Meta-analysis is a statistical method that quantitatively synthesizes, by calculating a combined result, the results of independent studies addressing a specific research question. The principle is simple: pooling data from several studies increases statistical power. However, a number of conditions must be assessed to ensure that the combined result is not biased and that the conclusions drawn are accurate. A key step is to explore the sources of heterogeneity and look for possible biases. Advances in bioinformatics and next-generation sequencing have led to important advances in the understanding of the role of the microbiota in health. Knowledge development is often based on the conclusions that can be drawn from all published data, i.e. meta-analyses. Yet, in microbiota studies, differences between studies (in terms of pipelines, characteristics, sequencing techniques, samplecollection sites, study populations, etc.) can be very high. An exploration of the sources of heterogeneity is essential to determine whether studies, even if they address the same research question, are comparable.Multivariate data analysis methods (such as principal components analysis for quantitative data, multiple correspondence analysis for categorical data, and factor analysis of mixed data, a mixture of the two) as well as penalized meta-regression (such as the Lasso) are applied to explore heterogeneity in microbiota meta-analyses. Data from recently published microbiome meta-analyses were re-analyzed with the developed tools. In this work, we illustrate the utility of multivariate data analysis methods and penalized meta-regression in exploring sources of heterogeneity in microbiota meta-analyses.< Réduire