fast.adonis: a computationally efficient non-parametric multivariate analysis of microbiome data for large-scale studies

Shilan Li; Emily Vogtmann; Barry I Graubard; Mitchell H Gail; Christian C Abnet; Jianxin Shi

doi:10.1093/bioadv/vbac044

fast.adonis: a computationally efficient non-parametric multivariate analysis of microbiome data for large-scale studies

Bioinform Adv. 2022 Jun 10;2(1):vbac044. doi: 10.1093/bioadv/vbac044. eCollection 2022.

Authors

Shilan Li^{1

2}, Emily Vogtmann¹, Barry I Graubard¹, Mitchell H Gail¹, Christian C Abnet¹, Jianxin Shi¹

Affiliations

¹ Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD 20892, USA.
² Department of Biostatistics, Bioinformatics and Biomathematics, Georgetown University, Washington, DC 20057, USA.

Abstract

Motivation: Nonparametric multivariate analysis has been widely used to identify variables associated with a dissimilarity matrix and to quantify their contribution. For very large studies ( $n \geq 5000$ ) and many explanatory variables, existing software packages (e.g. adonis and adonis2 in vegan) are computationally intensive when conducting sequential multivariate analysis with permutations or bootstrapping. Moreover, for subjects from a complex sampling design, we need to adjust for sampling weights to derive an unbiased estimate.

Results: We implemented an R function fast.adonis to overcome these computational challenges in large-scale studies. fast.adonis generates results consistent with adonis/adonis2 but much faster. For complex sampling studies, fast.adonis integrates sampling weights algebraically to mimic the source population; thus, analysis can be completed very fast without requiring a large amount of memory.

Availability and implementation: fast.adonis is implemented using R and is publicly available at https://github.com/jennylsl/fast.adonis.

Supplementary information: Supplementary data are available at Bioinformatics Advances online.

Published by Oxford University Press 2022.