Anduril 2: upgraded large-scale data integration framework.

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.

AML associated oncofusion proteins PML-RARA, AML1-ETO and CBFB-MYH11 target RUNX/ETS-factor binding sites to modulate H3ac levels and drive leukemogenesis.

Chromosomal translocations are one of the hallmarks of acute myeloid leukemia (AML), often leading to gene fusions and expression of an oncofusion protein. Over recent years it has become clear that most of the AML associated oncofusion proteins molecularly adopt distinct mechanisms for inducing leukemogenesis. Still these unique molecular properties of the chimeric proteins converge and give rise to a common pathogenic molecular mechanism. In the present study we compared genome-wide DNA binding and transcriptome data associated with AML1-ETO, CBFB-MYH11 and PML-RARA oncofusion protein expression to identify unique and common features. Our analyses revealed targeting of oncofusion binding sites to RUNX1 and ETS-factor occupied genomic regions. In addition, it revealed a highly comparable global histone acetylation pattern, similar expression of common target genes and related enrichment of several biological pathways critical for maintenance of AML, suggesting oncofusion proteins deregulate common gene programs despite their distinct binding signatures and mechanisms of action.

DNA methylation signature (SAM40) identifies subgroups of the Luminal A breast cancer samples with distinct survival.

Breast cancer patients with Luminal A disease generally have a good prognosis, but among this patient group are patients with good prognosis that are currently overtreated with adjuvant chemotherapy, and also patients that have a bad prognosis and should be given more aggressive treatment. There is no available method for subclassification of this patient group. Here we present a DNA methylation signature (SAM40) that segregates Luminal A patients based on prognosis, and identify one good prognosis group and one bad prognosis group. The prognostic impact of SAM40 was validated in four independent patient cohorts. Being able to subdivide the Luminal A patients may give the two-sided benefit of identifying one subgroup that may benefit from a more aggressive treatment than what is given today, and importantly, identifying a subgroup that may benefit from less treatment.

Data integration to prioritize drugs using genomics and curated data.

BACKGROUND: Genomic alterations affecting drug target proteins occur in several tumor types and are prime candidates for patient-specific tailored treatments. Increasingly, patients likely to benefit from targeted cancer therapy are selected based on molecular alterations. The selection of a precision therapy benefiting most patients is challenging but can be enhanced with integration of multiple types of molecular data. Data integration approaches for drug prioritization have successfully integrated diverse molecular data but do not take full advantage of existing data and literature. RESULTS: We have built a knowledge-base which connects data from public databases with molecular results from over 2200 tumors, signaling pathways and drug-target databases. Moreover, we have developed a data mining algorithm to effectively utilize this heterogeneous knowledge-base. Our algorithm is designed to facilitate retargeting of existing drugs by stratifying samples and prioritizing drug targets. We analyzed 797 primary tumors from The Cancer Genome Atlas breast and ovarian cancer cohorts using our framework. FGFR, CDK and HER2 inhibitors were prioritized in breast and ovarian data sets. Estrogen receptor positive breast tumors showed potential sensitivity to targeted inhibitors of FGFR due to activation of FGFR3. CONCLUSIONS: Our results suggest that computational sample stratification selects potentially sensitive samples for targeted therapies and can aid in precision medicine drug repositioning. Source code is available from http://csblcanges.fimm.fi/GOPredict/.

Identification of sample-specific regulations using integrative network level analysis.

