RT Journal Article
SR Electronic
T1 SIAMCAT: user-friendly and versatile machine learning workflows for statistically rigorous microbiome analyses
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 2020.02.06.931808
DO 10.1101/2020.02.06.931808
A1 Jakob Wirbel
A1 Konrad Zych
A1 Morgan Essex
A1 Nicolai Karcher
A1 Ece Kartal
A1 Guillem Salazar
A1 Peer Bork
A1 Shinichi Sunagawa
A1 Georg Zeller
YR 2020
UL http://biorxiv.org/content/early/2020/02/06/2020.02.06.931808.abstract
AB The human microbiome is increasingly mined for diagnostic and therapeutic biomarkers. However, computational tools tailored to such analyses are still scarce. Here, we present the SIAMCAT R package, a versatile and user-friendly toolbox for comparative metagenome analyses using machine learning (ML), statistical tests, and visualization. Based on a large meta-analysis of gut microbiome studies, we optimized the choice of ML algorithms and preprocessing routines for default workflow settings. Furthermore, we illustrate common pitfalls leading to overfitting and show how SIAMCAT safeguards against these to make statistically rigorous ML workflows broadly accessible. SIAMCAT is available from siamcat.embl.de and Bioconductor.