CoChat
Sign Up
Cited by 108Epigenetics and DNA MethylationGene expression and cancer classificationGenomics and Phylogenetic Studies

The DOI matches its Crossref record; 32 of the 32 references it cites were found in a registry or an index.

Batch effect correction for genome-wide methylation data with Illumina Infinium platform

Open access
Sun, Zhifu, Chai, High Seng, Wu, Yanhong, White, Wendy M, Donkena, Krishna V, Klein, Christopher J, Garovic, Vesna D, Therneau, Terry M, Kocher, Jean-Pierre A·2011·BMC Medical Genomics
DOI 10.1186/1755-8794-4-84PMC PMC3265417
Open this paper in CoChat — free accountOpen PDF

Abstract

BACKGROUND: Genome-wide methylation profiling has led to more comprehensive insights into gene regulation mechanisms and potential therapeutic targets. Illumina Human Methylation BeadChip is one of the most commonly used genome-wide methylation platforms. Similar to other microarray experiments, methylation data is susceptible to various technical artifacts, particularly batch effects. To date, little attention has been given to issues related to normalization and batch effect correction for this kind of data. METHODS: We evaluated three common normalization approaches and investigated their performance in batch effect removal using three datasets with different degrees of batch effects generated from HumanMethylation27 platform: quantile normalization at average β value (QNβ); two step quantile normalization at probe signals implemented in "lumi" package of R (lumi); and quantile normalization of A and B signal separately (ABnorm). Subsequent Empirical Bayes (EB) batch adjustment was also evaluated. RESULTS: Each normalization could remove a portion of batch effects and their effectiveness differed depending on the severity of batch effects in a dataset. For the dataset with minor batch effects (Dataset 1), normalization alone appeared adequate and "lumi" showed the best performance. However, all methods left substantial batch effects intact in the datasets with obvious batch effects and further correction was necessary. Without any correction, 50 and 66 percent of CpGs were associated with batch effects in Dataset 2 and 3, respectively. After QNβ, lumi or ABnorm, the number of CpGs associated with batch effects were reduced to 24, 32, and 26 percent for Dataset 2; and 37, 46, and 35 percent for Dataset 3, respectively. Additional EB correction effectively removed such remaining non-biological effects. More importantly, the two-step procedure almost tripled the numbers of CpGs associated with the outcome of interest for the two datasets. CONCLUSION: Genome-wide methylation data from Infinium Methylation BeadChip can be susceptible to batch effects with profound impacts on downstream analyses and conclusions. Normalization can reduce part but not all batch effects. EB correction along with normalization is recommended for effective batch effect removal.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1186/1755-8794-4-84

