Multi-group Analysis of Compositions of Microbiomes with Covariate Adjustments and Repeated Measures

Abstract

Microbiome differential abundance analysis methods for a pair of groups are well established in the literature. However, many microbiome studies involve multiple groups, sometimes even ordered groups, such as stages of a disease, and require different types of comparisons. Standard pairwise comparisons are not only inefficient in terms of power and false discovery rates, but they may not address the scientific question of interest. In this paper, we propose a general framework for performing a wide range of multi-group analyses with covariate adjustments and repeated measures. We demonstrate the effectiveness of our methodology through two real data sets. The first example explores the effects of aridity on the soil microbiome, and the second example investigates the effects of surgical interventions on the microbiome of IBD patients.

Gut microbes contribute to variation in foraging intensity in the honey bee,Apis mellifera

“…Microbiome beta diversity was analyzed using Permutation MANOVAs with 999 permutations ("adonis2", vegan package [78]) on clrtransformed ("transform", microbiome package [79]) raw read counts, followed by Pairwise Permutation MANOVAs with 999 permutations and FDR p-value adjustment ("pairwise.adonis", pairwiseAdonis package [80]), and visualized using Principal Components Analysis ("ordinate", phyloseq package [81]) with Aitchison distance ("distance", phyloseq package [81]). The relative abundance of individual microbes was analyzed using raw read counts and ANCOM-BC with FDR adjustment ("ancombc2", ANCOMBC package [62,65]) with task as a fixed effect and colony as random effect. Relative abundances were visualized using stacked barplots ("ggplot", ggplot2 package [82]).…”

Critical review of 16S rRNA gene sequencing workflow in microbiome studies: From primer selection to advanced data analysis

“…from Cappellato et al (2022),Fernandes et al (2014),Hugerth and Andersson (2017), Lin and Peddada (2020a), Lin and Peddada (2020b),Love et al (2014),Mallick et al (2021),Mandal et al (2015),Nearing et al (2022),Peddada and Lin (2023),Robinson et al (2010),Segata et al (2011),Weiss et al (2017),White et al (2009),Yang and Chen (2022) andZhou et al (2022).…”

“…More recently, numerous statistical methods have emerged that rely on the principles of CoDA. Among these, we highlight microbiome multivariable associations with linear models 2 (MaAsLin2) (Mallick et al., 2021), analysis of compositions of microbiomes (ANCOM) (Mandal et al., 2015), ANCOM with bias correction (ANCOM‐BC) (Lin & Peddada, 2020a), ANCOM‐BC2 (Peddada & Lin, 2023), ANOVA‐like differential expression 2 (ALDEx2) (Fernandes et al., 2014) and linear regression framework for differential abundance analysis (LinDA) (Zhou et al., 2022). As shown in Table 5, although these tools are diverse in the methods used to carry out differential abundance calculations, they all use the log‐ratio transformation approach.…”

“…In the literature, several papers have compared the performance of the tools described in our review for differential abundance analysis (Cappellato et al., 2022; Lin & Peddada, 2020a; Nearing et al., 2022; Peddada & Lin, 2023; Weiss et al., 2017). Specifically, these have focused primarily on assessing their power (i.e.…”

“…Among those that fall under the CoDA framework, although MaAsLin2 and ALDEx2 are effective in providing consistent results across studies (Nearing et al., 2022), their results concerning the control of FDRs are discrepant (Cappellato et al., 2022; Lin & Peddada, 2020b; Nearing et al., 2022). Lastly, ANCOM‐BC and LinDA have higher power than ANCOM‐BC2, but the latter performs better than the first two in controlling FDRs (Peddada & Lin, 2023). However, these conclusions should be taken cautiously due to the novel nature of many of these tools, which need to be evaluated in further independent studies.…”