Model-based clustering of DNA methylation array data: a recursive-partitioning algorithm for high-dimensional data arising as a mixture of beta distributions

Houseman, E. Andrés; Christensen, Brock C.; Yeh, Ru-Fang; Marsit, Carmen J.; Karagas, Margaret R.; Wrensch, Margaret; Nelson, Heather H.; Wiemels, Joseph L.; Zheng, Shichun; Wiencke, John K.; Kelsey, Karl T.

doi:10.1186/1471-2105-9-365

Cited by 184 publications

(216 citation statements)

References 27 publications

Supporting

Mentioning

212

Contrasting

Order By: Relevance

“…The RPMM algorithm was used because it is suitable for both b-distributions (methylation data) and Gaussian distributions (other data), and it gives an optimal number of clusters using a recursive-partitioning algorithm (24). The clustering gave three groups for the methylome (Me.1-.3), the transcriptome (T.1-.3), and the miRNome (Mi.1-.3) and four groups for the genome (G.1-.4; Fig.…”

Section: Resultsmentioning

confidence: 99%

“…Three methods were used to find unsupervised clusters within each omics: the recursively partitioned mixture model (RPMM) method (24), and two consensus clustering methods (25,26). There is a strong association between the three methods for the four omics (Fisher exact P values from 2.10EÀ04 to 2.49EÀ19; see Supplementary Material and Methods).…”

Section: Omic Analysismentioning

confidence: 99%

See 1 more Smart Citation

A Poor Prognosis Subtype of HNSCC Is Consistently Observed across Methylome, Transcriptome, and miRNome Analysis

Jung

Job

Ledrappier

et al. 2013

Clinical Cancer Research

View full text Add to dashboard Cite

Purpose: Distant metastasis after treatment is observed in about 20% of squamous cell carcinoma of the head and neck (HNSCC). In the absence of any validated robust biomarker, patients at higher risk for metastasis cannot be provided with tailored therapy. To identify prognostic HNSCC molecular subgroups and potential biomarkers, we have conducted genome-wide integrated analysis of four omic sets of data.Experimental Design: Using state-of-the-art technologies, a core set of 45 metastasizing and 55 nonmetastasizing human papillomavirus (HPV)-unrelated HNSCC patient samples were analyzed at four different levels: gene expression (transcriptome), DNA methylation (methylome), DNA copy number (genome), and microRNA (miRNA) expression (miRNome). Molecular subgroups were identified by a model-based clustering analysis. Their clinical relevance was evaluated by survival analysis, and functional significance by pathway enrichment analysis.Results: Patient subgroups selected by transcriptome, methylome, or miRNome integrated analysis are associated with shorter metastasis-free survival (MFS). A common subgroup, R1, selected by all three omic approaches, is statistically more significantly associated with MFS than any of the single omic-selected subgroups. R1 and non-R1 samples display similar DNA copy number landscapes, but more frequent chromosomal aberrations are observed in the R1 cluster (especially loss at 13q14.2-3). R1 tumors are characterized by alterations of pathways involved in cell-cell adhesion, extracellular matrix (ECM), epithelial-to-mesenchymal transition (EMT), immune response, and apoptosis.Conclusions: Integration of data across several omic profiles leads to better selection of patients at higher risk, identification of relevant molecular pathways of metastasis, and potential to discover biomarkers and drug targets. Clin Cancer Res; 19(15); 4174-84. Ó2013 AACR.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Omic Analysismentioning

confidence: 99%

A Poor Prognosis Subtype of HNSCC Is Consistently Observed across Methylome, Transcriptome, and miRNome Analysis

Jung

Job

Ledrappier

et al. 2013

Clinical Cancer Research

View full text Add to dashboard Cite

show abstract

“…For unsupervised class discovery within the four omics, three methods were used: the recursively partitioned mixture model (RPMM; ref. 17) and two consensus clustering methods (18,19). Only the results obtained with the third method were described in the article.…”

Section: Omics Analysismentioning

confidence: 99%

Molecular Subtypes of Clear Cell Renal Cell Carcinoma Are Associated with Sunitinib Response in the Metastatic Setting

Beuselinck

Job

Becht

et al. 2015

Clinical Cancer Research

249

201

View full text Add to dashboard Cite

Purpose: Selecting patients with metastatic clear-cell renal cell carcinoma (m-ccRCC) who might benefit from treatment with targeted tyrosine kinase inhibitors (TKI) is a challenge. Our aim was to identify molecular markers associated with outcome in patients with m-ccRCC treated with sunitinib.Experimental Design: We performed global transcriptome analyses on 53 primary resected ccRCC tumors from patients who developed metastatic disease and were treated with first-line sunitinib. We also determined chromosome copy-number aberrations, methylation status, and gene mutations in von Hippel-Lindau and PBRM1. Molecular data were analyzed in relation with response rate (RR), progression-free survival (PFS), and overall survival (OS). Validation was performed in 47 additional ccRCC samples treated in first-line metastatic setting with sunitinib.Results: Unsupervised transcriptome analysis identified 4 robust ccRCC subtypes (ccrcc1 to 4) related to previous molecular classifications that were associated with different responses to sunitinib treatment. ccrcc1/ccrcc4 tumors had a lower RR (P ¼ 0.005) and a shorter PFS and OS than ccrcc2/ccrcc3 tumors (P ¼ 0.001 and 0.0003, respectively). These subtypes were the only significant covariate in the multivariate Cox model for PFS and OS (P ¼ 0.017 and 0.006, respectively). ccrcc1/ccrcc4 tumors were characterized by a stem-cell polycomb signature and CpG hypermethylation, whereas ccrcc3 tumors, sensitive to sunitinib, did not exhibit cellular response to hypoxia. Moreover, ccrcc4 tumors exhibited sarcomatoid differentiation with a strong inflammatory, Th1-oriented but suppressive immune microenvironment, with high expression of PDCD1 (PD-1) and its ligands.Conclusions: ccRCC molecular subtypes are predictive of sunitinib response in metastatic patients, and could be used for personalized mRCC treatment with TKIs, demethylating or immunomodulatory drugs.

show abstract

“…106 Novel high-throughput nanopore sequencing variants, as well as diversity in technological platforms and in required sequencing depth, have also stimulated contributions to the recent literature. 107,108 Increasingly, sophisticated bioinformatic methods are required for downstream analyses as researchers attempt to interpret multi-locus methylation information from multiple samples, for example, methods such as model-based clustering described by Houseman et al, 109 tailored for data obtained with methylation-specific microarrays. Multivariate statistical methods, particularly for both supervised and unsupervised clustering, principal component analysis, regression and visualization tools, such as heatmaps, have proved vital to interpretation of outcomes for these complex data, 110 which are generated by combinations of epigenetic changes and molecular events.…”

Section: Targeted Analysis Methods and Toolsmentioning

confidence: 99%

Recent advances in computational epigenetics

Ruskin¹,

Barat²

2017

AGG

View full text Add to dashboard Cite

Model-based clustering of DNA methylation array data: a recursive-partitioning algorithm for high-dimensional data arising as a mixture of beta distributions

Cited by 184 publications

References 27 publications

A Poor Prognosis Subtype of HNSCC Is Consistently Observed across Methylome, Transcriptome, and miRNome Analysis

A Poor Prognosis Subtype of HNSCC Is Consistently Observed across Methylome, Transcriptome, and miRNome Analysis

Molecular Subtypes of Clear Cell Renal Cell Carcinoma Are Associated with Sunitinib Response in the Metastatic Setting

Recent advances in computational epigenetics

Contact Info

Product

Resources

About