Metabarcoding versus mapping unassembled shotgun reads for identification of prey consumed by arthropod epigeal predators

Abstract

Background A central challenge of DNA gut content analysis is to identify prey in a highly degraded DNA community. In this study, we evaluated prey detection using metabarcoding and a method of mapping unassembled shotgun reads (Lazaro). Results In a mock prey community, metabarcoding did not detect any prey, probably owing to primer choice and/or preferential predator DNA amplification, while Lazaro detected prey with accuracy 43–71%. Gut content analysis of field-collected arthropod epigeal predators (3 ants, 1 dermapteran, and 1 carabid) from agricultural habitats in Brazil (27 samples, 46–273 individuals per sample) revealed that 64% of the prey species detections by either method were not confirmed by melting curve analysis and 87% of the true prey were detected in common. We hypothesized that Lazaro would detect fewer true- and false-positive and more false-negative prey with greater taxonomic resolution than metabarcoding but found that the methods were similar in sensitivity, specificity, false discovery rate, false omission rate, and accuracy. There was a positive correlation between the relative prey DNA concentration in the samples and the number of prey reads detected by Lazaro, while this was inconsistent for metabarcoding. Conclusions Metabarcoding and Lazaro had similar, but partially complementary, detection of prey in arthropod predator guts. However, while Lazaro was almost 2× more expensive, the number of reads was related to the amount of prey DNA, suggesting that Lazaro may provide quantitative prey information while metabarcoding did not.

Quantitative prey species detection in predator guts across multiple trophic levels by mapping unassembled shotgun reads

“…The blastn xml output was converted to a tabular format and, to reduce the output file size, filtered for identity above 90% and overlap length above 100 bp (taxonomic resolution is poor at overlap lengths <100 bp, Meusnier et al, 2008). The reduced blastn files and associated metadata are given in Paula et al (2022b). In the reduced blastn output file, each read could match multiple reference sequences (each match is called a hit).…”

“…The complete bioinformatic pipelines using Java and R scripts are available in the GitHub repository: https://github.com/molec ularecolo gy/DDSS (snapshot available at Paula et al, 2022b). Raw sequence reads (Fastq) were deposited at GenBank in BioProject…”

“…The complete bioinformatic pipelines using Java and R scripts are available in the GitHub repository: https://github.com/molecular-ecology/DDSS (snapshot available at Paula et al, 2022b). Raw sequence reads (Fastq) were deposited at GenBank in BioProject PRJNA377711 with SRA access codes SRR5342614–SRR5342626 and SRR5342628–SRR5342629, and BioProject PRJNA377714 with SRA access codes SRR5350637–SRR5350641 and SRR5350643–SRR5350656.…”

“…Raw sequence reads (Fastq) were deposited at GenBank in BioProject PRJNA377711 with SRA access codes SRR5342614–SRR5342626 and SRR5342628–SRR5342629, and BioProject PRJNA377714 with SRA access codes SRR5350637–SRR5350641 and SRR5350643–SRR5350656. All sample metadata are given in Paula et al (2022b) with sample IDs SAMN06464667, SAMN06464677, SAMN06464679, SAMN06464681, SAMN06464684, SAMN06464686, SAMN06464689, SAMN06464704 and SAMN06464705, and the blastn output files that were used in the Lazaro analysis are in Blastn_files_for_threshold_determination.zip.…”

“…We showed in previous work (Paula et al, 2015(Paula et al, , 2016(Paula et al, , 2022a that direct DNA shotgun sequencing can be applied to characterize trophic interactions through gut content analysis and that a priori barcode selection, sample DNA enrichment and assembly of reads were not required. The independence from barcode(s) selection before sample preparation gives freedom to use any piece of genetic material (any kind of gene(s) or partial or full genomes) and multiple reference databases for data analysis.…”