Bayesian adaptive design cuts phenotyping burden in microbial genome-wide association studies

A preprint posted to bioRxiv describes BASS-GWAS (Bayesian Adaptive Sequential Sampling GWAS), a computational framework designed to reduce the phenotyping bottleneck in bacterial genome-wide association studies.

As collections of sequenced bacterial isolates have expanded, the main limiting factor for microbial GWAS has shifted from genome sequencing to the labour-intensive phenotyping of sufficient isolates to reach adequate statistical power. BASS-GWAS addresses this by coupling Bayesian adaptive experimental design with a sparse regression model. Rather than phenotyping isolates at random or exhaustively, the algorithm iteratively selects the isolates whose phenotypic data would be most informative, prioritising those expected to maximally update genetic effect estimates.

According to the preprint, the method efficiently recovered genetic associations in benchmarking analyses while requiring substantially fewer phenotyped isolates than conventional approaches. The authors suggest the framework could be applied broadly across microbial traits of biological and applied interest, including antimicrobial resistance and virulence phenotypes.

This work has not yet been peer-reviewed. The preprint is available at bioRxiv (DOI: 10.64898/2026.08.26.747358). It is likely to be of primary interest to researchers working in microbial genomics, statistical genetics, and experimental design.