Assessment of algorithms for high throughput detection of genomic copy number variation in oligonucleotide microarray data

Background

Genomic deletions and duplications are important in the pathogenesis of diseases, such as cancer and mental retardation, and have recently been shown to occur frequently in unaffected individuals as polymorphisms. Affymetrix GeneChip whole genome sampling analysis (WGSA) combined with 100 K single nucleotide polymorphism (SNP) genotyping arrays is one of several microarray-based approaches that are now being used to detect such structural genomic changes. The popularity of this technology and its associated open source data format have resulted in the development of an increasing number of software packages for the analysis of copy number changes using these SNP arrays.

Results

We evaluated four publicly available software packages for high throughput copy number analysis using synthetic and empirical 100 K SNP array data sets, the latter obtained from 107 mental retardation (MR) patients and their unaffected parents and siblings. We evaluated the software with regards to overall suitability for high-throughput 100 K SNP array data analysis, as well as effectiveness of normalization, scaling with various reference sets and feature extraction, as well as true and false positive rates of genomic copy number variant (CNV) detection.

Conclusion

We observed considerable variation among the numbers and types of candidate CNVs detected by different analysis approaches, and found that multiple programs were needed to find all real aberrations in our test set. The frequency of false positive deletions was substantial, but could be greatly reduced by using the SNP genotype information to confirm loss of heterozygosity.     

CopywriteR: DNA copy number detection from off-target sequence data

Current methods for detection of copy number variants (CNV) and aberrations (CNA) from targeted sequencing data are based on the depth of coverage of captured exons. Accurate CNA determination is complicated by uneven genomic distribution and non-uniform capture efficiency of targeted exons. Here we present CopywriteR, which eludes these problems by exploiting ‘off-target’ sequence reads. CopywriteR allows for extracting uniformly distributed copy number information, can be used without reference, and can be applied to sequencing data obtained from various techniques including chromatin immunoprecipitation and target enrichment on small gene panels. CopywriteR outperforms existing methods and constitutes a widely applicable alternative to available tools.

Electronic supplementary material

The online version of this article (doi:10.1186/s13059-015-0617-1) contains supplementary material, which is available to authorized users.     

rSW-seq: Algorithm for detection of copy number alterations in deep sequencing data

Background  

Recent advances in sequencing technologies have enabled generation of large-scale genome sequencing data. These data can be used to characterize a variety of genomic features, including the DNA copy number profile of a cancer genome. A robust and reliable method for screening chromosomal alterations would allow a detailed characterization of the cancer genome with unprecedented accuracy.     

Breaking the waves: improved detection of copy number variation from microarray-based comparative genomic hybridization

Background

Large-scale high throughput studies using microarray technology have established that copy number variation (CNV) throughout the genome is more frequent than previously thought. Such variation is known to play an important role in the presence and development of phenotypes such as HIV-1 infection and Alzheimer's disease. However, methods for analyzing the complex data produced and identifying regions of CNV are still being refined.

Results

We describe the presence of a genome-wide technical artifact, spatial autocorrelation or 'wave', which occurs in a large dataset used to determine the location of CNV across the genome. By removing this artifact we are able to obtain both a more biologically meaningful clustering of the data and an increase in the number of CNVs identified by current calling methods without a major increase in the number of false positives detected. Moreover, removing this artifact is critical for the development of a novel model-based CNV calling algorithm - CNVmix - that uses cross-sample information to identify regions of the genome where CNVs occur. For regions of CNV that are identified by both CNVmix and current methods, we demonstrate that CNVmix is better able to categorize samples into groups that represent copy number gains or losses.

Conclusion

Removing artifactual 'waves' (which appear to be a general feature of array comparative genomic hybridization (aCGH) datasets) and using cross-sample information when identifying CNVs enables more biological information to be extracted from aCGH experiments designed to investigate copy number variation in normal individuals.     

Fuzzy methods for the detection of copy number variations in comparative genomic hybridization arrays

Genomic copy number variations (CNVs) are considered as a significant source of genetic diversity and widely involved in gene expression and regulatory mechanism, genetic disorders and disease risk, susceptibility to certain diseases and conditions, and resistance to medical drugs. Many studies have targeted the identification, profiling, analysis, and associations of genetic CNVs. We propose herein two new fuzzy methods, taht is, one based on the fuzzy inference from the pre-processed input, and another based on fuzzy C-means clustering. Our solutions present a higher true positive rate and a lower false negative with no false positive, efficient performance and consumption of least resources.     

