A comparison of feature selection and classification methods in DNA methylation studies using the Illumina Infinium platform

BMC Bioinformatics, Apr 2012

Background The 27k Illumina Infinium Methylation Beadchip is a popular high-throughput technology that allows the methylation state of over 27,000 CpGs to be assayed. While feature selection and classification methods have been comprehensively explored in the context of gene expression data, relatively little is known as to how best to perform feature selection or classification in the context of Illumina Infinium methylation data. Given the rising importance of epigenomics in cancer and other complex genetic diseases, and in view of the upcoming epigenome wide association studies, it is critical to identify the statistical methods that offer improved inference in this novel context. Results Using a total of 7 large Illumina Infinium 27k Methylation data sets, encompassing over 1,000 samples from a wide range of tissues, we here provide an evaluation of popular feature selection, dimensional reduction and classification methods on DNA methylation data. Specifically, we evaluate the effects of variance filtering, supervised principal components (SPCA) and the choice of DNA methylation quantification measure on downstream statistical inference. We show that for relatively large sample sizes feature selection using test statistics is similar for M and β-values, but that in the limit of small sample sizes, M-values allow more reliable identification of true positives. We also show that the effect of variance filtering on feature selection is study-specific and dependent on the phenotype of interest and tissue type profiled. Specifically, we find that variance filtering improves the detection of true positives in studies with large effect sizes, but that it may lead to worse performance in studies with smaller yet significant effect sizes. In contrast, supervised principal components improves the statistical power, especially in studies with small effect sizes. We also demonstrate that classification using the Elastic Net and Support Vector Machine (SVM) clearly outperforms competing methods like LASSO and SPCA. Finally, in unsupervised modelling of cancer diagnosis, we find that non-negative matrix factorisation (NMF) clearly outperforms principal components analysis. Conclusions Our results highlight the importance of tailoring the feature selection and classification methodology to the sample size and biological context of the DNA methylation study. The Elastic Net emerges as a powerful classification algorithm for large-scale DNA methylation studies, while NMF does well in the unsupervised context. The insights presented here will be useful to any study embarking on large-scale DNA methylation profiling using Illumina Infinium beadarrays.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

http://www.biomedcentral.com/content/pdf/1471-2105-13-59.pdf

A comparison of feature selection and classification methods in DNA methylation studies using the Illumina Infinium platform

BMC Bioinformatics A comparison of feature selection and classification methods in DNA methylation studies using the Illumina Infinium platform Joanna Zhuang 0 1 2 Martin Widschwendter 1 Andrew E Teschendorff 0 2 0 Statistical Genomics Group, Paul O'Gorman Building, UCL Cancer Institute, University College London , 72 Huntley Street, London WC1E 6BT , UK 1 Department of Women's Cancer, UCL Elizabeth Garrett Anderson Institute for Women's Health, University College London , Room 340, 74 Huntley Street, London WC1E 6 AU , UK 2 Statistical Genomics Group, Paul O'Gorman Building, UCL Cancer Institute, University College London , 72 Huntley Street, London WC1E 6BT , UK Background: The 27k Illumina Infinium Methylation Beadchip is a popular high-throughput technology that allows the methylation state of over 27,000 CpGs to be assayed. While feature selection and classification methods have been comprehensively explored in the context of gene expression data, relatively little is known as to how best to perform feature selection or classification in the context of Illumina Infinium methylation data. Given the rising importance of epigenomics in cancer and other complex genetic diseases, and in view of the upcoming epigenome wide association studies, it is critical to identify the statistical methods that offer improved inference in this novel context. Results: Using a total of 7 large Illumina Infinium 27k Methylation data sets, encompassing over 1,000 samples from a wide range of tissues, we here provide an evaluation of popular feature selection, dimensional reduction and classification methods on DNA methylation data. Specifically, we evaluate the effects of variance filtering, supervised principal components (SPCA) and the choice of DNA methylation quantification measure on downstream statistical inference. We show that for relatively large sample sizes feature selection using test statistics is similar for M and b-values, but that in the limit of small sample sizes, M-values allow more reliable identification of true positives. We also show that the effect of variance filtering on feature selection is study-specific and dependent on the phenotype of interest and tissue type profiled. Specifically, we find that variance filtering improves the detection of true positives in studies with large effect sizes, but that it may lead to worse performance in studies with smaller yet significant effect sizes. In contrast, supervised principal components improves the statistical power, especially in studies with small effect sizes. We also demonstrate that classification using the Elastic Net and Support Vector Machine (SVM) clearly outperforms competing methods like LASSO and SPCA. Finally, in unsupervised modelling of cancer diagnosis, we find that non-negative matrix factorisation (NMF) clearly outperforms principal components analysis. Conclusions: Our results highlight the importance of tailoring the feature selection and classification methodology to the sample size and biological context of the DNA methylation study. The Elastic Net emerges as a powerful classification algorithm for large-scale DNA methylation studies, while NMF does well in the unsupervised context. The insights presented here will be useful to any study embarking on large-scale DNA methylation profiling using Illumina Infinium beadarrays. DNA methylation; Classification; Feature selection; Beadarrays - Background DNA methylation (DNAm) is one of the most important epigenetic mechanisms regulating gene expression, and aberrant DNAm has been implicated in the initiation and progression of human cancers [1,2]. DNAm changes have also been observed in normal tissue as a function of age [3-8], and age-associated DNAm markers have been proposed as early detection or cancer risk markers [3,6-8]. Proper statistical analysis of genome-wide DNA methylation profiles is therefore critically important for the discovery of novel DNAm based biomarkers. However, the nature of DNA methylation data presents novel statistical challenges and it is therefore unclear if popular statistical methods used in the gene expression community can be translated to the DNAm context [9]. The Illumina Infinium HumanMethylation27 BeadChip assay is a relatively recent high-throughput technology [10] that allows over 27,000 CpGs to be assayed. While a growing number of Infinium 27k data sets have been deposited in the public domain [3,4,11-15], relatively few studies have compared statistical analysis methods for this platform. In fact, most statistical reports on Infinium 27k DNAm data have focused on unsupervised clustering and normalisation methods [16-19], but as yet no study has performed a comprehensive comparison of feature selection and classification methods in this type of data. This is surprising given that feature selection and classification methods have been extensively explored in the context of gene expression data, see e.g. [20-33]. Moreover, feature selectio (...truncated)


This is a preview of a remote PDF: http://www.biomedcentral.com/content/pdf/1471-2105-13-59.pdf

Joanna Zhuang, Martin Widschwendter, Andrew E Teschendorff. A comparison of feature selection and classification methods in DNA methylation studies using the Illumina Infinium platform, BMC Bioinformatics, 2012, pp. 59, 13, DOI: 10.1186/1471-2105-13-59