Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index.

Batch effects removal for microbiome data via conditional quantile regression

Open access

Abstract

Batch effects in microbiome data arise from differential processing of specimens and can lead to spurious findings and obscure true signals. Strategies designed for genomic data to mitigate batch effects usually fail to address the zero-inflated and over-dispersed microbiome data. Most strategies tailored for microbiome data are restricted to association testing or specialized study designs, failing to allow other analytic goals or general designs. Here, we develop the Conditional Quantile Regression (ConQuR) approach to remove microbiome batch effects using a two-part quantile regression model. ConQuR is a comprehensive method that accommodates the complex distributions of microbial read counts by non-parametric modeling, and it generates batch-removed zero-inflated read counts that can be used in and benefit usual subsequent analyses. We apply ConQuR to simulated and real microbiome datasets and demonstrate its advantages in removing batch effects while preserving the signals of interest.

Overview

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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

Integrity signals

Registry

Crossref · 10.1038/s41467-022-33071-9

Cites

46 of 46 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Illustration of the ConQuR algorithm. Plots are based on real observations of Butyricimonas in the CARDIA study. a Two-step procedure. I. regression-step: (1) Use all available samples to fit the two-part quantile regression model; (2) For each sample, estimate the original likelihood of the taxon being present and the original distribution (by estimating a fine grid of percentiles) given the taxon is present. The two parts jointly determines the zero-inflated, over-dispersed conditional quantile function (the inverse of conditional distribution function) of the taxon count (Q^)^o. In the same manner, estimate the batch-free conditional quantile distribution (Q^)^c. II. Matching-step: locate the observed read count in (Q^)^o, and pick the value at the same location of (Q^)^c as the corrected read count. Repeat the procedure for each sample and then each taxon. b Three scenarios of matching. Left panel: Sample A has a less sparse and less outlying estimated batch-free distribution compared to the original one, so its observed measurement of zero is corrected to be a non-zero number. Middle panel: Sample B has a sparser and more outlying estimated batch-free distribution than the original one, so its observed non-zero count, located at a lower percentile of the original distribution, is corrected to be zero. Right panel: Sample C has a slightly less sparse and less outlying estimated batch-free distribution than the original one, so its observed non-zero count, located at a middle percentile of the original distribution, is corrected to be a smaller non-zero count.
Fig. 1 Illustration of the ConQuR algorithm. Plots are based on real observations of Butyricimonas in the CARDIA study. a Two-step procedure. I. regression-step: (1) Use all available samples to fit the two-part quantile regression model; (2) For each sample, estimate the original likelihood of the taxon being present and the original distribution (by estimating a fine grid of percentiles) given the taxon is present. The two parts jointly determines the zero-inflated, over-dispersed conditional quantile function (the inverse of conditional distribution function) of the taxon count (Q^)^o. In the same manner, estimate the batch-free conditional quantile distribution (Q^)^c. II. Matching-step: locate the observed read count in (Q^)^o, and pick the value at the same location of (Q^)^c as the corrected read count. Repeat the procedure for each sample and then each taxon. b Three scenarios of matching. Left panel: Sample A has a less sparse and less outlying estimated batch-free distribution compared to the original one, so its observed measurement of zero is corrected to be a non-zero number. Middle panel: Sample B has a sparser and more outlying estimated batch-free distribution than the original one, so its observed non-zero count, located at a lower percentile of the original distribution, is corrected to be zero. Right panel: Sample C has a slightly less sparse and less outlying estimated batch-free distribution than the original one, so its observed non-zero count, located at a middle percentile of the original distribution, is corrected to be a smaller non-zero count.CC BY · PubMed Central