Accurate in silico confirmation of rare copy number variant calls from exome sequencing data using transfer learning

Exome sequencing is widely used in genetic studies of human diseases and clinical genetic diagnosis. Accurate detection of copy number variants (CNVs) is important to fully utilize exome sequencing data. However, exome data are noisy. None of the existing methods alone can achieve both high precision and recall rate. A common practice is to perform heuristic filtration followed by manual inspection of read depth of putative CNVs. This approach does not scale in large studies. To address this issue, we developed a transfer learning method, CNV-espresso, for in silico confirming rare CNVs from exome sequencing data. CNV-espresso encodes candidate CNVs from exome data as images and uses pretrained convolutional neural network models to classify copy number states. We trained CNV-espresso using an offspring–parents trio exome sequencing dataset, with inherited CNVs as positives and CNVs with Mendelian errors as negatives. We evaluated the performance using additional samples that have both exome and whole-genome sequencing (WGS) data. Assuming the CNVs detected from WGS data as a proxy of ground truth, CNV-espresso significantly improves precision while keeping recall almost intact, especially for CNVs that span a small number of exons. CNV-espresso can effectively replace manual inspection of CNVs in large-scale exome sequencing studies.     

Sparse representation and Bayesian detection of genome copy number alterations from microarray data

MOTIVATION: Genomic instability in cancer leads to abnormal genome copy number alterations (CNA) that are associated with the development and behavior of tumors. Advances in microarray technology have allowed for greater resolution in detection of DNA copy number changes (amplifications or deletions) across the genome. However, the increase in number of measured signals and accompanying noise from the array probes present a challenge in accurate and fast identification of breakpoints that define CNA. This article proposes a novel detection technique that exploits the use of piece wise constant (PWC) vectors to represent genome copy number and sparse Bayesian learning (SBL) to detect CNA breakpoints. METHODS: First, a compact linear algebra representation for the genome copy number is developed from normalized probe intensities. Second, SBL is applied and optimized to infer locations where copy number changes occur. Third, a backward elimination (BE) procedure is used to rank the inferred breakpoints; and a cut-off point can be efficiently adjusted in this procedure to control for the false discovery rate (FDR). RESULTS: The performance of our algorithm is evaluated using simulated and real genome datasets and compared to other existing techniques. Our approach achieves the highest accuracy and lowest FDR while improving computational speed by several orders of magnitude. The proposed algorithm has been developed into a free standing software application (GADA, Genome Alteration Detection Algorithm). AVAILABILITY: http://biron.usc.edu/~piquereg/GADA     

GEAR: genomic enrichment analysis of regional DNA copy number changes

We developed an algorithm named GEAR (genomic enrichment analysis of regional DNA copy number changes) for functional interpretation of genome-wide DNA copy number changes identified by array-based comparative genomic hybridization. GEAR selects two types of chromosomal alterations with potential biological relevance, i.e. recurrent and phenotype-specific alterations. Then it performs functional enrichment analysis using a priori selected functional gene sets to identify primary and clinical genomic signatures. The genomic signatures identified by GEAR represent functionally coordinated genomic changes, which can provide clues on the underlying molecular mechanisms related to the phenotypes of interest. GEAR can help the identification of key molecular functions that are activated or repressed in the tumor genomes leading to the improved understanding on the tumor biology. AVAILABILITY: GEAR software is available with online manual in the website, http://www.systemsbiology.co.kr/GEAR/.     

Distinct genomic copy number in mitochondria of different mammalian organs

This study shows that mitochondria in liver, kidney, heart, and brain of the mouse have a distinct mitochondrial density. It also demonstrates that the mtDNA copy number per mitochondrion is organ-specific. A reliable method of determining mitochondrial density per organ is by stereological analysis of tissue sections while mtDNA quantitation is by the use of radiolabelled mtDNA probe. This is the first study in which a comprehensive examination of mitochondrial density and quantitation of mitochondrial genomes in mouse organs have been done. In summary the variability is not only in mitochondrial density but also in genomic copy number in mitochondria of various tissues.     