Cites

32 of 32 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Dataset 1 before and after normalization and batch effect correction. A: PCA plot for all 93 samples using all 27,578 CpGs. Different colors are for different batches. Nine pairs of technical replicates are marked as R1 to R9. The samples on Chip12 (circled with dash line) tend to separate from other samples. B. Density plot of samples from Chip12 and Chip26 shows minor distribution biases between the two chips. C: PCA plot of the 24 samples from Chip22 and 26 using all 27,578 CpGs. Two samples with an across bar are technical replicates. D: Box plot of pair-wise CpG errors between 9 pairs of technical replicates for unnormalized average β (red), QNβ (green), lumi (blue), and ABnorm (cyan). The unnormalized data has wider interquartile ranges and shifted medians from zero line. All normalized data have condensed interquartile ranges with medians adjusted close to zero line. E: Error means (lower pane) and average absolute deviations (upper panel) of 9 pairs of technical replicates before (red) and after three normalizations. Unnormalized data has the largest average absolute deviation for each of replicate pairs and shifted mean for most of the pairs. All normalized data show reduced average absolute deviations. F: Error means (lower pane) and average absolute deviations (upper panel) of 9 pairs of technical replicates before and after three normalizations plus EB correction. The normalized and EB correction data have almost identical error means and average absolute deviations compared to normalized data alone.
Figure 1 Dataset 1 before and after normalization and batch effect correction. A: PCA plot for all 93 samples using all 27,578 CpGs. Different colors are for different batches. Nine pairs of technical replicates are marked as R1 to R9. The samples on Chip12 (circled with dash line) tend to separate from other samples. B. Density plot of samples from Chip12 and Chip26 shows minor distribution biases between the two chips. C: PCA plot of the 24 samples from Chip22 and 26 using all 27,578 CpGs. Two samples with an across bar are technical replicates. D: Box plot of pair-wise CpG errors between 9 pairs of technical replicates for unnormalized average β (red), QNβ (green), lumi (blue), and ABnorm (cyan). The unnormalized data has wider interquartile ranges and shifted medians from zero line. All normalized data have condensed interquartile ranges with medians adjusted close to zero line. E: Error means (lower pane) and average absolute deviations (upper panel) of 9 pairs of technical replicates before (red) and after three normalizations. Unnormalized data has the largest average absolute deviation for each of replicate pairs and shifted mean for most of the pairs. All normalized data show reduced average absolute deviations. F: Error means (lower pane) and average absolute deviations (upper panel) of 9 pairs of technical replicates before and after three normalizations plus EB correction. The normalized and EB correction data have almost identical error means and average absolute deviations compared to normalized data alone.CC BY · PubMed Central
Figure 2 Dataset 2 before and after normalization and batch correction. A: The density plot of average β values for two chips of samples. All samples on Chip12 shift to the left, more CpGs with lower methylation values. B: PCA plot (first two components) for the 24 samples using 26,486 CpGs after excluding CpGs in sex chromosomes; samples on each chip cluster closely with the first component explaining 50.3% of variance. C: Density plot of lumi normalized data. The distribution bias has been greatly reduced but a significant portion remains. D: PCA plot of lumi normalized data still shows clear sample separation by batches using 26,486 CpGs after excluding CpGs in sex chromosomes. E: Density plot of average β after ABnorm. The distribution bias has been successfully removed. F: PCA plot of ABnorm data shows the clear remaining batch effects using 26,486 CpGs. G: The profiles of selected 20 CpGs that are associated with the batch effects after normalization. X-axis-samples ordered by Chip (11 or 12). Y-axis-methylation average β. Each line represents one CpG across samples. These CpGs are either all higher or lower on one chip than another. H: The profiles of the same 20 CpGs as G after normalization and EB correction. The systematic differences between the two chips have been removed.
Figure 2 Dataset 2 before and after normalization and batch correction. A: The density plot of average β values for two chips of samples. All samples on Chip12 shift to the left, more CpGs with lower methylation values. B: PCA plot (first two components) for the 24 samples using 26,486 CpGs after excluding CpGs in sex chromosomes; samples on each chip cluster closely with the first component explaining 50.3% of variance. C: Density plot of lumi normalized data. The distribution bias has been greatly reduced but a significant portion remains. D: PCA plot of lumi normalized data still shows clear sample separation by batches using 26,486 CpGs after excluding CpGs in sex chromosomes. E: Density plot of average β after ABnorm. The distribution bias has been successfully removed. F: PCA plot of ABnorm data shows the clear remaining batch effects using 26,486 CpGs. G: The profiles of selected 20 CpGs that are associated with the batch effects after normalization. X-axis-samples ordered by Chip (11 or 12). Y-axis-methylation average β. Each line represents one CpG across samples. These CpGs are either all higher or lower on one chip than another. H: The profiles of the same 20 CpGs as G after normalization and EB correction. The systematic differences between the two chips have been removed.CC BY · PubMed Central
Figure 3 Dataset 3 before and after normalization and batch effect correction. A: Box plot of raw average β for two chips of 24 samples. The medians between the two chips are similar, but the 3rd quartile values of Chip36 are much lower than Chip54. B: Box plot for the two chips of 24 samples after lumi normalization. C: The density plot of average β for 24 samples before normalization colored by batches. The distribution differs obviously between the chips. D: The density plot of "lumi" normalized average β for 24 samples shows a large portion of batch effects not corrected. E: unsupervised clustering using all 27,578 CpGs before normalization and EB correction shows the clear separation of samples by chips (Chip54 or Chip36); samples from the same tissue type tend to cluster within the same batch (* for normal prostates and others for tumors). F: unsupervised clustering after normalization and EB correction using all 27,578 CpGs shows the separation between batches is removed; samples from the same tissue type cluster closely (* for normal prostates and others for tumors). G: Methylation profiles of selected 20 CpGs that are significantly associated with the batch effects after normalization, showing the dramatic differences between the two chips. The letter "T" and "N" on x-axis represent tumor and normal sample. H: The methylation profiles of the same 20 CpGs as G after the additional EB correction. The systematic biases are successfully removed.
Figure 3 Dataset 3 before and after normalization and batch effect correction. A: Box plot of raw average β for two chips of 24 samples. The medians between the two chips are similar, but the 3rd quartile values of Chip36 are much lower than Chip54. B: Box plot for the two chips of 24 samples after lumi normalization. C: The density plot of average β for 24 samples before normalization colored by batches. The distribution differs obviously between the chips. D: The density plot of "lumi" normalized average β for 24 samples shows a large portion of batch effects not corrected. E: unsupervised clustering using all 27,578 CpGs before normalization and EB correction shows the clear separation of samples by chips (Chip54 or Chip36); samples from the same tissue type tend to cluster within the same batch (* for normal prostates and others for tumors). F: unsupervised clustering after normalization and EB correction using all 27,578 CpGs shows the separation between batches is removed; samples from the same tissue type cluster closely (* for normal prostates and others for tumors). G: Methylation profiles of selected 20 CpGs that are significantly associated with the batch effects after normalization, showing the dramatic differences between the two chips. The letter "T" and "N" on x-axis represent tumor and normal sample. H: The methylation profiles of the same 20 CpGs as G after the additional EB correction. The systematic biases are successfully removed.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 32 references

