Distance-based clustering of CGH data

MOTIVATION: We consider the problem of clustering a population of Comparative Genomic Hybridization (CGH) data samples. The goal is to develop a systematic way of placing patients with similar CGH imbalance profiles into the same cluster. Our expectation is that patients with the same cancer types will generally belong to the same cluster as their underlying CGH profiles will be similar. RESULTS: We focus on distance-based clustering strategies. We do this in two steps. (1) Distances of all pairs of CGH samples are computed. (2) CGH samples are clustered based on this distance. We develop three pairwise distance/similarity measures, namely raw, cosine and sim. Raw measure disregards correlation between contiguous genomic intervals. It compares the aberrations in each genomic interval separately. The remaining measures assume that consecutive genomic intervals may be correlated. Cosine maps pairs of CGH samples into vectors in a high-dimensional space and measures the angle between them. Sim measures the number of independent common aberrations. We test our distance/similarity measures on three well known clustering algorithms, bottom-up, top-down and k-means with and without centroid shrinking. Our results show that sim consistently performs better than the remaining measures. This indicates that the correlation of neighboring genomic intervals should be considered in the structural analysis of CGH datasets. The combination of sim with top-down clustering emerged as the best approach. AVAILABILITY: All software developed in this article and all the datasets are available from the authors upon request. CONTACT: juliu@cise.ufl.edu.     

Classification and feature selection algorithms for multi-class CGH data

Recurrent chromosomal alterations provide cytological and molecular positions for the diagnosis and prognosis of cancer. Comparative genomic hybridization (CGH) has been useful in understanding these alterations in cancerous cells. CGH datasets consist of samples that are represented by large dimensional arrays of intervals. Each sample consists of long runs of intervals with losses and gains. In this article, we develop novel SVM-based methods for classification and feature selection of CGH data. For classification, we developed a novel similarity kernel that is shown to be more effective than the standard linear kernel used in SVM. For feature selection, we propose a novel method based on the new kernel that iteratively selects features that provides the maximum benefit for classification. We compared our methods against the best wrapper-based and filter-based approaches that have been used for feature selection of large dimensional biological data. Our results on datasets generated from the Progenetix database, suggests that our methods are considerably superior to existing methods. AVAILABILITY: All software developed in this article can be downloaded from http://plaza.ufl.edu/junliu/feature.tar.gz.     

Array CGH data modeling and smoothing in Stationary Wavelet Packet Transform domain

Background

Array-based comparative genomic hybridization (array CGH) is a highly efficient technique, allowing the simultaneous measurement of genomic DNA copy number at hundreds or thousands of loci and the reliable detection of local one-copy-level variations. Characterization of these DNA copy number changes is important for both the basic understanding of cancer and its diagnosis. In order to develop effective methods to identify aberration regions from array CGH data, many recent research work focus on both smoothing-based and segmentation-based data processing. In this paper, we propose stationary packet wavelet transform based approach to smooth array CGH data. Our purpose is to remove CGH noise in whole frequency while keeping true signal by using bivariate model.

Results

In both synthetic and real CGH data, Stationary Wavelet Packet Transform (SWPT) is the best wavelet transform to analyze CGH signal in whole frequency. We also introduce a new bivariate shrinkage model which shows the relationship of CGH noisy coefficients of two scales in SWPT. Before smoothing, the symmetric extension is considered as a preprocessing step to save information at the border.

Conclusion

We have designed the SWTP and the SWPT-Bi which are using the stationary wavelet packet transform with the hard thresholding and the new bivariate shrinkage estimator respectively to smooth the array CGH data. We demonstrate the effectiveness of our approach through theoretical and experimental exploration of a set of array CGH data, including both synthetic data and real data. The comparison results show that our method outperforms the previous approaches.

