Machine learning approaches for the discovery of gene-gene interactions in disease data.

Because of the complexity of gene-phenotype relationships machine learning approaches have considerable appeal as a strategy for modelling interactions. A number of such methods have been developed and applied in recent years with some modest success. Progress is hampered by the challenges presented by the complexity of the disease genetic data, including phenotypic and genetic heterogeneity, polygenic forms of inheritance and variable penetrance, combined with the analytical and computational issues arising from the enormous number of potential interactions. We review here recent and current approaches focusing, wherever possible, on applications to real data (particularly in the context of genome-wide association studies) and looking ahead to the further challenges posed by next generation sequencing data.

Next generation exome sequencing of paediatric inflammatory bowel disease patients identifies rare and novel variants in candidate genes.

BACKGROUND: Multiple genes have been implicated by association studies in altering inflammatory bowel disease (IBD) predisposition. Paediatric patients often manifest more extensive disease and a particularly severe disease course. It is likely that genetic predisposition plays a more substantial role in this group. OBJECTIVE: To identify the spectrum of rare and novel variation in known IBD susceptibility genes using exome sequencing analysis in eight individual cases of childhood onset severe disease. DESIGN: DNA samples from the eight patients underwent targeted exome capture and sequencing. Data were processed through an analytical pipeline to align sequence reads, conduct quality checks, and identify and annotate variants where patient sequence differed from the reference sequence. For each patient, the entire complement of rare variation within strongly associated candidate genes was catalogued. RESULTS: Across the panel of 169 known IBD susceptibility genes, approximately 300 variants in 104 genes were found. Excluding splicing and HLA-class variants, 58 variants across 39 of these genes were classified as rare, with an alternative allele frequency of <5%, of which 17 were novel. Only two patients with early onset Crohn's disease exhibited rare deleterious variations within NOD2: the previously described R702W variant was the sole NOD2 variant in one patient, while the second patient also carried the L1007 frameshift insertion. Both patients harboured other potentially damaging mutations in the GSDMB, ERAP2 and SEC16A genes. The two patients severely affected with ulcerative colitis exhibited a distinct profile: both carried potentially detrimental variation in the BACH2 and IL10 genes not seen in other patients. CONCLUSION: For each of the eight individuals studied, all non-synonymous, truncating and frameshift mutations across all known IBD genes were identified. A unique profile of rare and potentially damaging variants was evident for each patient with this complex disease.

Epithelial-to-Mesenchymal Transition Supports Ovarian Carcinosarcoma Tumorigenesis and Confers Sensitivity to Microtubule Targeting with Eribulin.

Ovarian carcinosarcoma (OCS) is an aggressive and rare tumor type with limited treatment options. OCS is hypothesized to develop via the combination theory, with a single progenitor resulting in carcinomatous and sarcomatous components, or alternatively via the conversion theory, with the sarcomatous component developing from the carcinomatous component through epithelial-to-mesenchymal transition (EMT). In this study, we analyzed DNA variants from isolated carcinoma and sarcoma components to show that OCS from 18 women is monoclonal. RNA sequencing indicated that the carcinoma components were more mesenchymal when compared with pure epithelial ovarian carcinomas, supporting the conversion theory and suggesting that EMT is important in the formation of these tumors. Preclinical OCS models were used to test the efficacy of microtubule-targeting drugs, including eribulin, which has previously been shown to reverse EMT characteristics in breast cancers and induce differentiation in sarcomas. Vinorelbine and eribulin more effectively inhibited OCS growth than standard-of-care platinum-based chemotherapy, and treatment with eribulin reduced mesenchymal characteristics and N-MYC expression in OCS patient-derived xenografts. Eribulin treatment resulted in an accumulation of intracellular cholesterol in OCS cells, which triggered a downregulation of the mevalonate pathway and prevented further cholesterol biosynthesis. Finally, eribulin increased expression of genes related to immune activation and increased the intratumoral accumulation of CD8+ T cells, supporting exploration of immunotherapy combinations in the clinic. Together, these data indicate that EMT plays a key role in OCS tumorigenesis and support the conversion theory for OCS histogenesis. Targeting EMT using eribulin could help improve OCS patient outcomes. SIGNIFICANCE: Genomic analyses and preclinical models of ovarian carcinosarcoma support the conversion theory for disease development and indicate that microtubule inhibitors could be used to suppress EMT and stimulate antitumor immunity.

