Metabarcoding of a mock community of soil invertebrates: DNA extraction, false positives, and data filtration

Abstract

Metabarcoding is a powerful tool widely used to analyze soil communities but methodological difficulties can introduce biases, resulting in inaccurate estimation of species diversity. We evaluated five DNA extraction methods, two PCR programs, and three post-bioinformatic processing techniques (multiple databases, filtering based on the level of similarity and on the number of reads obtained) using a mock community of soil invertebrates containing 24 species, 23 genera, and 15 families of arthropods and annelids. The choice of DNA extraction kit significantly influenced the accuracy of the metabarcoding results. Kits developed for the isolation of DNA from soil and alkaline lysis without purification showed worse results compared to those designed for isolating DNA from tissues. In the best variant of sample processing, metabarcoding yielded correct identification of 67 % species and 94 % families. Post-bioinformatic filtering dramatically increased the proportion of true positives achieving 100 % at the family, over 73 % at the genus, and more than 60 % at the species level. Thus, an acceptable level of accuracy may be achieved in field surveys with relatively little time and effort.