Taxonomic-free metagenome GWAS to identify gut microbiome functions influencing host phenotypes

Abstract

Genome-wide association studies (GWAS) have been pivotal for uncovering the genetics of human phenotypes, and there is now growing interest in applying GWAS-like methods to explore the role of the microbiome in human health. Here, we developed a taxonomy free GWAS approach that uses k-mers, i.e. DNA words of length k that can capture single nucleotide polymorphisms, insertion or deletion events, and gene presence-absence, to interrogate the gut microbiome. We applied this method to 26 traits spanning demographics, physiological measurements, health, and lifestyle in 938 healthy participants from the Milieu Interieur cohort. We generated a k-mer abundance matrix encompassing 97 million distinct k-mers. GWAS of the 26 traits identified significant associations for seven of them: age, sex, depression, appetite, cooked meat consumption, soda intake, and smoking. By modeling the correlation structure among k-mers, we identified a modest number of independent signals and conducted a comprehensive in silico functional annotation of these signals, revealing potential mechanisms of host-microbiota interaction. Overall, our analyses demonstrate that k-mers can capture biologically relevant functions shared across multiple taxa and provide a refined modeling framework that complements the standard taxonomic-based screening approach.