BACKGROUND: Histologically similar tumors even from the same anatomical position may still show high variability at molecular level hindering analysis of genome-wide data. Leveling the analysis to a gene regulatory network instead of focusing on single genes has been suggested to overcome the heterogeneity issue although the majority of the network methods require large datasets. Network methods that are able to function at a single sample level are needed to overcome the heterogeneity and sample size issues. METHODS: We present a novel network method, Differentially Expressed Regulation Analysis (DERA) that integrates expression data to biological network information at a single sample level. The sample-specific networks are subsequently used to discover samples with similar molecular functions by identification of regulations that are shared between samples or are specific for a subgroup. RESULTS: We applied DERA to identify key regulations in triple negative breast cancer (TNBC), which is characterized by lack of estrogen receptor, progesterone receptor and HER2 expression and has poorer prognosis than the other breast cancer subtypes. DERA identified 110 core regulations consisting of 28 disconnected subnetworks for TNBC. These subnetworks are related to oncogenic activity, proliferation, cancer survival, invasiveness and metastasis. Our analysis further revealed 31 regulations specific for TNBC as compared to the other breast cancer subtypes and thus form a basis for understanding TNBC. We also applied DERA to high-grade serous ovarian cancer (HGS-OvCa) data and identified several common regulations between HGS-OvCa and TNBC. The performance of DERA was compared to two pathway analysis methods GSEA and SPIA and our results shows better reproducibility and higher sensitivity in a small sample set. CONCLUSIONS: We present a novel method called DERA to identify subnetworks that are similarly active for a group of samples. DERA was applied to breast cancer and ovarian cancer data showing our method is able to identify reliable and potentially important regulations with high reproducibility. R package is available at http://csbi.ltdk.helsinki.fi/pub/czliu/DERA/.

Identification of genetic markers with synergistic survival effect in cancer.

BACKGROUND: Cancers are complex diseases arising from accumulated genetic mutations that disrupt intracellular signaling networks. While several predisposing genetic mutations have been found, these individual mutations account only for a small fraction of cancer incidence and mortality. With large-scale measurement technologies, such as single nucleotide polymorphism (SNP) microarrays, it is now possible to identify combinatorial effects that have significant impact on cancer patient survival. RESULTS: The identification of synergetic functioning SNPs on genome-scale is a computationally daunting task and requires advanced algorithms. We introduce a novel algorithm, Geninter, to identify SNPs that have synergetic effect on survival of cancer patients. Using a large breast cancer cohort we generate a simulator that allows assessing reliability and accuracy of Geninter and logrank test, which is a standard statistical method to integrate genetic and survival data. CONCLUSIONS: Our results show that Geninter outperforms the logrank test and is able to identify SNP-pairs with synergetic impact on survival.

Germline variation in TP53 regulatory network genes associates with breast cancer survival and treatment outcome.

Germline variation in the TP53 network genes PRKAG2, PPP2R2B, CCNG1, PIAS1 and YWHAQ was previously suggested to have an impact on drug response in vitro. Here, we investigated the effect on breast cancer survival of germline variation in these genes in 925 Finnish breast cancer patients and further analyzed five single nucleotide polymorphisms (SNPs) in PRKAG2 (rs1029946, rs4726050, rs6464153, rs7789699) and PPP2R2B (rs10477313) for 10-year survival in breast cancer patients, interaction with TP53 R72P and MDM2-SNP309, outcome after specific adjuvant therapy and correlation to tumor characteristics in 4,701 invasive cases from four data sets. We found evidence for carriers of PRKAG2-rs1029946 and PRKAG2-rs4726050 having improved survival in the pooled data (HR 0.53, 95% CI 0.3-0.9; p = 0.023 for homozygous carriers of the rare G-allele and HR 0.85, 95% CI 0.7-0.9; p = 0.049 for carriers of the rare G allele, respectively). PRKAG2-rs4726050 showed a significant interaction with MDM2-SNP309, with PRKAG2-rs4726050 rare G-allele having a dose-dependent effect for better breast cancer survival confined only to MDM2 SNP309 rare G-allele carriers (HR 0.45, 95% CI 0.2-0.7; p = 0.001). This interaction also emerged as an independent predictor of better survival (p = 0.047). PPP2R2B-rs10477313 rare A-allele was found to predict better survival (HR 0.82, 95% CI 0.6-0.9; p = 0.018), especially after hormonal therapy (HR 0.66, 95% CI 0.5-0.9; p = 0.048). These findings warrant further studies and suggest that genetic markers in TP53 network genes such as PRKAG2 and PPP2R2B might affect prognosis and treatment outcome in breast cancer patients.

FoxA1 specifies unique androgen and glucocorticoid receptor binding events in prostate cancer cells.