     

Comparing genomes with duplications: a computational complexity point of view

In this paper, we are interested in the computational complexity of computing (dis)similarity measures between two genomes when they contain duplicated genes or genomic markers, a problem that happens frequently when comparing whole nuclear genomes. Recently, several methods ( [1], [2]) have been proposed that are based on two steps to compute a given (dis)similarity measure M between two genomes G_1 and G_2: first, one establishes a oneto- one correspondence between genes of G_1 and genes of G_2 ; second, once this correspondence is established, it defines explicitly a permutation and it is then possible to quantify their similarity using classical measures defined for permutations, like the number of breakpoints. Hence these methods rely on two elements: a way to establish a one-to-one correspondence between genes of a pair of genomes, and a (dis)similarity measure for permutations. The problem is then, given a (dis)similarity measure for permutations, to compute a correspondence that defines an optimal permutation for this measure. We are interested here in two models to compute a one-to-one correspondence: the exemplar model, where all but one copy are deleted in both genomes for each gene family, and the matching model, that computes a maximal correspondence for each gene family. We show that for these two models, and for three (dis)similarity measures on permutations, namely the number of common intervals, the maximum adjacency disruption (MAD) number and the summed adjacency disruption (SAD) number, the problem of computing an optimal correspondence is NP-complete, and even APXhard for the MAD number and SAD number.     

Weighted clustering of called array CGH data

Array comparative genomic hybridization (aCGH) is a laboratory technique to measure chromosomal copy number changes. A clear biological interpretation of the measurements is obtained by mapping these onto an ordinal scale with categories loss/normal/gain of a copy. The pattern of gains and losses harbors a level of tumor specificity. Here, we present WECCA (weighted clustering of called aCGH data), a method for weighted clustering of samples on the basis of the ordinal aCGH data. Two similarities to be used in the clustering and particularly suited for ordinal data are proposed, which are generalized to deal with weighted observations. In addition, a new form of linkage, especially suited for ordinal data, is introduced. In a simulation study, we show that the proposed cluster method is competitive to clustering using the continuous data. We illustrate WECCA using an application to a breast cancer data set, where WECCA finds a clustering that relates better with survival than the original one.     

Regression as a method to predict copy numbers in comparative genomic hybridization studies on bacteria

Comparative genomic hybridizations (CGH) using microarrays are performed with bacteria in order to determine the level of genomic similarity between various strains. The microarrays applied in CGH experiments are constructed on the basis of the genome sequence of one strain, which is used as a control, or reference, in each experiment. A strain being compared with the known strain is called the unknown strain. The ratios of fluorescent intensities obtained from the spots on the microarrays can be used to determine which genes are divergent in the unknown strain, as well as to predict the copy number of actual genes in the unknown strain. In this paper, we focus on the prediction of gene copy number based on data from CGH experiments. We assumed a linear connection between the log2 of the copy number and the observed log2-ratios, then predictors based on the factor analysis model and the linear random model were proposed in an attempt to identify the copy numbers. These predictors were compared to using the ratio of the intensities directly. Simulations indicated that the proposed predictors improved the prediction of the copy number in most situations. The predictors were applied on CGH data obtained from experiments with Enterococcus faecalis strains in order to determine copy number of relevant genes in five different strains.     

Estimation of Copy Number Alterations from Exome Sequencing Data

Exome sequencing constitutes an important technology for the study of human hereditary diseases and cancer. However, the ability of this approach to identify copy number alterations in primary tumor samples has not been fully addressed. Here we show that somatic copy number alterations can be reliably estimated using exome sequencing data through a strategy that we have termed exome2cnv. Using data from 86 paired normal and primary tumor samples, we identified losses and gains of complete chromosomes or large genomic regions, as well as smaller regions affecting a minimum of one gene. Comparison with high-resolution comparative genomic hybridization (CGH) arrays revealed a high sensitivity and a low number of false positives in the copy number estimation between both approaches. We explore the main factors affecting sensitivity and false positives with real data, and provide a side by side comparison with CGH arrays. Together, these results underscore the utility of exome sequencing to study cancer samples by allowing not only the identification of substitutions and indels, but also the accurate estimation of copy number alterations.     

