Identifying lineage effects when controlling for population structure improves power in bacterial association studies

Abstract

Bacteria pose unique challenges for genome-wide association studies because of strong structuring into distinct strains and substantial linkage disequilibrium across the genome1,2. Although methods developed for human studies can correct for strain structure3,4, this risks considerable loss-of-power because genetic differences between strains often contribute substantial phenotypic variability5. Here, we propose a new method that captures lineage-level associations even when locus-specific associations cannot be fine-mapped. We demonstrate its ability to detect genes and genetic variants underlying resistance to 17 antimicrobials in 3,144 isolates from four taxonomically diverse clonal and recombining bacteria: Mycobacterium tuberculosis, Staphylococcus aureus, Escherichia coli and Klebsiella pneumoniae. Strong selection, recombination and penetrance confer high power to recover known antimicrobial resistance mechanisms and reveal a candidate association between the outer membrane porin nmpC and cefazolin resistance in E. coli. Hence, our method pinpoints locus-specific effects where possible and boosts power by detecting lineage-level differences when fine-mapping is intractable.

High risk Epstein‐Barr virus variants characterized by distinct polymorphisms in the EBER locus are strongly associated with nasopharyngeal carcinoma

“…The association between NPC and the variants with MAF > 0.05 and missingness <0.1 was tested by a GWAS. To control for population structure, the GWAS was performed using a linear mixed model implemented in GEMMA provided by bugwas . A relatedness matrix amongst genomes was generated with the variants.…”

A genome-wide association study identifies a horizontally transferred bacterial surface adhesin gene associated with antimicrobial resistant strains

“…Extraction of 52,363 bi-allelic SNPs in the genomes, and the calculation of an n  ×  n relatedness matrix that summarizes all genetic covariance among the strains, was performed to control for background population structure. Using the linear mixed regression model in the R package bugwas 13, which uses the relatedness matrix to model the background random effect, we found that the association was highly significant for 469 out of 488 kmers after false discovery rate (FDR) correction (P FDR  < 0.05).…”

“…A complementary approach is to explore novel genetic elements, which could be associated with antimicrobial resistance in strains not possessing the commonly known resistance features. Methods for applying genome-wide association studies (GWAS) to bacteria have recently been developed1213. In this study, we utilized these methods to identify novel genetic elements associated with carbapenem resistance, and evaluated effectiveness of this approach.…”

“…We applied the method to genome sequences from carefully selected carbapenem-resistant strains that did not possess the commonly reported resistance features as well as an equivalent number of carbapenem-susceptible strains, spanning various genotypes, from the PATRIC database. We took a kmer-based approach1213, in which the genome sequence of each isolate was fragmented into unique, overlapping, 31-bp DNA motifs or kmers, that could be used to identify any kind of genetic variation such as single nucleotide polymorphisms (SNPs), indels, and the presence or absence of a whole gene or gene region. We then explored the DNA motifs that were significantly associated with carbapenem resistance.…”

“…We then explored the DNA motifs that were significantly associated with carbapenem resistance. We accounted for the inter-dependence of the strains and population structure, following a recently developed method13 that directly incorporates the relatedness of all of strains and employs statistical tests to examine potential lineage effect as well as a locus effect for a given kmer. Validation and annotation were performed using multiple complete genomes from other strains.…”