Batch effects correction for microbiome data with Dirichlet-multinomial regression

Abstract Microbial communities have been increasingly studied in recent years to investigate their role in ecological habitats. However, microbiome studies are difficult to reproduce or replicate as they may suffer from confounding factors that are unavoidable in practice and originate from biological, technical or computational sources. In this review, we define batch effects as unwanted variation introduced by confounding factors that are not related to any factors of interest. Computational and analytical methods are required to remove or account for batch effects. However, inherent microbiome data characteristics (e.g. sparse, compositional and multivariate) challenge the development and application of batch effect adjustment methods to either account or correct for batch effects. We present commonly encountered sources of batch effects that we illustrate in several case studies. We discuss the limitations of current methods, which often have assumptions that are not met due to the peculiarities of microbiome data. We provide practical guidelines for assessing the efficiency of the methods based on visual and numerical outputs and a thorough tutorial to reproduce the analyses conducted in this review.

Download Full-text

Erratum to: An integrative Bayesian Dirichlet-multinomial regression model for the analysis of taxonomic abundances in microbiome data

BMC Bioinformatics ◽

10.1186/s12859-017-1606-z ◽

2017 ◽

Vol 18 (1) ◽

Cited By ~ 2

Author(s):

W. Duncan Wadsworth ◽

Raffaele Argiento ◽

Michele Guindani ◽

Jessica Galloway-Pena ◽

Samuel A. Shelburne ◽

...

Keyword(s):

Regression Model ◽

Multinomial Regression ◽

Multinomial Regression Model ◽

Microbiome Data

Download Full-text

Cluster analysis of microbiome data by using mixtures of Dirichlet–multinomial regression models

Journal of the Royal Statistical Society Series C (Applied Statistics) ◽

10.1111/rssc.12432 ◽

2020 ◽

Vol 69 (5) ◽

pp. 1163-1187

Author(s):

Sanjeena Subedi ◽

Drew Neish ◽

Stephen Bak ◽

Zeny Feng

Keyword(s):

Cluster Analysis ◽

Regression Models ◽

Multinomial Regression ◽

Microbiome Data

Download Full-text

Batch effects removal for microbiome data via conditional quantile regression (ConQuR)

10.1101/2021.09.23.461592 ◽

2021 ◽

Author(s):

Wodan Ling ◽

Ni Zhao ◽

Anju Lulla ◽

Anna M. Plantinga ◽

Weijia Fu ◽

...

Keyword(s):

Quantile Regression ◽

Genomic Analysis ◽

Data Sets ◽

Batch Effects ◽

Differential Processing ◽

Conditional Quantile ◽

Association Testing ◽

Quantile Regression Model ◽

Comprehensive Method ◽

Microbiome Data

Batch effects in microbiome data arise from differential processing of specimens and can lead to spurious findings and obscure true signals. Most existing strategies for mitigating batch effects rely on approaches designed for genomic analysis, failing to address the zero-inflated and over-dispersed microbiome data. Strategies tailored for microbiome data are restricted to association testing, failing to allow other analytic goals such as visualization. We develop the Conditional Quantile Regression (ConQuR) approach to remove microbiome batch effects using a two-part quantile regression model. It is a fundamental advancement in the field because it is the first comprehensive method that accommodates the complex distributions of microbial read counts, and it generates batch-removed zero-inflated read counts that can be used in and benefit all usual subsequent analyses. We apply ConQuR to real microbiome data sets and demonstrate its state-of-the-art performance in removing batch effects while preserving or even amplifying the signals of interest.