Optimization of DOP-PCR amplification of DNA for high-resolution comparative genomic hybridization analysis

BACKGROUND: Whole-genome amplification of minute samples of DNA for the use in comparative genomic hybridization (CGH) analysis has found widespread use, but the method has not been well validated. METHODS: Four protocols for degenerate oligonucleotide primed polymerase chain reaction (DOP-PCR) and fluorescence labeling were applied to test DNA from normal and K-562 cells. The DNA products were used for CGH analysis. RESULTS: The DOP-PCR-amplified DNA from each protocol produced hybridizations with different qualities. These could be seen primarily as differences in background staining and signal-to-noise ratios, but also as characteristic deviations of normal/normal hybridizations. One DOP-PCR-protocol was further investigated. We observed concordance between CGH results using unamplified and DOP-PCR-amplified DNA. An example of an analysis of an invasive carcinoma of the breast supports the practical value of this approach. CONCLUSIONS: DOP-PCR-amplified DNA is applicable for high- resolution CGH, the results being similar to those of CGH using unamplified DNA.     

An algorithm to infer similarity among cell types and organisms by examining the most expressed sequences

Following sequence alignment, clustering algorithms are among the most utilized techniques in gene expression data analysis. Clustering gene expression patterns allows researchers to determine which gene expression patterns are alike and most likely to participate in the same biological process being investigated. Gene expression data also allow the clustering of whole samples of data, which makes it possible to find which samples are similar and, consequently, which sampled biological conditions are alike. Here, a novel similarity measure calculation and the resulting rank-based clustering algorithm are presented. The clustering was applied in 418 gene expression samples from 13 data series spanning three model organisms: Homo sapiens, Mus musculus, and Arabidopsis thaliana. The initial results are striking: more than 91% of the samples were clustered as expected. The MESs (most expressed sequences) approach outperformed some of the most used clustering algorithms applied to this kind of data such as hierarchical clustering and K-means. The clustering performance suggests that the new similarity measure is an alternative to the traditional correlation/distance measures typically used in clustering algorithms.     

Using Hamming Distance as Information for SNP-Sets Clustering and Testing in Disease Association Studies

The availability of high-throughput genomic data has led to several challenges in recent genetic association studies, including the large number of genetic variants that must be considered and the computational complexity in statistical analyses. Tackling these problems with a marker-set study such as SNP-set analysis can be an efficient solution. To construct SNP-sets, we first propose a clustering algorithm, which employs Hamming distance to measure the similarity between strings of SNP genotypes and evaluates whether the given SNPs or SNP-sets should be clustered. A dendrogram can then be constructed based on such distance measure, and the number of clusters can be determined. With the resulting SNP-sets, we next develop an association test HDAT to examine susceptibility to the disease of interest. This proposed test assesses, based on Hamming distance, whether the similarity between a diseased and a normal individual differs from the similarity between two individuals of the same disease status. In our proposed methodology, only genotype information is needed. No inference of haplotypes is required, and SNPs under consideration do not need to locate in nearby regions. The proposed clustering algorithm and association test are illustrated with applications and simulation studies. As compared with other existing methods, the clustering algorithm is faster and better at identifying sets containing SNPs exerting a similar effect. In addition, the simulation studies demonstrated that the proposed test works well for SNP-sets containing a large proportion of neutral SNPs. Furthermore, employing the clustering algorithm before testing a large set of data improves the knowledge in confining the genetic regions for susceptible genetic markers.     

From array to array: Confirmation of genomic gains and losses discovered by array-based comparative genomic hybridization utilizing fluorescence in situ hybridization on tissue microarrays

The combination of array-based comparative genomic hybridization (CGH) with fluorescence in situ hybridization utilizing custom-designed bacterial artificial chromosome (BAC) probes applied to tissue microarrays represents a powerful compendium of techniques–greatly enhancing the throughput of genomic analysis and subsequent target validation. Such approach can be automated at various levels and allows managing large volume of targets and samples in a few experiments. As such, this approach facilitates discovery, validation and implementation of findings in the process of identification of new diagnostic, prognostic and potentially therapeutic molecular markers.     