Fig. 2 Evaluation on the simulated data. There are 6 simulation scenarios with 2 conditions and 2 batches, based on the starting data processed from the MOMS-PI study. Simulation scenarios are: A. Condition FC = 16, Batch FC = 1 (Null), B. Condition FC = 64, Batch FC = 4 (Condition Effect > Batch Effect), C. Condition FC = 4, Batch FC = 64 (Condition Effect < Batch Effect), where Condition and Batch are simulated from joint Bernoulli distribution with pCondition = 0.5, pBatch = 0.5, and OR = 1.25; Scenarios D, E, F are similar to Scenarios A, B, C, respectively, but pBatch=1/(1+exp(−libsize^s)), making batch variability incorporate library size variability. In the following plots, the scenarios are arranged on the x-axis with the order A, D, B, E, C, F because the two Nulls are allocated together, followed by Condition Effect > Batch Effect, and then Condition Effect < Batch Effect. Color and the name of the corresponding method are shown on the right within the graph. a Average proportions of data variability explained by Batch and Condition, quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarity. Lower batch variability with preserved or increased condition variability is preferred. b Average cross-validated area under the receiver operating characteristic curve (ROC-AUC) of predicting Condition from the taxa read counts via random forest. Higher ROC-AUC indicates a better prediction combining sensitivity and specificity. c The average false discovery rate (FDR, solid line) and sensitivity (dashed line) of association analysis between taxa relative abundance and Condition. Approaches with FDR attained around the nominal level 0.05 are valid, and among the valid approaches, higher sensitivity is preferred.
Fig. 2 Evaluation on the simulated data. There are 6 simulation scenarios with 2 conditions and 2 batches, based on the starting data processed from the MOMS-PI study. Simulation scenarios are: A. Condition FC = 16, Batch FC = 1 (Null), B. Condition FC = 64, Batch FC = 4 (Condition Effect > Batch Effect), C. Condition FC = 4, Batch FC = 64 (Condition Effect < Batch Effect), where Condition and Batch are simulated from joint Bernoulli distribution with pCondition = 0.5, pBatch = 0.5, and OR = 1.25; Scenarios D, E, F are similar to Scenarios A, B, C, respectively, but pBatch=1/(1+exp(−libsize^s)), making batch variability incorporate library size variability. In the following plots, the scenarios are arranged on the x-axis with the order A, D, B, E, C, F because the two Nulls are allocated together, followed by Condition Effect > Batch Effect, and then Condition Effect < Batch Effect. Color and the name of the corresponding method are shown on the right within the graph. a Average proportions of data variability explained by Batch and Condition, quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarity. Lower batch variability with preserved or increased condition variability is preferred. b Average cross-validated area under the receiver operating characteristic curve (ROC-AUC) of predicting Condition from the taxa read counts via random forest. Higher ROC-AUC indicates a better prediction combining sensitivity and specificity. c The average false discovery rate (FDR, solid line) and sensitivity (dashed line) of association analysis between taxa relative abundance and Condition. Approaches with FDR attained around the nominal level 0.05 are valid, and among the valid approaches, higher sensitivity is preferred.CC BY · PubMed Central
Fig. 3 Evaluation on the CARDIA data. a PCoA plots clustered by batch ID (corresponding colors are shown at the bottom within the graph), based on Bray-Curtis dissimilarity on raw count data (top panel), Aitchison dissimilarity on the corresponding relative abundance data (middle panel), and GUniFrac dissimilarity on the corresponding relative abundance data (bottom panel). Each point represents a sample and each ellipse represents a batch, with the centroid indicating the mean. As an ellipse connects the 95% percentile of points for each batch, the size of the ellipse indicates the dispersion, and the angle indicates higher-order features of the batch. Better alignment of the ellipses is preferred. b Proportions of data variability explained by batch ID and systolic blood pressure (SBP), quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarities. Lower variability explained by batch ID with preserved or increased variability explained by SBP is preferred. c Cross-validated root of mean squared error (RMSE) of predicting SBP based on the taxa read counts via random forest, where n = 5 folds of cross-validation. Lower values indicate stronger predictive signals of SBP in the microbial profiles. Definitions of the boxplot elements: the center line indicates median, the box limits are upper and lower quartiles, whiskers are the 1.5 interquartile range, and points beyond the whiskers are outliers.
Fig. 3 Evaluation on the CARDIA data. a PCoA plots clustered by batch ID (corresponding colors are shown at the bottom within the graph), based on Bray-Curtis dissimilarity on raw count data (top panel), Aitchison dissimilarity on the corresponding relative abundance data (middle panel), and GUniFrac dissimilarity on the corresponding relative abundance data (bottom panel). Each point represents a sample and each ellipse represents a batch, with the centroid indicating the mean. As an ellipse connects the 95% percentile of points for each batch, the size of the ellipse indicates the dispersion, and the angle indicates higher-order features of the batch. Better alignment of the ellipses is preferred. b Proportions of data variability explained by batch ID and systolic blood pressure (SBP), quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarities. Lower variability explained by batch ID with preserved or increased variability explained by SBP is preferred. c Cross-validated root of mean squared error (RMSE) of predicting SBP based on the taxa read counts via random forest, where n = 5 folds of cross-validation. Lower values indicate stronger predictive signals of SBP in the microbial profiles. Definitions of the boxplot elements: the center line indicates median, the box limits are upper and lower quartiles, whiskers are the 1.5 interquartile range, and points beyond the whiskers are outliers.CC BY · PubMed Central
