Mining the coding and non-coding genome for cancer drivers

Optimized high-throughput screening of non-coding variants identified from genome-wide association studies

The vast majority of disease-associated single nucleotide polymorphisms (SNP) identified from genome-wide association studies (GWAS) are localized in non-coding regions. A significant fraction of these variants impact transcription factors binding to enhancer elements and alter gene expression. To functionally interrogate the activity of such variants we developed snpSTARRseq, a high-throughput experimental method that can interrogate the functional impact of hundreds to thousands of non-coding variants on enhancer activity. snpSTARRseq dramatically improves signal-to-noise by utilizing a novel sequencing and bioinformatic approach that increases both insert size and the number of variants tested per loci. Using this strategy, we interrogated known prostate cancer (PCa) risk-associated loci and demonstrated that 35% of them harbor SNPs that significantly altered enhancer activity. Combining these results with chromosomal looping data we could identify interacting genes and provide a mechanism of action for 20 PCa GWAS risk regions. When benchmarked to orthogonal methods, snpSTARRseq showed a strong correlation with in vivo experimental allelic-imbalance studies whereas there was no correlation with predictive in silico approaches. Overall, snpSTARRseq provides an integrated experimental and computational framework to functionally test non-coding genetic variants.

Non-Coding Variants in Cancer: Mechanistic Insights and Clinical Potential for Personalized Medicine

The cancer genome is characterized by extensive variability, in the form of Single Nucleotide Polymorphisms (SNPs) or structural variations such as Copy Number Alterations (CNAs) across wider genomic areas. At the molecular level, most SNPs and/or CNAs reside in non-coding sequences, ultimately affecting the regulation of oncogenes and/or tumor-suppressors in a cancer-specific manner. Notably, inherited non-coding variants can predispose for cancer decades prior to disease onset. Furthermore, accumulation of additional non-coding driver mutations during progression of the disease, gives rise to genomic instability, acting as the driving force of neoplastic development and malignant evolution. Therefore, detection and characterization of such mutations can improve risk assessment for healthy carriers and expand the diagnostic and therapeutic toolbox for the patient. This review focuses on functional variants that reside in transcribed or not transcribed non-coding regions of the cancer genome and presents a collection of appropriate state-of-the-art methodologies to study them.

2-kupl: mapping-free variant detection from DNA-seq data of matched samples

BACKGROUND

The detection of genome variants, including point mutations, indels and structural variants, is a fundamental and challenging computational problem. We address here the problem of variant detection between two deep-sequencing (DNA-seq) samples, such as two human samples from an individual patient, or two samples from distinct bacterial strains. The preferred strategy in such a case is to align each sample to a common reference genome, collect all variants and compare these variants between samples. Such mapping-based protocols have several limitations. DNA sequences with large indels, aggregated mutations and structural variants are hard to map to the reference. Furthermore, DNA sequences cannot be mapped reliably to genomic low complexity regions and repeats.

RESULTS

We introduce 2-kupl, a k-mer based, mapping-free protocol to detect variants between two DNA-seq samples. On simulated and actual data, 2-kupl achieves higher accuracy than other mapping-free protocols. Applying 2-kupl to prostate cancer whole exome sequencing data, we identify a number of candidate variants in hard-to-map regions and propose potential novel recurrent variants in this disease.

CONCLUSIONS

We developed a mapping-free protocol for variant calling between matched DNA-seq samples. Our protocol is suitable for variant detection in unmappable genome regions or in the absence of a reference genome.

ncVarDB: a manually curated database for pathogenic non-coding variants and benign controls

Variants within the non-coding genome are frequently associated with phenotypes in genome-wide association studies. These non-coding regions may be involved in the regulation of gene expression, encode functional non-coding RNAs, or influence splicing and other cellular functions. We have curated a list of characterized non-coding human genome variants based on the published evidence that indicates phenotypic consequences of the variation. In order to minimize annotation errors, two curators have independently verified the supporting evidence for pathogenicity of each non-coding variant in the published literature. The database consists of 721 non-coding variants linked to the published literature describing the evidence of functional consequences. We have also sampled 7228 covariate-matched benign controls, that have a population frequency of over 5%, from the single nucleotide polymorphism database (dbSNP151) database. These were sampled controlling for potential confounding factors such as linkage with pathogenic variants, annotation type (untranslated region, intron, intergenic, etc.) and variant type (substitution or indel). The dataset presented here represents a curated repository, with a potential use for the training or evaluation of algorithms used in the prediction of non-coding variant functionality.

Database URL: https://github.com/Gardner-BinfLab/ncVarDB.

Functional coding/non-coding variants in EGFR, ROS1 and ALK genes and their role in liquid biopsy as a personalized therapy

Personalized medicine holds promise to tailor the treatment options for patients' unique genetic make-up, behavioral and environmental background. Liquid biopsy is non-invasive technique and precise diagnosis and treatment approach. Significantly, NGS technologies have revolutionized the genomic medicine by novel identifying SNPs, indel mutations in both coding and non-coding regions and also a promising technology to accelerate the early detection and finding new biomarkers for diagnosis and treatment. The number of the bioinformatics tools have been rapidly increasing with the aim of learning more about the detected mutations either they have a pathogenic role or not. EGFR, ROS1 and ALK genes are members of the RTK family. Until now, mutations within these genes have been associated with many cancers and involved in resistance formation to TKIs. This review article summarized the findings about the mostly investigated variations in EGFR, ROS1 and ALK genes and their potential role in liquid biopsy approach.

