Scalable fold-change estimation from microbial metagenomics
Teichman, S.; Willis, A. D.
Show abstract
Microbial sequencing data are subject to sample-specific and species-specific distortions that complicate inference about true microbial abundances. Recent work established assumptions under which relative fold-differences in microbial abundances can be estimated from sequencing data, accompanied by an estimation and testing method called radEmu. While radEmu has excellent error rates, its computational cost can be prohibitive for large-scale analyses. Here we introduce fastEmu, a scalable approximation to radEmu's testing methodology. fastEmu uses a small, data-driven subset of taxa to approximate the computationally expensive test statistic. In a metagenomic analysis, we show a reduction in per-taxon runtime from 4-8 hours to 0.5-45 seconds, and simulations illustrate no adverse impact on estimation error or Type I error rate control. fastEmu enables testing differential abundance for thousands of microbial taxa while retaining clear parameter interpretation and error rate control.
Matching journals
The top 6 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Species-level microbial sequence classification is improved by source-environment information 97%
- Genomic GC bias correction improves species abundance estimation from metagenomic data 95%
- Systematic evaluation of metatranscriptomic differential gene expression in silico, in vitro, and in vivo enables elucidation of inter-species cross-feeding 95%
Similar papers in this journal
- Functional Analysis of Metagenomes by Likelihood Inference (FAMLI) Successfully Compensates for Multi-Mapping Short Reads from Metagenomic Samples 96%
- Feature selection and causal analysis for microbiome studies in the presence of confounding using standardization 93%
- A negative binomial latent factor model for paired microbiome sequencing data 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.