Fig. 4 Evaluation on the HIVRC data. a PCoA plots clustered by study ID (corresponding colors are shown at the bottom within the graph), based on Bray-Curtis dissimilarity on raw count data (top panel) and Aitchison dissimilarity on the corresponding relative abundance data (bottom panel). Each point represents a sample and each ellipse represents a batch, with the centroid indicating the mean. As an ellipse connects the 95% percentile of points for each batch, the size of the ellipse indicates the dispersion, and the angle indicates higher-order features of the batch. Better alignment of the ellipses is preferred. b Proportions of data variability explained by study ID and HIV status, quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarities. Lower variability explained by study ID with preserved or increased variability explained by HIV status is preferred. c Cross-validated area under the receiver operating characteristic curve (ROC-AUC) of predicting HIV status based on the taxa read counts via random forest. Higher ROC-AUC indicates stronger predictive signal of HIV status in the microbial profiles.
Fig. 4 Evaluation on the HIVRC data. a PCoA plots clustered by study ID (corresponding colors are shown at the bottom within the graph), based on Bray-Curtis dissimilarity on raw count data (top panel) and Aitchison dissimilarity on the corresponding relative abundance data (bottom panel). Each point represents a sample and each ellipse represents a batch, with the centroid indicating the mean. As an ellipse connects the 95% percentile of points for each batch, the size of the ellipse indicates the dispersion, and the angle indicates higher-order features of the batch. Better alignment of the ellipses is preferred. b Proportions of data variability explained by study ID and HIV status, quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarities. Lower variability explained by study ID with preserved or increased variability explained by HIV status is preferred. c Cross-validated area under the receiver operating characteristic curve (ROC-AUC) of predicting HIV status based on the taxa read counts via random forest. Higher ROC-AUC indicates stronger predictive signal of HIV status in the microbial profiles.CC BY · PubMed Central
Fig. 5 Evaluation on the MOUTH data. a PCoA plots clustered by batch ID (corresponding colors are shown at the bottom within the graph), based on Bray-Curtis dissimilarity on raw count data (top panel), Aitchison dissimilarity on the corresponding relative abundance data (middle panel), and GUniFrac dissimilarity on the corresponding relative abundance data (bottom panel). Each point represents a sample, and each ellipse represents a batch with the centroid indicating the mean. As an ellipse connects the 95% percentile of points for each batch, the size of the ellipse indicates the dispersion, and the angle indicates higher-order features of the batch. Better alignment of the ellipses is preferred. b Proportions of data variability explained by batch ID and cigarette smoking (CIG) status, quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarities. Lower variability explained by batch ID with preserved or increased variability explained by CIG status is preferred. c Cross-validated cross-entropy of predicting CIG status based on the taxa read counts via random forest, where n = 5 folds of cross-validation. Lower values indicate stronger predictive signals of CIG status in the microbial profiles. Definitions of the boxplot elements: the center line indicates median, the box limits are upper and lower quartiles, whiskers are the 1.5 interquartile range, and points beyond the whiskers are outliers.
Fig. 5 Evaluation on the MOUTH data. a PCoA plots clustered by batch ID (corresponding colors are shown at the bottom within the graph), based on Bray-Curtis dissimilarity on raw count data (top panel), Aitchison dissimilarity on the corresponding relative abundance data (middle panel), and GUniFrac dissimilarity on the corresponding relative abundance data (bottom panel). Each point represents a sample, and each ellipse represents a batch with the centroid indicating the mean. As an ellipse connects the 95% percentile of points for each batch, the size of the ellipse indicates the dispersion, and the angle indicates higher-order features of the batch. Better alignment of the ellipses is preferred. b Proportions of data variability explained by batch ID and cigarette smoking (CIG) status, quantified by PERMANOVA R² in either Bray-Curtis or Aitchison dissimilarities. Lower variability explained by batch ID with preserved or increased variability explained by CIG status is preferred. c Cross-validated cross-entropy of predicting CIG status based on the taxa read counts via random forest, where n = 5 folds of cross-validation. Lower values indicate stronger predictive signals of CIG status in the microbial profiles. Definitions of the boxplot elements: the center line indicates median, the box limits are upper and lower quartiles, whiskers are the 1.5 interquartile range, and points beyond the whiskers are outliers.CC BY · PubMed Central