Heterozygous frameshift variants in HNRNPA2B1 cause early-onset oculopharyngeal muscular dystrophy.

Missense variants in RNA-binding proteins (RBPs) underlie a spectrum of disease phenotypes, including amyotrophic lateral sclerosis, frontotemporal dementia, and inclusion body myopathy. Here, we present ten independent families with a severe, progressive muscular dystrophy, reminiscent of oculopharyngeal muscular dystrophy (OPMD) but of much earlier onset, caused by heterozygous frameshift variants in the RBP hnRNPA2/B1. All disease-causing frameshift mutations abolish the native stop codon and extend the reading frame, creating novel transcripts that escape nonsense-mediated decay and are translated to produce hnRNPA2/B1 protein with the same neomorphic C-terminal sequence. In contrast to previously reported disease-causing missense variants in HNRNPA2B1, these frameshift variants do not increase the propensity of hnRNPA2 protein to fibrillize. Rather, the frameshift variants have reduced affinity for the nuclear import receptor karyopherin ß2, resulting in cytoplasmic accumulation of hnRNPA2 protein in cells and in animal models that recapitulate the human pathology. Thus, we expand the phenotypes associated with HNRNPA2B1 to include an early-onset form of OPMD caused by frameshift variants that alter its nucleocytoplasmic transport dynamics.

FAK regulates IL-33 expression by controlling chromatin accessibility at c-Jun motifs.

Focal adhesion kinase (FAK) localizes to focal adhesions and is overexpressed in many cancers. FAK can also translocate to the nucleus, where it binds to, and regulates, several transcription factors, including MBD2, p53 and IL-33, to control gene expression by unknown mechanisms. We have used ATAC-seq to reveal that FAK controls chromatin accessibility at a subset of regulated genes. Integration of ATAC-seq and RNA-seq data showed that FAK-dependent chromatin accessibility is linked to differential gene expression, including of the FAK-regulated cytokine and transcriptional regulator interleukin-33 (Il33), which controls anti-tumor immunity. Analysis of the accessibility peaks on the Il33 gene promoter/enhancer regions revealed sequences for several transcription factors, including ETS and AP-1 motifs, and we show that c-Jun, a component of AP-1, regulates Il33 gene expression by binding to its enhancer in a FAK kinase-dependent manner. This work provides the first demonstration that FAK controls transcription via chromatin accessibility, identifying a novel mechanism by which nuclear FAK regulates biologically important gene expression.

Role of PLEXIND1/TGFß Signaling Axis in Pancreatic Ductal Adenocarcinoma Progression Correlates with the Mutational Status of KRAS.

PLEXIND1 is upregulated in several cancers, including pancreatic ductal adenocarcinoma (PDAC). It is an established mediator of semaphorin signaling, and neuropilins are its known coreceptors. Herein, we report data to support the proposal that PLEXIND1 acts as a transforming growth factor beta (TGFß) coreceptor, modulating cell growth through SMAD3 signaling. Our findings demonstrate that PLEXIND1 plays a pro-tumorigenic role in PDAC cells with oncogenic KRAS (KRASmut). We show in KRASmut PDAC cell lines (PANC-1, AsPC-1,4535) PLEXIND1 downregulation results in decreased cell viability (in vitro) and reduced tumor growth (in vivo). Conversely, PLEXIND1 acts as a tumor suppressor in the PDAC cell line (BxPC-3) with wild-type KRAS (KRASwt), as its reduced expression results in higher cell viability (in-vitro) and tumor growth (in vivo). Additionally, we demonstrate that PLEXIND1-mediated interactions can be selectively disrupted using a peptide based on its C-terminal sequence (a PDZ domain-binding motif), an outcome that may possess significant therapeutic implications. To our knowledge, this is the first report showing that (1) PLEXIND1 acts as a TGFß coreceptor and mediates SMAD3 signaling, and (2) differential roles of PLEXIND1 in PDAC cell lines correlate with KRASmut and KRASwt status.

Repression of the Type I Interferon Pathway Underlies MYC- and KRAS-Dependent Evasion of NK and B Cells in Pancreatic Ductal Adenocarcinoma.