Array-based comparative genomic hybridization and copy number variation in cancer research

Array-based comparative genomic hybridization (aCGH) is a molecular cytogenetic technique used in detecting and mapping DNA copy number alterations. aCGH is able to interrogate the entire genome at a previously unattainable, high resolution and has directly led to the recent appreciation of a novel class of genomic variation: copy number variation (CNV) in mammalian genomes. All forms of DNA variation/polymorphism are important for studying the basis of phenotypic diversity among individuals. CNV research is still at its infancy, requiring careful collation and annotation of accumulating CNV data that will undoubtedly be useful for accurate interpretation of genomic imbalances identified during cancer research.     

Minimum error calibration and normalization for genomic copy number analysis

BackgroundCopy number variations (CNV) are regional deviations from the normal autosomal bi-allelic DNA content. While germline CNVs are a major contributor to genomic syndromes and inherited diseases, the majority of cancers accumulate extensive “somatic” CNV (sCNV or CNA) during the process of oncogenetic transformation and progression. While specific sCNV have closely been associated with tumorigenesis, intriguingly many neoplasias exhibit recurrent sCNV patterns beyond the involvement of a few cancer driver genes. Currently, CNV profiles of tumor samples are generated using genomic micro-arrays or high-throughput DNA sequencing. Regardless of the underlying technology, genomic copy number data is derived from the relative assessment and integration of multiple signals, with the data generation process being prone to contamination from several sources. Estimated copy number values have no absolute or strictly linear correlation to their corresponding DNA levels, and the extent of deviation differs between sample profiles, which poses a great challenge for data integration and comparison in large scale genome analysis.ResultsIn this study, we present a novel method named “Minimum Error Calibration and Normalization for Copy Numbers Analysis” (Mecan4CNA). It only requires CNV segmentation files as input, is platform independent, and has a high performance with limited hardware requirements. For a given multi-sample copy number dataset, Mecan4CNA can batch-normalize all samples to the corresponding true copy number levels of the main tumor clones. Experiments of Mecan4CNA on simulated data showed an overall accuracy of 93% and 91% in determining the normal level and single copy alteration (i.e. duplication or loss of one allele), respectively. Comparison of estimated normal levels and single copy alternations with existing methods and karyotyping data on the NCI-60 tumor cell line produced coherent results. To estimate the method's impact on downstream analyses, we performed GISTIC analyses on the original and Mecan4CNA normalized data from the Cancer Genome Atlas (TCGA) where the normalized data showed prominent improvements of both sensitivity and specificity in detecting focal regions.ConclusionsMecan4CNA provides an advanced method for CNA data normalization, especially in meta-analyses involving large profile numbers and heterogeneous source data quality. With its informative output and visualization options, Mecan4CNA also can improve the interpretation of individual CNA profiles. Mecan4CNA is freely available as a Python package and through its code repository on Github.     

Exploiting noise in array CGH data to improve detection of DNA copy number change

Developing effective methods for analyzing array-CGH data to detect chromosomal aberrations is very important for the diagnosis of pathogenesis of cancer and other diseases. Current analysis methods, being largely based on smoothing and/or segmentation, are not quite capable of detecting both the aberration regions and the boundary break points very accurately. Furthermore, when evaluating the accuracy of an algorithm for analyzing array-CGH data, it is commonly assumed that noise in the data follows normal distribution. A fundamental question is whether noise in array-CGH is indeed Gaussian, and if not, can one exploit the characteristics of noise to develop novel analysis methods that are capable of detecting accurately the aberration regions as well as the boundary break points simultaneously? By analyzing bacterial artificial chromosomes (BACs) arrays with an average 1 mb resolution, 19 k oligo arrays with the average probe spacing <100 kb and 385 k oligo arrays with the average probe spacing of about 6 kb, we show that when there are aberrations, noise in all three types of arrays is highly non-Gaussian and possesses long-range spatial correlations, and that such noise leads to worse performance of existing methods for detecting aberrations in array-CGH than the Gaussian noise case. We further develop a novel method, which has optimally exploited the character of the noise, and is capable of identifying both aberration regions as well as the boundary break points very accurately. Finally, we propose a new concept, posteriori signal-to-noise ratio (p-SNR), to assign certain confidence level to an aberration region and boundaries detected.     