Download Full-text

A multivariate method to correct for batch effects in microbiome data

10.1101/2020.10.27.358283 ◽

2020 ◽

Author(s):

Yiwen Wang ◽

Kim-Anh Lê Cao

Keyword(s):

Correction Method ◽

Expression Data ◽

Batch Effects ◽

Batch Correction ◽

Treatment Variation ◽

Corrected Data ◽

Component Estimation ◽

Microbiome Data ◽

Treatment Designs ◽

Latent Components

AbstractMicrobial communities are highly dynamic and sensitive to changes in the environment. Thus, microbiome data are highly susceptible to batch effects, defined as sources of unwanted variation that are not related to, and obscure any factors of interest. Existing batch correction methods have been primarily developed for gene expression data. As such, they do not consider the inherent characteristics of microbiome data, including zero inflation, overdispersion and correlation between variables. We introduce a new multivariate and non-parametric batch correction method based on Partial Least Squares Discriminant Analysis. PLSDA-batch first estimates treatment and batch variation with latent components to then subtract batch variation from the data. The resulting batch effect corrected data can then be input in any downstream statistical analysis. Two variants are also proposed to handle unbalanced batch x treatment designs and to include variable selection during component estimation. We compare our approaches with existing batch correction methods removeBatchEffect and ComBat on simulated and three case studies. We show that our three methods lead to competitive performance in removing batch variation while preserving treatment variation, and especially when batch effects have high variability. Reproducible code and vignettes are available on GitHub.

Download Full-text

Variable selection for sparse Dirichlet-multinomial regression with an application to microbiome data analysis

The Annals of Applied Statistics ◽

10.1214/12-aoas592 ◽

2013 ◽

Vol 7 (1) ◽

pp. 418-442 ◽

Cited By ~ 96

Author(s):

Jun Chen ◽

Hongzhe Li

Keyword(s):

Data Analysis ◽

Variable Selection ◽

Multinomial Regression ◽

Selection For ◽

Microbiome Data ◽

Microbiome Data Analysis

Download Full-text

Trajectories of eating behavior during COVID-19 lockdown: Longitudinal analyses of 22,374 adults in the UK

10.31234/osf.io/kmgzw ◽

2020 ◽

Author(s):

Moritz Herle ◽

Andrea Smith ◽

Feifei Bu ◽

Andrew Steptoe ◽

Daisy Fancourt

Keyword(s):

Eating Behavior ◽

Eating Behaviors ◽

Latent Class ◽

Stressful Life Events ◽

Panel Study ◽

Social Study ◽

Multinomial Regression ◽

Design Data ◽

Trajectories Of Change ◽

The Uk

Background: The COVID-19 pandemic has led to the implementation of stay-at-home and lockdown measures. It is currently unknown if the experience of lockdown leads to long term changes in individual’s eating behaviors.Objective: The objectives of this study were: i) to derive longitudinal trajectories of change in eating during UK lockdown, and ii) to identify risk factors associated with eating behavior trajectories. Design: Data from 22,374 UK adults from the UCL COVID-19 Social study (a panel study collecting weekly data during the pandemic) were analyzed from 28th March to 29th May 2020. Latent Class Growth Analysis was used to derive trajectories of change in eating. These were then associated with prior socio-economic, heath-related and psychological factors using multinomial regression models. Results: Analyses suggested five trajectories, with the majority (64%) showing no change in eating. In contrast, one trajectory was marked by persistently eating more, whereas another by persistently eating less. Overall, participants with greater depressive symptoms were more likely to report any change in eating. Loneliness was linked to persistently eating more (OR= 1.07), whereas being single or divorced, as well as stressful life events, were associated with consistently eating less (OR= 1.69). Overall, higher education status was linked to lower odds of changing eating behavior (OR= 0.54-0.77). Secondary exploratory analyses suggest that participants self-reported to have overweight were most common amongst the consistently overeaters, whereas underweight participants persistently ate less. Conclusion: In this study, we found that one third of the sample report changes in quantities eaten throughout the first UK lockdown period. Findings highlight the importance of adjusting public health programs to support eating behaviors in future lockdowns both in this and potential future pandemics. This is particularly important as part of on-going preventive efforts to prevent nutrition-related chronic diseases.

Download Full-text