Cited by

Sorted by citations

Items 101–107 of 107

Related

See all

Metadata

Authors
Sun, Z., Chai, H. S., Wu, Y., White, W. M., Donkena, K. V., Klein, C. J., Garovic, V. D., Therneau, T. M., & Kocher, J.-P. A.
Affiliations
Mayo Clinic (US)
Corresponding author
Kocher, Jean-Pierre A
ORCID
Venue
BMC Medical Genomics
Publisher
Springer Science and Business Media LLC
Published
2011-12
Volume
4
Issue
1
Pages
84-84
ISSN
1755-8794 (ISSN-L)
DOI
10.1186/1755-8794-4-84
PMC
PMC3265417
OpenAlex
W2108947463
Cited by
108
References
32
Citations by year
2026: 1 · 2025: 2 · 2024: 4 · 2023: 1 · 2022: 6 · 2021: 4 · 2020: 3 · 2019: 9 · 2018: 9 · 2017: 6 · 2016: 13 · 2015: 13 · 2014: 11 · 2013: 17 · 2012: 9
FWCI
4.81
Type
journal article
Language
en
Open access
Licence
Locations
  • BMC Medical Genomics (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Mayo Clinic
  • Mayo Foundation for Medical Education and Research
  • Center for Individualized Medicine, Mayo Clinic
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Epigenetics and DNA Methylation
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
Keywords
batch effect correction, DNA methylation, genome-wide methylation profiling, quantile normalization, data normalization, CpG methylation
MeSH terms
Adult, Humans, Male, Reproducibility of Results, Genome, Human, CpG Islands, DNA Methylation, Oligonucleotide Array Sequence Analysis, Databases, Genetic

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review