Construction and use of spotted large-insert clone DNA microarrays for the detection of genomic copy number changes

Microarray-based comparative genomic hybridization has become a widespread method for the analysis of DNA copy number changes across the human genome. Initial methods for microarray construction using large-insert clones required the preparation of DNA from large-scale cultures. This rapidly became an expensive and time-consuming process when expanded to the number of clones needed for higher resolution arrays. To overcome this problem, several PCR-based strategies have been developed to enable array construction from small amounts of cloned DNA. Here, we describe the construction of microarrays composed of human-specific large-insert clones (40-200 kb) using a specific degenerate oligonucleotide PCR strategy. In addition, we also describe array hybridization using manual and automated procedures and methods for array analysis. The technology and protocols described in this article can easily be adapted for other species dependent on the availability of clone libraries. According to our protocols, the procedure will take approximately 3 days from labeling the DNA to scanning the hybridized slides.     

cnvHiTSeq: integrative models for high-resolution copy number variation detection and genotyping using population sequencing data

Recent advances in sequencing technologies provide the means for identifying copy number variation (CNV) at an unprecedented resolution. A single next-generation sequencing experiment offers several features that can be used to detect CNV, yet current methods do not incorporate all available signatures into a unified model. cnvHiTSeq is an integrative probabilistic method for CNV discovery and genotyping that jointly analyzes multiple features at the population level. By combining evidence from complementary sources, cnvHiTSeq achieves high genotyping accuracy and a substantial improvement in CNV detection sensitivity over existing methods, while maintaining a low false discovery rate. cnvHiTSeq is available at http://sourceforge.net/projects/cnvhitseq     

Identification of germline genomic copy number variation in familial pancreatic cancer

Adenocarcinoma of the pancreas is a significant cause of cancer mortality, and up to 10?% of cases appear to be familial. Heritable genomic copy number variants (CNVs) can modulate gene expression and predispose to disease. Here, we identify candidate predisposition genes for familial pancreatic cancer (FPC) by analyzing germline losses or gains present in one or more high-risk patients and absent in a large control group. A total of 120 FPC cases and 1,194 controls were genotyped on the Affymetrix 500K array, and 36 cases and 2,357 controls were genotyped on the Affymetrix 6.0 array. Detection of CNVs was performed by multiple computational algorithms and partially validated by quantitative PCR. We found no significant difference in the germline CNV profiles of cases and controls. A total of 93 non-redundant FPC-specific CNVs (53 losses and 40 gains) were identified in 50 cases, each CNV present in a single individual. FPC-specific CNVs overlapped the coding region of 88 RefSeq genes. Several of these genes have been reported to be differentially expressed and/or affected by copy number alterations in pancreatic adenocarcinoma. Further investigation in high-risk subjects may elucidate the role of one or more of these genes in genetic predisposition to pancreatic cancer.     

Inferring copy number and genotype in tumour exome data

Background

Using whole exome sequencing to predict aberrations in tumours is a cost effective alternative to whole genome sequencing, however is predominantly used for variant detection and infrequently utilised for detection of somatic copy number variation.

Results

We propose a new method to infer copy number and genotypes using whole exome data from paired tumour/normal samples. Our algorithm uses two Hidden Markov Models to predict copy number and genotypes and computationally resolves polyploidy/aneuploidy, normal cell contamination and signal baseline shift. Our method makes explicit detection on chromosome arm level events, which are commonly found in tumour samples. The methods are combined into a package named ADTEx (Aberration Detection in Tumour Exome). We applied our algorithm to a cohort of 17 in-house generated and 18 TCGA paired ovarian cancer/normal exomes and evaluated the performance by comparing against the copy number variations and genotypes predicted using Affymetrix SNP 6.0 data of the same samples. Further, we carried out a comparison study to show that ADTEx outperformed its competitors in terms of precision and F-measure.

Conclusions

Our proposed method, ADTEx, uses both depth of coverage ratios and B allele frequencies calculated from whole exome sequencing data, to predict copy number variations along with their genotypes. ADTEx is implemented as a user friendly software package using Python and R statistical language. Source code and sample data are freely available under GNU license (GPLv3) at http://adtex.sourceforge.net/.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-732) contains supplementary material, which is available to authorized users.