Preterm birth is the leading cause of infant mortality worldwide and bacterial infections are known to be its main cause, thus even a slight improvement in the statistical power for detection of patterns associated to the microbial population dynamics can make a considerable difference. We will show examples of how carefully modeling the noise in measurements of microbiota of pregnant women using a combination of oligotyping, variance stabilizing transformations, mixture models and finite state Markov chains can detect perturbations to the normal stability of the microbiome and pinpoint the relevant biomarkers for preterm birth. We use this as an example of pipelines for the analysis of heterogeneous data using the phyloseq package and interactive graphical explorations using D3.
This work involves data collected by D Relman and D. DiGiulio software developed by PJ McMurdie and statistical analyses done jointly with B Callahan.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.