A benchmark study of scoring methods for non-coding mutations

Motivation

Detailed knowledge of coding sequences has led to different candidate models for pathogenic variant prioritization. Several deleteriousness scores have been proposed for the non-coding part of the genome, but no large-scale comparison has been realized to date to assess their performance.

Results

We compared the leading scoring tools (CADD, FATHMM-MKL, Funseq2 and GWAVA) and some recent competitors (DANN, SNP and SOM scores) for their ability to discriminate assumed pathogenic variants from assumed benign variants (using the ClinVar, COSMIC and 1000 genomes project databases). Using the ClinVar benchmark, CADD was the best tool for detecting the pathogenic variants that are mainly located in protein coding gene regions. Using the COSMIC benchmark, FATHMM-MKL, GWAVA and SOMliver outperformed the other tools for pathogenic variants that are typically located in lincRNAs, pseudogenes and other parts of the non-coding genome. However, all tools had low precision, which could potentially be improved by future non-coding genome feature discoveries. These results may have been influenced by the presence of potential benign variants in the COSMIC database. The development of a gold standard as consistent as ClinVar for these regions will be necessary to confirm our tool ranking.

Availability and implementation

The Snakemake, C++ and R codes are freely available from https://github.com/Oncostat/BenchmarkNCVTools and supported on Linux.

Contact

damien.drubay@gustaveroussy.fr or stefan.michiels@gustaveroussy.fr.

Supplementary information

Supplementary data are available at Bioinformatics online.

Integration of Random Forest Classifiers and Deep Convolutional Neural Networks for Classification and Biomolecular Modeling of Cancer Driver Mutations

Development of machine learning solutions for prediction of functional and clinical significance of cancer driver genes and mutations are paramount in modern biomedical research and have gained a significant momentum in a recent decade. In this work, we integrate different machine learning approaches, including tree based methods, random forest and gradient boosted tree (GBT) classifiers along with deep convolutional neural networks (CNN) for prediction of cancer driver mutations in the genomic datasets. The feasibility of CNN in using raw nucleotide sequences for classification of cancer driver mutations was initially explored by employing label encoding, one hot encoding, and embedding to preprocess the DNA information. These classifiers were benchmarked against their tree-based alternatives in order to evaluate the performance on a relative scale. We then integrated DNA-based scores generated by CNN with various categories of conservational, evolutionary and functional features into a generalized random forest classifier. The results of this study have demonstrated that CNN can learn high level features from genomic information that are complementary to the ensemble-based predictors often employed for classification of cancer mutations. By combining deep learning-generated score with only two main ensemble-based functional features, we can achieve a superior performance of various machine learning classifiers. Our findings have also suggested that synergy of nucleotide-based deep learning scores and integrated metrics derived from protein sequence conservation scores can allow for robust classification of cancer driver mutations with a limited number of highly informative features. Machine learning predictions are leveraged in molecular simulations, protein stability, and network-based analysis of cancer mutations in the protein kinase genes to obtain insights about molecular signatures of driver mutations and enhance the interpretability of cancer-specific classification models.

Epigenomic annotation of noncoding mutations identifies mutated pathways in primary liver cancer

Evidence that noncoding mutation can result in cancer driver events is mounting. However, it is more difficult to assign molecular biological consequences to noncoding mutations than to coding mutations, and a typical cancer genome contains many more noncoding mutations than protein-coding mutations. Accordingly, parsing functional noncoding mutation signal from noise remains an important challenge. Here we use an empirical approach to identify putatively functional noncoding somatic single nucleotide variants (SNVs) from liver cancer genomes. Annotation of candidate variants by publicly available epigenome datasets finds that 40.5% of SNVs fall in regulatory elements. When assigned to specific regulatory elements, we find that the distribution of regulatory element mutation mirrors that of nonsynonymous coding mutation, where few regulatory elements are recurrently mutated in a patient population but many are singly mutated. We find potential gain-of-binding site events among candidate SNVs, suggesting a mechanism of action for these variants. When aggregating noncoding somatic mutation in promoters, we find that genes in the ERBB signaling and MAPK signaling pathways are significantly enriched for promoter mutations. Altogether, our results suggest that functional somatic SNVs in cancer are sporadic, but occasionally occur in regulatory elements and may affect phenotype by creating binding sites for transcriptional regulators. Accordingly, we propose that noncoding mutation should be formally accounted for when determining gene- and pathway-mutation burden in cancer.