MYC is implicated in the development and progression of pancreatic cancer, yet the precise level of MYC deregulation required to contribute to tumor development has been difficult to define. We used modestly elevated expression of human MYC, driven from the Rosa26 locus, to investigate the pancreatic phenotypes arising in mice from an approximation of MYC trisomy. We show that this level of MYC alone suffices to drive pancreatic neuroendocrine tumors, and to accelerate progression of KRAS-initiated precursor lesions to metastatic pancreatic ductal adenocarcinoma (PDAC). Our phenotype exposed suppression of the type I interferon (IFN) pathway by the combined actions of MYC and KRAS, and we present evidence of repressive MYC-MIZ1 complexes binding directly to the promoters of the genes encodiing the type I IFN regulators IRF5, IRF7, STAT1, and STAT2. Derepression of IFN regulator genes allows pancreatic tumor infiltration by B and natural killer (NK) cells, resulting in increased survival. SIGNIFICANCE: We define herein a novel mechanism of evasion of NK cell-mediated immunity through the combined actions of endogenously expressed mutant KRAS and modestly deregulated expression of MYC, via suppression of the type I IFN pathway. Restoration of IFN signaling may improve outcomes for patients with PDAC.This article is highlighted in the In This Issue feature, p. 747.

Feasibility and clinical utility of endoscopic ultrasound guided biopsy of pancreatic cancer for next-generation molecular profiling.

Next-generation sequencing is enabling molecularly guided therapy for many cancer types, yet failure rates remain relatively high in pancreatic cancer (PC). The aim of this study is to investigate the feasibility of genomic profiling using endoscopic ultrasound (EUS) biopsy samples to facilitate personalised therapy for PC. Ninty-five patients underwent additional research biopsies at the time of diagnostic EUS. Diagnostic formalin-fixed (FFPE) and fresh frozen EUS samples underwent DNA extraction, quantification and targeted gene sequencing. Whole genome (WGS) and RNA sequencing was performed as proof of concept. Only 2 patients (2%) with a diagnosis of PC had insufficient material for targeted sequencing in both FFPE and frozen specimens. Targeted panel sequencing (n=54) revealed mutations in PC genes (KRAS, GNAS, TP53, CDKN2A, SMAD4) in patients with histological evidence of PC, including potentially actionable mutations (BRCA1, BRCA2, ATM, BRAF). WGS (n=5) of EUS samples revealed mutational signatures that are potential biomarkers of therapeutic responsiveness. RNA sequencing (n=35) segregated patients into clinically relevant molecular subtypes based on transcriptome. Integrated multi-omic analysis of PC using standard EUS guided biopsies offers clinical utility to guide personalized therapy and study the molecular pathology in all patients with PC.

Deleterious coding variants in multi-case families with non-syndromic cleft lip and/or palate phenotypes.

Nonsyndromic Cleft Lip and/or Palate (NSCLP) is regarded as a multifactorial condition in which clefting is an isolated phenotype, distinguished from the largely monogenic, syndromic forms which include clefts among a spectrum of phenotypes. Nonsyndromic clefting has been shown to arise through complex interactions between genetic and environmental factors. However, there is increasing evidence that the broad NSCLP classification may include a proportion of cases showing familial patterns of inheritance and contain highly penetrant deleterious variation in specific genes. Through exome sequencing of multi-case families ascertained in Bogota, Colombia, we identify 28 non-synonymous single nucleotide variants that are considered damaging by at least one predictive score. We discuss the functional impact of candidate variants identified. In one family we find a coding variant in the MSX1 gene which is predicted damaging by multiple scores. This variant is in exon 2, a highly conserved region of the gene. Previous sequencing has suggested that mutations in MSX1 may account for ~2% of NSCLP. Our analysis further supports evidence that a proportion of NSCLP cases arise through monogenic coding mutations, though further work is required to unravel the complex interplay of genetics and environment involved in facial clefting.

CXCR2 Inhibition Profoundly Suppresses Metastases and Augments Immunotherapy in Pancreatic Ductal Adenocarcinoma.

CXCR2 has been suggested to have both tumor-promoting and tumor-suppressive properties. Here we show that CXCR2 signaling is upregulated in human pancreatic cancer, predominantly in neutrophil/myeloid-derived suppressor cells, but rarely in tumor cells. Genetic ablation or inhibition of CXCR2 abrogated metastasis, but only inhibition slowed tumorigenesis. Depletion of neutrophils/myeloid-derived suppressor cells also suppressed metastasis suggesting a key role for CXCR2 in establishing and maintaining the metastatic niche. Importantly, loss or inhibition of CXCR2 improved T cell entry, and combined inhibition of CXCR2 and PD1 in mice with established disease significantly extended survival. We show that CXCR2 signaling in the myeloid compartment can promote pancreatic tumorigenesis and is required for pancreatic cancer metastasis, making it an excellent therapeutic target.