Haplotype-based linkage disequilibrium mapping via direct data mining

MOTIVATION: With the availability of large-scale, high-density single-nucleotide polymorphism markers and information on haplotype structures and frequencies, a great challenge is how to take advantage of haplotype information in the association mapping of complex diseases in case-control studies. RESULTS: We present a novel approach for association mapping based on directly mining haplotypes (i.e. phased genotype pairs) produced from case-control data or case-parent data via a density-based clustering algorithm, which can be applied to whole-genome screens as well as candidate-gene studies in small genomic regions. The method directly explores the sharing of haplotype segments in affected individuals that are rarely present in normal individuals. The measure of sharing between two haplotypes is defined by a new similarity metric that combines the length of the shared segments and the number of common alleles around any marker position of the haplotypes, which is robust against recent mutations/genotype errors and recombination events. The effectiveness of the approach is demonstrated by using both simulated datasets and real datasets. The results show that the algorithm is accurate for different population models and for different disease models, even for genes with small effects, and it outperforms some recently developed methods.     

Integrating multi-attribute similarity networks for robust representation of the protein space

MOTIVATION: A global view of the protein space is essential for functional and evolutionary analysis of proteins. In order to achieve this, a similarity network can be built using pairwise relationships among proteins. However, existing similarity networks employ a single similarity measure and therefore their utility depends highly on the quality of the selected measure. A more robust representation of the protein space can be realized if multiple sources of information are used. RESULTS: We propose a novel approach for analyzing multi-attribute similarity networks by combining random walks on graphs with Bayesian theory. A multi-attribute network is created by combining sequence and structure based similarity measures. For each attribute of the similarity network, one can compute a measure of affinity from a given protein to every other protein in the network using random walks. This process makes use of the implicit clustering information of the similarity network, and we show that it is superior to naive, local ranking methods. We then combine the computed affinities using a Bayesian framework. In particular, when we train a Bayesian model for automated classification of a novel protein, we achieve high classification accuracy and outperform single attribute networks. In addition, we demonstrate the effectiveness of our technique by comparison with a competing kernel-based information integration approach.     

Variation in gene content among geographically diverse Sulfolobus isolates

The ability of competitive (i.e., comparative) genomic hybridization (CGH) to assess similarity across entire microbial genomes suggests that it should reveal diversification within and between natural populations of free-living prokaryotes. We used CGH to measure relatedness of genomes drawn from Sulfolobus populations that had been shown in a previous study to be diversified along geographical lines. Eight isolates representing a wide range of spatial separation were compared with respect to gene-specific tags based on a closely related reference strain ( Sulfolobus solfataricus P2). For the purpose of assessing genetic divergence, 232 loci identified as polymorphic were assigned one of two alleles based on the corresponding fluorescence intensities from the arrays. Clustering of these binary genotypes was stable with respect to changes in the threshold and similarity criteria, and most of the groupings were consistent with an isolation-by-distance model of diversification. These results indicate that increasing spatial separation of geothermal sites correlates not only with minor sequence polymorphisms in conserved genes of Sulfolobus (demonstrated in the previous study), but also with the regions of difference (RDs) that occur between genomes of conspecifics. In view of the abundance of RDs in prokaryotic genomes and the relevance that some RDs may have for ecological adaptation, the results further suggest that CGH on microarrays may have advantages for investigating patterns of diversification in other free-living archaea and bacteria.     

A prognostic model for the combined analysis of gene expression profiling in hepatocellular carcinoma

Microarray techniques using cDNA array and comparative genomic hybridization (CGH) have been developed for several discovery applications. They are frequently applied for the prediction and diagnosis of cancer in recent years. Many studies have shown that integrating genomic data from different sources may increase the reliability of gene expression analysis results in understanding cancer progression. Therefore, developing a good prognostic model dealing simultaneously with different types of dataset is important. The challenge with these types of data is high background noise. We describe an analytical two-stage framework with a multi-parallel data analysis method named wavelet-based generalized singular value decomposition and shaving method (WGSVD-shaving). This method is proposed for de-noising and dimension-reduction during early stage prognosis modeling. We also applied a supervised gene clustering technique with penalized logistic regression with Cox-model on an integrated data. We show the accuracy of the method using a simulated dataset with a case study on Hepatocelluar Carcinoma (HCC) cDNA and CGH data. The method shows improved results from GSVD-shaving and has application in the discovery of candidate genes associated with cancer.     

