Fig. 3

Comparison of taxonomic profiling results of 2bRAD-M, shotgun WGS, and 16S rRNA sequencing methods in mock microbial communities. a 2bRAD-M performance in samples with low amount (LoA), high DNA degradation (HiD), and high host DNA contamination (HoC) based on a mock community of five bacteria species (Mock-CAS). LoA are samples with a gradient of total DNA concentrations (50 ng, 20 ng, 10 ng, 1 ng, 0.1 ng, 0.01 ng, and 0.001 ng); HiD includes samples with fragmented bacterial DNA (50 bp or 150 bp). HoC are samples with a mixture of human DNA (90% or 99%) and bacterial DNA. Three technical replicates were included for each group. The L2 score that measures the microbial composition similarity between 2bRAD-M and WMS is shown on the head of each stacked bar plot. The false-positive identification rates of reads (i.e., reads mapped to species not in the mock community) are very low (0.9% in WMS and 1% in 2bRAD-M). b Benchmarking 2bRAD-M against conventional metagenomic approaches using the mock community of MSA 1002. Each stacked bar plot in the left panel shows the resulting taxonomic profile from a benchmarked method in a library preparation or bioinformatic setting. The white blank refers to false-positive identifications. For WMS data, we employed various bioinformatic tools (e.g., MetaPhlAn2, Bracken, Kraken2, and mOTUs2) to generate taxonomic profiles on MSA 1002. For 2bRAD-M data, we applied our own bioinformatic pipeline to generate the taxonomic profiles of MSA 1002 with low to high DNA amounts, and under various degrees of host DNA contamination. In the right panel, bars in each row indicate the corresponding precision, recall, L2 similarity, Bray-Curtis dissimilarity, and Shannon index similarity of the predicted taxonomic profile as compared to the ground truth. The bars in the “Shannon” column represent the correlation between the Shannon index of the profiling results and the ground truth (see the “Materials and Methods”)