Quantifying the cumulative effect of low-penetrance genetic variants on breast cancer risk.

Many common diseases have a complex genetic basis in which large numbers of genetic variations combine with environmental factors to determine risk. However, quantifying such polygenic effects has been challenging. In order to address these difficulties we developed a global measure of the information content of an individual's genome relative to a reference population, which may be used to assess differences in global genome structure between cases and appropriate controls. Informally this measure, which we call relative genome information (RGI), quantifies the relative "disorder" of an individual's genome. In order to test its ability to predict disease risk we used RGI to compare single-nucleotide polymorphism genotypes from two independent samples of women with early-onset breast cancer with three independent sets of controls. We found that RGI was significantly elevated in both sets of breast cancer cases in comparison with all three sets of controls, with disease risk rising sharply with RGI. Furthermore, these differences are not due to associations with common variants at a small number of disease-associated loci, but rather are due to the combined associations of thousands of markers distributed throughout the genome. Our results indicate that the information content of an individual's genome may be used to measure the risk of a complex disease, and suggest that early-onset breast cancer has a strongly polygenic component.

A genome wide meta-analysis study for identification of common variation associated with breast cancer prognosis.

OBJECTIVE: Genome wide association studies (GWAs) of breast cancer mortality have identified few potential associations. The concordance between these studies is unclear. In this study, we used a meta-analysis of two prognostic GWAs and a replication cohort to identify the strongest associations and to evaluate the loci suggested in previous studies. We attempt to identify those SNPs which could impact overall survival irrespective of the age of onset. METHODS: To facilitate the meta-analysis and to refine the association signals, SNPs were imputed using data from the 1000 genomes project. Cox-proportional hazard models were used to estimate hazard ratios (HR) in 536 patients from the POSH cohort (Prospective study of Outcomes in Sporadic versus Hereditary breast cancer) and 805 patients from the HEBCS cohort (Helsinki Breast Cancer Study). These hazard ratios were combined using a Mantel-Haenszel fixed effects meta-analysis and a p-value threshold of 5×10(-8) was used to determine significance. Replication was performed in 1523 additional patients from the POSH study. RESULTS: Although no SNPs achieved genome wide significance, three SNPs have significant association in the replication cohort and combined p-values less than 5.6×10(-6). These SNPs are; rs421379 which is 556 kb upstream of ARRDC3 (HR = 1.49, 95% confidence interval (CI) = 1.27-1.75, P = 1.1×10(-6)), rs12358475 which is between ECHDC3 and PROSER2 (HR = 0.75, CI = 0.67-0.85, P = 1.8×10(-6)), and rs1728400 which is between LINC00917 and FOXF1. CONCLUSIONS: In a genome wide meta-analysis of two independent cohorts from UK and Finland, we identified potential associations at three distinct loci. Phenotypic heterogeneity and relatively small sample sizes may explain the lack of genome wide significant findings. However, the replication at three SNPs in the validation cohort shows promise for future studies in larger cohorts. We did not find strong evidence for concordance between the few associations highlighted by previous GWAs of breast cancer survival and this study.

Support Vector Machine classifier for estrogen receptor positive and negative early-onset breast cancer.

Two major breast cancer sub-types are defined by the expression of estrogen receptors on tumour cells. Cancers with large numbers of receptors are termed estrogen receptor positive and those with few are estrogen receptor negative. Using genome-wide single nucleotide polymorphism genotype data for a sample of early-onset breast cancer patients we developed a Support Vector Machine (SVM) classifier from 200 germline variants associated with estrogen receptor status (p<0.0005). Using a linear kernel Support Vector Machine, we achieved classification accuracy exceeding 93%. The model indicates that polygenic variation in more than 100 genes is likely to underlie the estrogen receptor phenotype in early-onset breast cancer. Functional classification of the genes involved identifies enrichment of functions linked to the immune system, which is consistent with the current understanding of the biological role of estrogen receptors in breast cancer.