%0 Journal Article
%T Microbiome networks and change-point analysis reveal key community changes associated with cystic fibrosis pulmonary exacerbations
%A Bryan Coburn
%A D. Elizabeth Tullis
%A David M. Hwang
%A David S. Guttman
%A Hannah Li
%A Julio Diaz Caballero
%A Mehdi Layeghifard
%A Pauline W. Wang
%A Shawn T. Clark
%A Sylva L. Donaldson
%A Valerie Waters
%A Yu Zhang
%A Yvonne C. W. Yau
%J Archive of "NPJ Biofilms and Microbiomes".
%D 2019
%R 10.1038/s41522-018-0077-y
%X Association between key taxa and clinical states of CF patients. Key taxa were identified using abundance-based, prevalence-based, and hub-based strategies as described in the text. a A representative network of a single patient¡¯s microbiome with hubs representing taxa at the genus level and edges representing co-occurrence. The top ten key taxa detected by the three strategies are highlighted in the network using colored pie charts with yellow, green, and blue identifying taxa as top ten via abundance-based, prevalence-based, or hub-based methods, respectively. The names of the top 10 key taxa identified by any of the three strategy are shown in the key on the bottom. b Changes in the relative abundance of well-known taxa over time compared to changes in the clinical state of a single patient. The solid red lines represent the samples collected over time that do not significantly change in their abundances. Breaks in the red lines indicate changes in abundance identified by the change-point detection algorithm. c Cumulative cross-covariance measures of association between key taxa and clinical states of all CF patients in the discovery data set (longitudinal microbiome data from 18 CF patients). The boxes are interquartile range boxes representing the middle 50% of the data values. The line in the middle of each box represents the median (i.e. middle quartile). The whiskers extending from either side of each box represent the ranges for the top and bottom 25% of the data, excluding outliers which are shown as hollow points. The y-axis represents the cumulative sum of cross-covariances for the identified key taxa. The higher this measure, the stronger is the association between changes in the abundances of the top key taxa and changes in the clinical states. The x axis represents top one to top five ranked key taxa found by the four strategies, such that the ¡®Top n¡¯ category (where, 1£¿<£¿=£¿n£¿>£¿=£¿5) shows the sum of the cross-covariance measures of the n most important taxa. Thus, we expect an increase in the cumulative cross-covariance measures with increasing number of key taxa, n. We also identified a change-point standard for each patient consisted of the 10 taxa whose changes in abundance showed the strongest association to changes in clinical state. The Mann-Whitney-Wilcoxon test was used for statistical hypothesis testing (p£¿<£¿0.05; The asterisks represent the significant differences between the strategies). d Validation data set from Carmody et al.11 including longitudinal data from four patients is presented in the same manner as described in
%U https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6341074/