Figure 1. Overview of TQMD phases and heuristics.

(A) The preparation phase consists in downloading newly released prokaryotic genomes from NCBI RefSeq and to pre-compute all per-genome information required to run the second phase: k-mer composition, assembly quality, annotation richness, contamination level (and completeness) level. Pre-computed information for single genomes is stored in a relational database associated with TQMD. (B) The dereplication phase then retrieves this information for all genomes to dereplicate from the database and clusters the genomes from pairwise distances computed on the fly. Cluster representatives (one per cluster created) are chosen for each cluster based on the single-genome metrics computed during the preparation phase. The dereplication is iterative and the process repeats until representative genomes cannot be dereplicated anymore, which produces the final list of representatives. Parallelized steps are shown as overlaid boxes.