highlight mutations affecting Deltex1 (DTX1), which functions as an E3 ubiquitin ligase and targets Notch1 intracellular domain for degradation. In the study cohort, 14% of the patients had a non-synonymous DTX1 mutation. Mutations were equally distributed between GCB and ABC DLBCLs and low (0-2) and high (3-5) International Prognostic Index (IPI) risk groups. According to Kaplan-Meier analysis, the 5-year PFS rate for the patients with DTX1 mutations was significantly worse as compared to ones without mutations (46% vs 78%, p=0.010). The corresponding OS values were 54% and 85% (p=0.006). In multivariate analysis with IPI, DTX1 mutation status remained as an independent prognostic factor for both PFS (RR, 2.83; 95% CI, 1.17-6.85, p=0.021) and OS (RR, 3.51; 95% CI, 1.41-8.71, p=0.007). Mutational status was also found to correlate with the expression levels; the patients with mutations had lower DTX1 mRNA levels (p=0.006). When DTX1 mutations were analyzed according to different functional domains, 11 out of 16 (65%) were located in the first exon encoding the WWE 1 domain and a binding site for intracellular domain of Notch1. Interestingly, the patients with mutations in exon 1 had a significantly worse survival in comparison to other patients (Figure 1). Taken together, the results demonstrate that DTX1 mutations in DLBCL are an independent predictor of survival after chemoimmunotherapy, and mutations enriched in exon 1 identify cases with dismal prognosis. Figure 1 Figure 1. Disclosures No relevant conflicts of interest to declare.

Abstract 1557 Background: The aim was to compare differentially expressed exons and splicing variants between diffuse large B-cell lymphoma (DLBCL) patients, who had relapsed or remained in remission after dose dense chemoimmunotherapy. Design and methods: We performed genome-wide exon array analysis from four DLBCL cell lines and 38 tumor tissues from young (1) disease. The patients were treated in a Nordic phase II protocol with six courses of R-CHOEP-14 followed by systemic central nervous system prophylaxis with one course of high dose methotrexate and high dose cytarabine. At the time of the analysis, median follow up was 34 months, predicted 3-year progression free survival (PFS) 78% and overall survival (OS) 79%. RNA for quantitative PCR validation was available from 20 patients. Two DLBCL cell lines and eight patient samples were further analyzed with high throughput RNA sequencing. In addition, microarray data set from 233 DLBCL patients treated with chemoimmunotherapy (Lymphoma/Leukemia Molecular Profiling Project (LLMPP)) was utilized for validation. Results: Differentially expressed exons between relapsed and non-relapsed patients were screened using criteria of p ≤ 0.05 and fold change ≥1.6 converting to 566 differentially expressed genes, of which 131 coded proteins. One of the identified genes with possible alternative splicing was TUBB2B, which encodes therapeutic target of taxanes and vinca alkaloids. The expression of TUBB2B, and specifically the expression of exon 3, was found to be suppressed in relapsed patients in comparison to patients remaining in remission. Differential expression of TUBB2B whole transcript and exon 3 was confirmed with RNAseq and quantitative PCR. Studies in lymphoma cell lines provided further support for the existence of different TUBB2B isoforms. According to Kaplan Meier estimates the patients with high (>median) expression levels of TUBB2B exon 3 had better 3-year PFS and lymphoma related OS rates than the patients with low expression levels (95% vs. 61%, p=0.015 for PFS, 100% vs. 75%, p=0.024 for OS). The prognostic significance of TUBB2B gene expression was validated in LLMPP data set (3-year OS 80% vs. 67%, p=0.040). Conclusions: The results provide evidence that differential expression and splicing of TUBB2B gene can discriminate the outcome of homogenously treated high risk DLBCL patients. Disclosures: No relevant conflicts of interest to declare.