Verification

Claims the paper makes

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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 46 references

Research works printed but not linked: 1

  1. Zou, H. & Yuan, M. Composite quantile regression and the oracle model selection theory. Ann. Stat. 36, 1108–1126 (2008).

Cited by

See all 155

Metadata

Authors
Ling, W., Lu, J., Zhao, N., Lulla, A., Plantinga, A. M., Fu, W., Zhang, A., Liu, H., Song, H., Li, Z., Chen, J., Randolph, T. W., Koay, W. L. A., White, J. R., Launer, L. J., Fodor, A. A., Meyer, K. A., & Wu, M. C.
Affiliations
  • Fred Hutch Cancer Center (US)
  • Johns Hopkins University (US)
  • Williams College (US)
  • University of Washington (US)
  • University of Florida Health (US)
  • Mayo Clinic (US)
  • Children's National (US)
  • Resphera Biosciences (US)
  • National Institutes of Health (US)
  • University of North Carolina at Charlotte (US)
Corresponding author
  • Zhao, Ni
  • Wu, Michael C.
Venue
Nature Communications
Published
2022-09-15
Volume
13
Issue
1
Pages
5418-5418
ISSN
2041-1723 (ISSN-L)
OpenAlex
W4296029444
Cited by
136
References
46
Citations by year
2026: 43 · 2025: 39 · 2024: 41 · 2023: 12 · 2022: 1
FWCI
7.76
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Heart, Lung, and Blood Institute (R01HL155417, HHSN268201800004I, HHSN268201800005I, K01HL127159, and HHSN268201800007I, HHSN268201800006I, HHSN268201800007I, HHSN268201800003I)
  • National Institute of General Medical Sciences (R01GM123014, R01GM129512)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gut microbiota and health
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
Keywords
batch effects, microbiome data, conditional quantile regression, two-part model, zero-inflated count data, overdispersion, nonparametric models, batch effect correction
MeSH terms
Research Design, Microbiota

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Batch effects removal for microbiome data via conditional quantile regression. https://app.cochat.ai/papers/10.1038/s41467-022-33071-9
The paper
Ling, W., Lu, J., Zhao, N., Lulla, A., Plantinga, A. M., Fu, W., Zhang, A., Liu, H., Song, H., Li, Z., Chen, J., Randolph, T. W., Koay, W. L. A., White, J. R., Launer, L. J., Fodor, A. A., Meyer, K. A., & Wu, M. C. (2022). Batch effects removal for microbiome data via conditional quantile regression. Nature Communications. https://doi.org/10.1038/s41467-022-33071-9

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.