maldipickr streamlines high-throughput isolation by dereplicating bacterial MALDI-TOF spectra

Abstract

Multi-cellular organisms are colonized by microbial communities that influence their health. One approach to investigate these microbiomes consists of isolating single microbes from the community for further characterization and use in functional studies.One major hurdle in this process is to rapidly identify bacteria. Depending on the isolation workflow, this usually necessitates reducing the redundancy of bacterial isolates, as many of the pure cultures obtained originate from clonal cells of the same organism. The Bruker MALDI-TOF Biotyper is used routinely for fast and cheap culture identification to the species level, which is however bound to database completion. Bioinformatic tools to dereplicate spectra have been developed, but there is urgent need for a cross-platform open software that enables transparent and reproducible dereplication scenarios.

 

Therefore, we implemented the R package "maldipickr", which rapidly analyses spectra to select species-level representatives. This package covers the preprocessing and dereplication steps needed to cluster MALDI-TOF spectra based on their similarity before final cherry-picking. It enables the easy exclusion of spectra and/or clusters to accommodate flexible cherry-picking strategies.

 

We processed 80 spectra from a 10-species mock dataset including a novel species. maldipickr correctly picked 10 or 11 isolates depending on the similarity threshold (Adjusted Rand Index: 1 or 0.986). Another tool, SPeDE  picked 13 or 10 isolates depending on the settings (ARI: 0.921 or 0.971). Cherry-picking using the Biotyper taxonomic identifications inflated the tally to 36 picked species (ARI: 0.702). maldipickr is available on CRAN for easy installation and is being benchmarked with increasingly complex datasets.