Introduction Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous disease both clinically and biologically. Based on gene expression profiling, two major molecular subtypes, activated B-cell (ABC)- and germinal center B-cell (GCB)-like DLBCLs, have been recognized. More recently, next generation sequencing has also shed light on the mutational heterogeneity of these lymphomas. While several oncogenes are recurrently mutated and aberrantly expressed in DLBCL, little is known of their functional impact and association with survival. Design and Methods To identify mutated oncogenes in DLBCL, we performed whole-exome and RNA sequencing screen for tumor tissue and matched normal DNA from eight high-risk (aaIPI >2) DLBCL patients less than 65 years old. The patients were treated in the Nordic phase II protocol with dose-dense chemoimmunotherapy followed by systemic central nervous system (CNS) prophylaxis. Initial screen was expanded to a Nordic discovery cohort consisting of 38 high-risk DLBCL patients treated in the same protocol. Gene expression was analyzed from the tumor tissue using Affymetrix Human Exon 1.0 ST arrays, quantified with MEAP (Multiple Exon Array Preprocessing, probe annotation version 70) algorithm, and expression of the mutated genes identified. Samples were classified into GCB, ABC and nonclassified subgroups using the Lymphochip gene predictor. At the time of the analysis, median follow-up time was 65 months and predicted 5-year OS 76%. Survival association of the identified mutations and gene expression was validated using RNAseq data from 92 (Cancer Genome Characterization Initiative (CGCI) repository), and oligonucleotide-based microarray expression data from 233 (Lymphoma/Leukemia Molecular Profiling Project, LLMPP) DLBCL patients, respectively. Patients were treated with chemoimmunotherapy in both validation cohorts. Results In our initial screen 922 genes were affected by somatic mutations. The genes included both novel and previously identified lymphoma-related oncogenes, such as MYD88, CD79B and DTX1. We highlight mutations affecting GC marker Kelch-like family member 6 (KLHL6), which is a predicted oncogene and involved in B-cell receptor (BCR) signaling in DLBCL. In the CGCI validation cohort 16% of the patients carried non-silent KLHL6 mutations. According to Kaplan Meier analysis, the patients with KLHL6 mutations had significantly worse survival in comparison to the patients without mutations (5-year OS; 53% vs 84%; p = .010). In multivariate analysis, the prognostic impact of KLHL6 mutations on the survival was independent of IPI (OS; RR, 3.060; 95% CI, 1.233-7.598; p = .016). KLHL6 mutations were equally distributed between the molecular subtypes. However, when the molecular subtypes were analyzed separately, KLHL6 mutations remained predictive for poor survival only in the ABC-like DLBCL (OS; RR, 6.727; 95% CI 1.576-28.714; p= .010). KLHL6 expression was further analyzed in the Nordic discovery cohort. Consistent with its role as a GC marker, KLHL6 expression tended to be higher in the GCB- than the ABC-like DLBCLs. Overall, low KLHL6 expression was associated with poor survival (5-year OS; 93% vs 65%; p = .044). The difference in survival was restricted to the ABC subtype. The prognostic impact of KLHL6 expression in the patients with the ABC-like DLBCL was confirmed in independent LLMPP validation cohort. Conclusions The results show that although high KLHL6 expression is characteristic for the GCB DLBCL, mutations and low expression of KLHL6 are clinically more relevant in the ABC subtype pinpointing the patients with the most dismal prognosis. Disclosures Leppä: CTI Life Sciences: Honoraria; Roche: Honoraria, Other: Travel Expenses, Research Funding; Takeda: Honoraria, Other: Travel expenses; Bayer: Honoraria, Research Funding; Mundipharma: Research Funding; Janssen: Research Funding; Amgen: Research Funding.

Background Alternative splicing (AS) increases dramatically the diversity of the proteome by allowing a single gene to generate multiple transcripts and functionally diverse protein isoforms. Although previous studies have provided evidence that AS is often deregulated in cancers, the role of AS events in diffuse large B-cell lymphoma (DLBCL) remains largely unexplored. Here, we design and apply a novel method to analyze the global prognostically significant mRNA and AS variation by comparing differentially expressed exons and splicing variants between diffuse large B-cell lymphoma (DLBCL) patients, who have relapsed (poor prognosis group) or remained in remission (good prognosis group) after chemoimmunotherapy. Design and methods We performed genome-wide exon array analysis on RNAs isolated from 38 tumor tissues from young (1) disease. The patients were treated in a Nordic phase II protocol with six courses of R-CHOEP-14 followed by systemic central nervous system prophylaxis with one course of high dose methotrexate and high dose cytarabine. At the time of the analysis, median follow up was 50 months, predicted 5-year progression free survival (PFS) 74% and overall survival (OS) 75%. To identify genes with differential expression or with differential splicing between patients with poor or good prognosis, a novel algorithm based on a splicing index was developed. Results With a criteria of p Conclusions AS index is a more potential discriminator than conventional gene profiling between good and bad prognosis patients in DLBCL, and may provide a new class of cancer biomarker candidates. Disclosures: No relevant conflicts of interest to declare.