The forkhead protein FoxA1 has functions other than a pioneer factor, in that its depletion brings about a significant redistribution in the androgen receptor (AR) and glucocorticoid receptor (GR) cistromes. In this study, we found a novel function for FoxA1 in defining the cell-type specificity of AR- and GR-binding events in a distinct fashion, namely, for AR in LNCaP-1F5 cells and for GR in VCaP cells. We also found different, cell-type and receptor-specific compilations of cis-elements enriched adjacent to the AR- and GR-binding sites. The AR pathway is central in prostate cancer biology, but the role of GR is poorly known. We find that AR and GR cistromes and transcription programs exhibit significant overlap, and GR regulates a large number of genes considered to be AR pathway-specific. This raises questions about the role of GR in maintaining the AR pathway under androgen-deprived conditions in castration-resistant prostate cancer patients. However, in the presence of androgen, ligand-occupied GR acts as a partial antiandrogen and attenuates the AR-dependent transcription program. .

An optimized isolation of biotinylated cell surface proteins reveals novel players in cancer metastasis.

Details of metastasis, the deadliest aspect of cancer, are unclear. Cell surface proteins play central roles in adhesive contacts between the tumor cell and the stroma during metastasis. We optimized a fast, small-scale isolation of biotinylated cell surface proteins to reveal novel metastasis-associated players from an isogenic pair of human MDA-MB-435 cancer cells with opposite metastatic phenotypes. Isolated proteins were trypsin digested and analyzed using LC-MS/MS followed by quantitation with the Progenesis LC-MS software. Sixteen proteins displayed over twofold expression differences between the metastatic and non-metastatic cells. Interestingly, overexpression of most of them (14/16) in the metastatic cells indicates a gain of novel surface protein profile as compared to the non-metastatic ones. All five validated, differentially expressed proteins showed higher expression in the metastatic cells in culture, and four of these were further validated in vivo. Moreover, we analyzed expression of two of the identified proteins, CD109 and ITGA6 in 3-dimensional cultures of six melanoma cell lines. Both proteins marked the surface of cells derived from melanoma metastasis over cells derived from primary melanoma. The unbiased identification and validation of both known and novel metastasis-associated proteins indicate a reliable approach for the identification of differentially expressed surface proteins.

Dual role of FoxA1 in androgen receptor binding to chromatin, androgen signalling and prostate cancer.

High androgen receptor (AR) level in primary tumour predicts increased prostate cancer-specific mortality. However, the mechanisms that regulate AR function in prostate cancer are poorly known. We report here a new paradigm for the forkhead protein FoxA1 action in androgen signalling. Besides pioneering the AR pathway, FoxA1 depletion elicited extensive redistribution of AR-binding sites (ARBs) on LNCaP-1F5 cell chromatin that was commensurate with changes in androgen-dependent gene expression signature. We identified three distinct classes of ARBs and androgen-responsive genes: (i) independent of FoxA1, (ii) pioneered by FoxA1 and (iii) masked by FoxA1 and functional upon FoxA1 depletion. FoxA1 depletion also reprogrammed AR binding in VCaP cells, and glucocorticoid receptor binding and glucocorticoid-dependent signalling in LNCaP-1F5 cells. Importantly, FoxA1 protein level in primary prostate tumour had significant association to disease outcome; high FoxA1 level was associated with poor prognosis, whereas low FoxA1 level, even in the presence of high AR expression, predicted good prognosis. The role of FoxA1 in androgen signalling and prostate cancer is distinctly different from that in oestrogen signalling and breast cancer.

Role of RNA binding protein HuR in ductal carcinoma in situ of the breast.

