Date | October 7th, 2019 |
Seminar Room | Weniger Hall 149 |
Refreshments Room | Weniger Hall 245 (Statistics Conference room) |
Tea and Refreshments with Faculty and Speaker | 3:15pm to 3:45 pm |
Seminar Start | 4:00pm to 4:50pm |
Cost | Free and open to the public |
Bias in compositional data and implications for microbiome science
High-throughput sequencing has advanced our understanding of the role that bacteria and archaea play in marine, terrestrial and host-associated health. However, high throughput sequencing methods distort the composition of bacterial communities. The analysis of appropriate control data clearly illustrates that observed relative abundances are biased estimates of true relative abundances, with certain taxa consistently overobserved or underobserved compared to their true abundance. We propose a statistical model for bias in compositional data, and methods to estimate true community composition using calibration controls. Compositional bias has serious implications for the reproducibility of microbiome science, which we illustrate using data from the Microbiome Quality Control Project. We conclude with recommendations for the analysis and experimental design of studies involving compositional data.
For more information about Amy Willis, click here.