A guided tour into genomic contamination detection

Nowadays, genomes constitute the basis of many research endeavors. This is especially true since the decrease of sequencing cost has led to an explosion of the number of genomes in public repositories. Estimating the contamination level, i.e. the inclusion of unwanted DNA in genomic materials, of this deluge of data has become a field in itself, with numerous algorithms now available and an increasing rate of publication over the years. As newly released tools do not simply replace older ones, but have their own scope, it becomes difficult for scientists to efficiently determine which tool to use in their study. Recently, we published an overview of the main characteristics and applicability of 18 algorithms dedicated to the estimation of genomic contamination[1]. For instance, the conceptual differences between database-free tools and those associated with a reference database have an effect on detection sensitivity, as do the difference between genome-wide and marker-based methods. Beyond this typology of tools, we present here a new analysis designed to compare algorithms on a large simulated dataset derived from empirical data. This protocol, dubbed CRACOT for CRitical Assessment of COntamination detection at multiple Taxonomic levels, further reveals both under- and over-detection by even the most commonly used algorithms, with simulated contamination events ranging from inter-phylum to inter-species.

[1] Cornet L, Baurain D. Contamination detection in genomic data: more is not enough. Genome Biology. 2022;23:60.