HuR is a ubiquitously expressed RNA-binding protein that modulates gene expression at the post-transcriptional level. It is predominantly nuclear, but can shuttle between the nucleus and the cytoplasm. While in the cytoplasm HuR can stabilize its target transcripts, many of which encode proteins involved in carcinogenesis. While cytoplasmic HuR expression is a marker of reduced survival in breast cancer, its role in precursor lesions of malignant diseases is unclear. To address this we explored HuR expression in atypical ductal hyperplasia (ADH) and in ductal in situ carcinomas (DCIS). We show that cytoplasmic HuR expression is elevated in both ADH and DCIS when compared to normal controls, and that this expression associated with high grade, progesterone receptor negativity and microinvasion and/or tumour-positive sentinel nodes of the DCIS. To study the mechanisms of HuR in breast carcinogenesis, HuR expression was silenced in an immortalized breast epithelial cell line (184B5Me), which led to reduction in anchorage-independent growth, increased programmed cell death and inhibition of invasion. In addition, we identified two novel target transcripts (CTGF and RAB31) that are regulated by HuR and that bind HuR protein in this cell line. Our results show that HuR is aberrantly expressed at early stages of breast carcinogenesis and that its inhibition can lead to suppression of this process. ArrayExpress Accession No. E-MEXP-3035.

High frequency of TTK mutations in microsatellite-unstable colorectal cancer and evaluation of their effect on spindle assembly checkpoint.

Frameshift mutations frequently accumulate in microsatellite-unstable colorectal cancers (MSI CRCs) typically leading to downregulation of the target genes due to nonsense-mediated messenger RNA decay. However, frameshift mutations that occur in the 3' end of the coding regions can escape decay, which has largely been ignored in previous works. In this study, we characterized nonsense-mediated decay-escaping frameshift mutations in MSI CRC in an unbiased, genome wide manner. Combining bioinformatic search with expression profiling, we identified genes that were predicted to escape decay after a deletion in a microsatellite repeat. These repeats, located in 258 genes, were initially sequenced in 30 MSI CRC samples. The mitotic checkpoint kinase TTK was found to harbor decay-escaping heterozygous mutations in exon 22 in 59% (105/179) of MSI CRCs, which is notably more than previously reported. Additional novel deletions were found in exon 5, raising the mutation frequency to 66%. The exon 22 of TTK contains an A(9)-G(4)-A(7) locus, in which the most common mutation was a mononucleotide deletion in the A(9) (c.2560delA). When compared with identical non-coding repeats, TTK was found to be mutated significantly more often than expected without selective advantage. Since TTK inhibition is known to induce override of the mitotic spindle assembly checkpoint (SAC), we challenged mutated cancer cells with the microtubule-stabilizing drug paclitaxel. No evidence of checkpoint weakening was observed. As a conclusion, heterozygous TTK mutations occur at a high frequency in MSI CRCs. Unexpectedly, the plausible selective advantage in tumourigenesis does not appear to be related to SAC.

Large-scale data integration framework provides a comprehensive view on glioblastoma multiforme.

BACKGROUND: Coordinated efforts to collect large-scale data sets provide a basis for systems level understanding of complex diseases. In order to translate these fragmented and heterogeneous data sets into knowledge and medical benefits, advanced computational methods for data analysis, integration and visualization are needed. METHODS: We introduce a novel data integration framework, Anduril, for translating fragmented large-scale data into testable predictions. The Anduril framework allows rapid integration of heterogeneous data with state-of-the-art computational methods and existing knowledge in bio-databases. Anduril automatically generates thorough summary reports and a website that shows the most relevant features of each gene at a glance, allows sorting of data based on different parameters, and provides direct links to more detailed data on genes, transcripts or genomic regions. Anduril is open-source; all methods and documentation are freely available. RESULTS: We have integrated multidimensional molecular and clinical data from 338 subjects having glioblastoma multiforme, one of the deadliest and most poorly understood cancers, using Anduril. The central objective of our approach is to identify genetic loci and genes that have significant survival effect. Our results suggest several novel genetic alterations linked to glioblastoma multiforme progression and, more specifically, reveal Moesin as a novel glioblastoma multiforme-associated gene that has a strong survival effect and whose depletion in vitro significantly inhibited cell proliferation. All analysis results are available as a comprehensive website. CONCLUSIONS: Our results demonstrate that integrated analysis and visualization of multidimensional and heterogeneous data by Anduril enables drawing conclusions on functional consequences of large-scale molecular data. Many of the identified genetic loci and genes having significant survival effect have not been reported earlier in the context of glioblastoma multiforme. Thus, in addition to generally applicable novel methodology, our results provide several glioblastoma multiforme candidate genes for further studies.Anduril is available at http://csbi.ltdk.helsinki.fi/anduril/The glioblastoma multiforme analysis results are available at http://csbi.ltdk.helsinki.fi/anduril/tcga-gbm/

Computational identification of cancer susceptibility loci.