Prioritization of non-coding disease-causing variants and long non-coding RNAs in liver cancer

There are multiple bioinformatics tools available for the detection of coding driver mutations in cancers. However, the prioritization of pathogenic non-coding variants remains a challenging and demanding task. The present study was performed to discriminate non-coding disease-causing mutations and prioritize potential cancer-implicated long non-coding RNAs (lncRNAs) in liver cancer using a logistic regression model. A logistic regression model was constructed by combining 19,153 disease-associated ClinVar and human gene mutation database pathogenic variants as the response variable and non-coding features as the predictor variable. Genome-wide association study (GWAS) disease or trait-associated variants and recurrent somatic mutations were used to validate the model. Non-coding gene features with the highest fractions of load were characterized and potential cancer-associated lncRNA candidates were prioritized by combining the fraction of high-scoring regions and average score predicted by the logistic regression model. H3K9me3 and conserved regions were the most negatively and positively informative for the model, respectively. The area under the receiver operating characteristic curve of the model was 0.92. The average score of GWAS disease-associated variants was significantly increased compared with neutral single nucleotide polymorphisms (5.8642 vs. 5.4707; P<0.001), the average score of recurrent somatic mutations of liver cancer was significantly increased compared with non-recurrent somatic mutations (5.4101 vs. 5.2768; P=0.0125). The present study found regions in lncRNAs and introns/untranslated regions of protein coding genes where mutations are most likely to be damaging. In total, 847 lncRNAs were filtered out from the background. Characterization of this subset of lncRNAs showed that these lncRNAs are more conservative, less mutated and more highly expressed compared with other control lncRNAs. In addition, 23 of these lncRNAs were differentially expressed between 12 pairs of liver cancer and adjacent normal specimens. The logistic regression model is a useful tool to prioritize non-coding pathogenic variants and lncRNAs, and paves the way for the detection of non-coding driver lncRNAs in liver cancer.

Functional annotation of noncoding variants and prioritization of cancer-associated lncRNAs in lung cancer

Multiple computational tools have been widely applied to the detection of coding driver mutations in cancer; however, the prioritization of pathogenic non-coding variants remains a difficult and demanding task. The present study was performed to distinguish non-coding disease-causing mutations from neutral ones, and to prioritize potential cancer-associated long non-coding RNAs (lncRNAs) with a logistic regression model in lung cancer. A logistic regression model was constructed, combining 19,153 disease-associated ClinVar and Human Gene Mutation Database pathogenic variants as the response variable and non-coding features as the predictor variable. Validation of the model was conducted with genome-wide association study (GWAS) disease- or trait-associated single nucleotide polymorphisms (SNPs) and recurrent somatic mutations. High scoring regions were characterized with respect to their distribution in various features and gene classes; potential cancer-associated lncRNA candidates were prioritized, combining the fraction of high-scoring regions and average score predicted by the logistic regression model. H3K79me2 was the most negative factor that contributed to the model, while conserved regions were most positively informative to the model. The area under the receiver operating characteristic curve of the model was 0.89. The model assigned a significantly higher score to GWAS SNPs and recurrent somatic mutations compared with neutral SNPs (mean, 5.9012 vs. 5.5238; P<0.001, Mann-Whitney U test) and non-recurrent mutations (mean, 5.4677 vs. 5.2277, P<0.001, Mann-Whitney U test), respectively. It was observed that regions, including splicing sites and untranslated regions, and gene classes, including cancer genes and cancer-associated lncRNAs, had an increased enrichment of high-scoring regions. In total, 2,679 cancer-associated lncRNAs were determined and characterized. A total of 104 of these lncRNAs were differentially expressed between lung cancer and normal specimens. The logistic regression model is a useful and efficient scoring system to prioritize non-coding pathogenic variants and lncRNAs, and may provide the basis for detecting non-coding driver lncRNAs in lung cancer.

A novel cell cycle-associated lncRNA, HOXA11-AS, is transcribed from the 5-prime end of the HOXA transcript and is a biomarker of progression in glioma

The comprehensive lncRNA expression signature in glioma has not yet been fully elucidated. We performed a high-throughput microarray to detect the ncRNA expression profiles of 220 human glioma tissues. Here, we found that a novel lncRNA, HOXA11-AS, was the antisense transcript of the HOX11 gene. It was shown that HOXA11-AS was closely associated with glioma grade and poor prognosis. Multivariate Cox regression analysis revealed that HOXA11-AS was an independent prognostic factor in glioblastoma multiforme patients, and its expression was correlated with the glioma molecular subtypes of the Cancer Genome Atlas. Gene set enrichment analysis indicated that the gene sets most correlated with HOXA11-AS expression were involved in cell cycle progression. Over-expression of the HOXA11-AS transcript promoted cell proliferation in vitro, while knockdown of HOXA11-AS expression repressed cell proliferation via regulation of cell cycle progression. The growth-promoting and growth-inhibiting effects of HOXA11-AS were also demonstrated in a xenograft mouse model. Our data confirms, for the first time, that HOXA11-AS is an important long non-coding RNA that primarily serves as a prognostic factor for glioma patient survival. HOXA11-AS could serve as a biomarker for identifying glioma molecular subtypes and as therapeutic target for glioma patients.