Bioinformatics Seminars

Bioinformatics Seminar

Time: 11AM
Venue: Zoom Webinar

16 February 2021

Statistical and Computational Methods for Managing Batch Effects in Microbiome Data

Yiwen Wang
Melbourne Integrative Genomics

Microbial studies have been investigating the link between microbial composition and phenotypes, including human diseases. As microbial composition is highly dynamic and sensitive to small variations existing in the environment, microbiome data are highly susceptible to batch effects. Batch effects can be defined as any unwanted sources of variation obscuring the biological factors of interest. Most methods have been primarily proposed for differential abundance analysis, but they limit the breadth of statistical analysis that can be performed to answer biological questions. Another alternative is to use methods that correct for batch effects, however these methods were not developed to take into account the inherent characteristics of microbiome data.

In this talk, I will present different techniques to manage batch effects in microbiome data, starting from data characteristics, methods assumptions, detection and visualisation of batch effects. I will make the distinction between accounting for and correcting for batch effects. I will then introduce a multivariate method I recently developed for batch effect correction, based on PLS-Discriminant Analysis. Each step will be illustrated in one or several case studies.


Search past seminars