The identification of novel cancer susceptibility syndromes and genes from very limited numbers of study individuals has become feasible through the use of high-throughput genotype microarrays. With such an approach, highly sensitive genome-wide computational methods are needed to identify the regions of interest. We have developed novel methods to identify and compare homozygous and compound heterozygous regions between cases and controls, to facilitate the identification of recessively inherited cancer susceptibility loci. As our approach is optimized for sensitivity, it creates many hits that may be unrelated to the phenotype of interest. We compensate for this compromised specificity by the automated use of additional sources of biological information along with a ranking function to focus on the most relevant regions. The methods are demonstrated here by comparing colorectal cancer patients to controls.

Integrative platform to translate gene sets to networks.

SUMMARY: We have implemented a computational platform (Moksiskaan) that integrates pathway, protein-protein interaction, genome and literature mining data to result in comprehensive networks for a list of genes or proteins. Moksiskaan is able to generate hypothetical pathways for these genes or proteins as well as estimate their activation statuses using regulation information in pathway repositories. An automatically generated result document provides a detailed description of the query genes, biological processes and drug targets. Moksiskaan networks can be downloaded to Cytoscape for further analysis. To demonstrate the utility of Moksiskaan, we use gene microarray and clinical data from >200 glioblastoma multiforme primary tumor samples and translate the resulting set of 124 survival-associated genes to a network. AVAILABILITY AND IMPLEMENTATION: Moksiskaan and user guide are freely available under GNU General Public License at http://csbi.ltdk.helsinki.fi/moksiskaan/

Fast gene ontology based clustering for microarray experiments.

BACKGROUND: Analysis of a microarray experiment often results in a list of hundreds of disease-associated genes. In order to suggest common biological processes and functions for these genes, Gene Ontology annotations with statistical testing are widely used. However, these analyses can produce a very large number of significantly altered biological processes. Thus, it is often challenging to interpret GO results and identify novel testable biological hypotheses. RESULTS: We present fast software for advanced gene annotation using semantic similarity for Gene Ontology terms combined with clustering and heat map visualisation. The methodology allows rapid identification of genes sharing the same Gene Ontology cluster. CONCLUSION: Our R based semantic similarity open-source package has a speed advantage of over 2000-fold compared to existing implementations. From the resulting hierarchical clustering dendrogram genes sharing a GO term can be identified, and their differences in the gene expression patterns can be seen from the heat map. These methods facilitate advanced annotation of genes resulting from data analysis.

Computational identification of candidate loci for recessively inherited mutation using high-throughput SNP arrays.

MOTIVATION: Single nucleic polymorphisms (SNPs) are one of the most abundant genetic variations in the human genome. Recently, several platforms for high-throughput SNP analysis have become available, capable of measuring thousands of SNPs across the genome. Tools for analysing and visualizing these large genetic data sets in biologically relevant manner are rare. This hinders effective use of the SNP-array data in research on complex diseases, such as cancer. RESULTS: We describe a computational framework to analyse and visualize SNP-array data, and link the results in relevant databases. Our major objective is to develop methods for identifying DNA regions that likely harbour recessive mutations. Thus, the algorithms are designed to have high sensitivity and the identified regions are ranked using a scoring algorithm. We have also developed annotation tools that automatically query gene IDs, exon counts, microarray probe IDs, etc. In our case study, we apply the methods for identifying candidate regions for recessively inherited colorectal cancer predisposition and suggest directions for wet-lab experiments. AVAILABILITY: R-package implementation is available at http://www.ltdk.helsinki.fi/sysbio/csb/downloads/CohortComparator/