Identifying patterns of DNA for tumor diagnosis using capillary electrophoresis-amplified fragment length polymorphism (CE-AFLP) screening

Amplified Fragment Length Polymorphism (AFLP) screening is a genome-wide genotyping strategy that has been widely used in plants and bacteria, but little has been reported concerning its use in humans. We investigated if the AFLP procedure could be coupled with high-throughput capillary electrophoresis (CE) for use in tumor diagnosis and classification. Using CE-AFLP, a series of molecular 'fingerprints' were generated for a set of gastric tumor and normal genomic DNA samples. The CE-AFLP procedure was qualitatively and quantitatively robust, and a variety of clustering tools were used to identify a specific DNA marker 'pattern' of 20 features that classified the tumor and normal samples to reasonable degrees of accuracy (Sensitivity 95%, Specificity 80%). The CE-AFLP-based approach also correctly classified 16 tumor samples, which in a previous study had exhibited no detectable genomic aberrations by comparative genome hybridization (CGH). This is the first reported application of CE-AFLP screening in tumor diagnosis. As the procedure is relatively inexpensive and requires minimal prior sequence knowledge and biological material, we suggest that CE-AFLP-based protocols may represent a promising new approach for DNA-based cancer screening and diagnosis.     

Modeling and Visualizing Uncertainty in Gene Expression Clusters Using Dirichlet Process Mixtures

Although the use of clustering methods has rapidly become one of the standard computational approaches in the literature of microarray gene expression data, little attention has been paid to uncertainty in the results obtained. Dirichlet process mixture (DPM) models provide a nonparametric Bayesian alternative to the bootstrap approach to modeling uncertainty in gene expression clustering. Most previously published applications of Bayesian model-based clustering methods have been to short time series data. In this paper, we present a case study of the application of nonparametric Bayesian clustering methods to the clustering of high-dimensional nontime series gene expression data using full Gaussian covariances. We use the probability that two genes belong to the same cluster in a DPM model as a measure of the similarity of these gene expression profiles. Conversely, this probability can be used to define a dissimilarity measure, which, for the purposes of visualization, can be input to one of the standard linkage algorithms used for hierarchical clustering. Biologically plausible results are obtained from the Rosetta compendium of expression profiles which extend previously published cluster analyses of this data.     

Weighted relative entropy for alignment-free sequence comparison based on Markov model

In this paper, we introduce a probabilistic measure for computing the similarity between two biological sequences without alignment. The computation of the similarity measure is based on the Kullback-Leibler divergence of two constructed Markov models. We firstly validate the method on clustering nine chromosomes from three species. Secondly, we give the result of similarity search based on our new method. We lastly apply the measure to the construction of phylogenetic tree of 48 HEV genome sequences. Our results indicate that the weighted relative entropy is an efficient and powerful alignment-free measure for the analysis of sequences in the genomic scale.     

Structural unbalanced chromosome rearrangements resolved by comparative genomic hybridization.

The identification of unbalanced structural chromosome rearrangements using conventional cytogenetic techniques depends on recognition of the unknown material from its banding pattern. Even with optimally banded chromosomes, when large chromosome segments are involved, cytogeneticists may not always be able to determine the origin of extrachromosomal material and supernumerary chromosomes. We report here on the application of comparative genomic hybridization (CGH), a new molecular-cytogenetic assay capable of detecting chromosomal gains and losses, to six clinical samples suspected of harboring unbalanced structural chromosome abnormalities. CGH provided essential information on the nature of the unbalanced aberration investigated in five of the six samples. This approach has proved its ability to resolve complex karyotypes and to provide information when metaphase chromosomes are not available. In cases where metaphase chromosome spreads were available, confirmation of CGH results was easily obtained by fluorescence in situ hybridization (FISH) using specific probes. Thus the combined use of CGH and FISH provided an efficient method for resolving the origin of aberrant chromosomal material unidentified by conventional cytogenetic analysis.