Normalizing bead-based microRNA expression data: a measurement error model-based approach - PubMed (original) (raw)

Normalizing bead-based microRNA expression data: a measurement error model-based approach

Bin Wang et al. Bioinformatics. 2011.

Abstract

Motivation: Compared with complementary DNA (cDNA) or messenger RNA (mRNA) microarray data, microRNA (miRNA) microarray data are harder to normalize due to the facts that the total number of miRNAs is small, and that the majority of miRNAs usually have low expression levels. In bead-based microarrays, the hybridization is completed in several pools. As a result, the number of miRNAs tested in each pool is even smaller, which poses extra difficulty to intrasample normalization and ultimately affects the quality of the final profiles assembled from various pools. In this article, we consider a measurement error model-based method for bead-based microarray intrasample normalization.

Results: In this study, results from quantitative real-time PCR (qRT-PCR) assays are used as 'gold standards' for validation. The performance of the proposed measurement error model-based method is evaluated via a simulation study and real bead-based miRNA expression data. Simulation results show that the new method performs well to assemble complete profiles from subprofiles from various pools. Compared with two intrasample normalization methods recommended by the manufacturer, the proposed approach produces more robust final complete profiles and results in better agreement with the qRT-PCR results in identifying differentially expressed miRNAs, and hence improves the reproducibility between the two microarray platforms. Meaningful results are obtained by the proposed intrasample normalization method, together with quantile normalization as a subsequent complemental intersample normalization method.

Availability: Datasets and R package are available at http://gauss.usouthal.edu/publ/beadsme/.

PubMed Disclaimer

Figures

Fig. 1.

Fig. 1.

All results are based on the experimental data for one specimen. Plot (a) shows the original intensity distributions of the four samples (before normalization). Plot (b) shows the logarithmic intensity distributions of the four samples (before normalization). Each of plots (c) through (f) shows the logarithmic intensity distributions of the five subprofiles of the four samples.

Fig. 2.

Fig. 2.

Profiles of the four normalizers in the five pools (in columns) for the four samples (in rows) for 1 of the 10 specimens. Original scale nMFIs are shown in the _x_-axis.

Fig. 3.

Fig. 3.

Box-plots of the D's based on 10 000 repeats: left, nmean; middle, nmed; right, nme.

Fig. 4.

Fig. 4.

Weighted Kappa coefficient comparisons for different intrasample normalization methods, under various FC and RQ cutoffs, and with Qnorm as intersample normalization method.

Similar articles

Cited by

References

    1. Bolstad B.M., et al. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics. 2003;19:185–193. - PubMed
    1. Bruheim S., et al. Gene expression profiles classify human osteosarcoma xenografts according to sensitivity to doxorubicin, cisplatin, and ifosfamide. Clin. Cancer Res. 2009;15:7161–7169. - PubMed
    1. Cohen J. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 1960;20:37–46.
    1. Davison T.S., et al. Analyzing micro-RNA expression using microarrays. Methods Enzymol. 2006;411:14–34. - PubMed
    1. Dudoit S., et al. Statistical methods for identifying genes with differential expression in replicated cDNA microarray experiments. Stat. Sin. 2002;12:111–139.

Publication types

MeSH terms

Substances

LinkOut - more resources