Limits...
Autoregressive higher-order hidden Markov models: exploiting local chromosomal dependencies in the analysis of tumor expression profiles.

Seifert M, Abou-El-Ardat K, Friedrich B, Klink B, Deutsch A - PLoS ONE (2014)

Bottom Line: The performance benefit is attributed to the simultaneous usage of higher-order state-transitions in combination with autoregressive emissions.This benefit could not be reached by using each of these two features independently.An implementation is available under www.jstacs.de/index.php/ARHMM.

View Article: PubMed Central - PubMed

Affiliation: Center for Information Services and High Performance Computing, Dresden University of Technology, Dresden, Germany.

ABSTRACT
Changes in gene expression programs play a central role in cancer. Chromosomal aberrations such as deletions, duplications and translocations of DNA segments can lead to highly significant positive correlations of gene expression levels of neighboring genes. This should be utilized to improve the analysis of tumor expression profiles. Here, we develop a novel model class of autoregressive higher-order Hidden Markov Models (HMMs) that carefully exploit local data-dependent chromosomal dependencies to improve the identification of differentially expressed genes in tumor. Autoregressive higher-order HMMs overcome generally existing limitations of standard first-order HMMs in the modeling of dependencies between genes in close chromosomal proximity by the simultaneous usage of higher-order state-transitions and autoregressive emissions as novel model features. We apply autoregressive higher-order HMMs to the analysis of breast cancer and glioma gene expression data and perform in-depth model evaluation studies. We find that autoregressive higher-order HMMs clearly improve the identification of overexpressed genes with underlying gene copy number duplications in breast cancer in comparison to mixture models, standard first- and higher-order HMMs, and other related methods. The performance benefit is attributed to the simultaneous usage of higher-order state-transitions in combination with autoregressive emissions. This benefit could not be reached by using each of these two features independently. We also find that autoregressive higher-order HMMs are better able to identify differentially expressed genes in tumors independent of the underlying gene copy number status in comparison to the majority of related methods. This is further supported by the identification of well-known and of previously unreported hotspots of differential expression in glioblastomas demonstrating the efficacy of autoregressive higher-order HMMs for the analysis of individual tumor expression profiles. Moreover, we reveal interesting novel details of systematic alterations of gene expression levels in known cancer signaling pathways distinguishing oligodendrogliomas, astrocytomas and glioblastomas. An implementation is available under www.jstacs.de/index.php/ARHMM.

Show MeSH

Related in: MedlinePlus

Local chromosomal dependencies of gene expression levels in different types of cancer.Spatial correlations of expression levels of genes in increasing chromosomal order up to ten were quantified by an average autocorrelation function that considers each chromosome-specific expression profile in each individual tumor sample. The autocorrelation function quantifies the similarity of gene expression levels of neighboring genes on a chromosome in a fixed distance. Corresponding average autocorrelation functions are shown for three types of cancer (i) different types of gliomas (red) [33], (ii) breast cancer expression profiles (orange) [3] and (iii) glioblastoma expression profiles (grey) [4]. Additionally, the green curve represents the average autocorrelation function of normal brain reference gene expression profiles taken from [33]. Due to chromosomal aberrations in gliomas, expression levels of genes in close chromosomal proximity tend to show greater similarity in gliomas (red) than in corresponding normal brain tissues (green). Moreover, the black curve represents mean values and standard deviations of the average autocorrelation function for randomly permuted glioma gene expression profiles from [33] across 100 repeats. The observation of significant local chromosomal dependencies in tumor expression profiles compared to permuted expression profiles motivates the development of autoregressive higher-order HMMs for the analysis of tumor expression profiles.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4067306&req=5

pone-0100295-g001: Local chromosomal dependencies of gene expression levels in different types of cancer.Spatial correlations of expression levels of genes in increasing chromosomal order up to ten were quantified by an average autocorrelation function that considers each chromosome-specific expression profile in each individual tumor sample. The autocorrelation function quantifies the similarity of gene expression levels of neighboring genes on a chromosome in a fixed distance. Corresponding average autocorrelation functions are shown for three types of cancer (i) different types of gliomas (red) [33], (ii) breast cancer expression profiles (orange) [3] and (iii) glioblastoma expression profiles (grey) [4]. Additionally, the green curve represents the average autocorrelation function of normal brain reference gene expression profiles taken from [33]. Due to chromosomal aberrations in gliomas, expression levels of genes in close chromosomal proximity tend to show greater similarity in gliomas (red) than in corresponding normal brain tissues (green). Moreover, the black curve represents mean values and standard deviations of the average autocorrelation function for randomly permuted glioma gene expression profiles from [33] across 100 repeats. The observation of significant local chromosomal dependencies in tumor expression profiles compared to permuted expression profiles motivates the development of autoregressive higher-order HMMs for the analysis of tumor expression profiles.

Mentions: To overcome this, we develop a novel model class of autoregressive higher-order HMMs enabling an improved modeling of local dependencies between successive measurements. Autoregressive higher-order HMMs simultaneously utilize higher-order state-transitions in combination with autoregressive emissions as novel model features. Globally, this model class has very general modeling capabilities including mixture models, standard first-order HMMs and higher-order HMMs as special cases. We motivate the development of autoregressive higher-order HMMs by considering the analysis of individual tumor expression profiles in which local dependencies of gene expression levels are frequently caused by deletions and duplications of underlying chromosomal regions. The existence of such local chromosomal dependencies between expression levels of genes in close chromosomal proximity is clearly shown for three different types of cancer in Figure 1. Additionally, based on initial findings on the importance of integrating prior knowledge on the distribution of differentially expressed genes into the training of HMMs [13], we here also specifically design an efficient Bayesian Baum-Welch training for autoregressive higher-order HMMs.


Autoregressive higher-order hidden Markov models: exploiting local chromosomal dependencies in the analysis of tumor expression profiles.

Seifert M, Abou-El-Ardat K, Friedrich B, Klink B, Deutsch A - PLoS ONE (2014)

Local chromosomal dependencies of gene expression levels in different types of cancer.Spatial correlations of expression levels of genes in increasing chromosomal order up to ten were quantified by an average autocorrelation function that considers each chromosome-specific expression profile in each individual tumor sample. The autocorrelation function quantifies the similarity of gene expression levels of neighboring genes on a chromosome in a fixed distance. Corresponding average autocorrelation functions are shown for three types of cancer (i) different types of gliomas (red) [33], (ii) breast cancer expression profiles (orange) [3] and (iii) glioblastoma expression profiles (grey) [4]. Additionally, the green curve represents the average autocorrelation function of normal brain reference gene expression profiles taken from [33]. Due to chromosomal aberrations in gliomas, expression levels of genes in close chromosomal proximity tend to show greater similarity in gliomas (red) than in corresponding normal brain tissues (green). Moreover, the black curve represents mean values and standard deviations of the average autocorrelation function for randomly permuted glioma gene expression profiles from [33] across 100 repeats. The observation of significant local chromosomal dependencies in tumor expression profiles compared to permuted expression profiles motivates the development of autoregressive higher-order HMMs for the analysis of tumor expression profiles.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4067306&req=5

pone-0100295-g001: Local chromosomal dependencies of gene expression levels in different types of cancer.Spatial correlations of expression levels of genes in increasing chromosomal order up to ten were quantified by an average autocorrelation function that considers each chromosome-specific expression profile in each individual tumor sample. The autocorrelation function quantifies the similarity of gene expression levels of neighboring genes on a chromosome in a fixed distance. Corresponding average autocorrelation functions are shown for three types of cancer (i) different types of gliomas (red) [33], (ii) breast cancer expression profiles (orange) [3] and (iii) glioblastoma expression profiles (grey) [4]. Additionally, the green curve represents the average autocorrelation function of normal brain reference gene expression profiles taken from [33]. Due to chromosomal aberrations in gliomas, expression levels of genes in close chromosomal proximity tend to show greater similarity in gliomas (red) than in corresponding normal brain tissues (green). Moreover, the black curve represents mean values and standard deviations of the average autocorrelation function for randomly permuted glioma gene expression profiles from [33] across 100 repeats. The observation of significant local chromosomal dependencies in tumor expression profiles compared to permuted expression profiles motivates the development of autoregressive higher-order HMMs for the analysis of tumor expression profiles.
Mentions: To overcome this, we develop a novel model class of autoregressive higher-order HMMs enabling an improved modeling of local dependencies between successive measurements. Autoregressive higher-order HMMs simultaneously utilize higher-order state-transitions in combination with autoregressive emissions as novel model features. Globally, this model class has very general modeling capabilities including mixture models, standard first-order HMMs and higher-order HMMs as special cases. We motivate the development of autoregressive higher-order HMMs by considering the analysis of individual tumor expression profiles in which local dependencies of gene expression levels are frequently caused by deletions and duplications of underlying chromosomal regions. The existence of such local chromosomal dependencies between expression levels of genes in close chromosomal proximity is clearly shown for three different types of cancer in Figure 1. Additionally, based on initial findings on the importance of integrating prior knowledge on the distribution of differentially expressed genes into the training of HMMs [13], we here also specifically design an efficient Bayesian Baum-Welch training for autoregressive higher-order HMMs.

Bottom Line: The performance benefit is attributed to the simultaneous usage of higher-order state-transitions in combination with autoregressive emissions.This benefit could not be reached by using each of these two features independently.An implementation is available under www.jstacs.de/index.php/ARHMM.

View Article: PubMed Central - PubMed

Affiliation: Center for Information Services and High Performance Computing, Dresden University of Technology, Dresden, Germany.

ABSTRACT
Changes in gene expression programs play a central role in cancer. Chromosomal aberrations such as deletions, duplications and translocations of DNA segments can lead to highly significant positive correlations of gene expression levels of neighboring genes. This should be utilized to improve the analysis of tumor expression profiles. Here, we develop a novel model class of autoregressive higher-order Hidden Markov Models (HMMs) that carefully exploit local data-dependent chromosomal dependencies to improve the identification of differentially expressed genes in tumor. Autoregressive higher-order HMMs overcome generally existing limitations of standard first-order HMMs in the modeling of dependencies between genes in close chromosomal proximity by the simultaneous usage of higher-order state-transitions and autoregressive emissions as novel model features. We apply autoregressive higher-order HMMs to the analysis of breast cancer and glioma gene expression data and perform in-depth model evaluation studies. We find that autoregressive higher-order HMMs clearly improve the identification of overexpressed genes with underlying gene copy number duplications in breast cancer in comparison to mixture models, standard first- and higher-order HMMs, and other related methods. The performance benefit is attributed to the simultaneous usage of higher-order state-transitions in combination with autoregressive emissions. This benefit could not be reached by using each of these two features independently. We also find that autoregressive higher-order HMMs are better able to identify differentially expressed genes in tumors independent of the underlying gene copy number status in comparison to the majority of related methods. This is further supported by the identification of well-known and of previously unreported hotspots of differential expression in glioblastomas demonstrating the efficacy of autoregressive higher-order HMMs for the analysis of individual tumor expression profiles. Moreover, we reveal interesting novel details of systematic alterations of gene expression levels in known cancer signaling pathways distinguishing oligodendrogliomas, astrocytomas and glioblastomas. An implementation is available under www.jstacs.de/index.php/ARHMM.

Show MeSH
Related in: MedlinePlus