Dietary metabarcoding – the process of taxonomic identification of food species from DNA in consumer guts or feces – has been rapidly adopted by ecologists to gain insights into biocontrol, invasive species, and the structure of food webs. However, an outstanding issue with metabarcoding is the semi-quantitative nature of the data it provides: because metabarcoding is likely to produce false positives for some prey more often than for other prey, we cannot infer relative frequencies of prey in the diet. To correct for this, we can adjust detected prey frequencies using DNA detectability half-lives unique to each predator-prey combination. Because the feeding experiments required to deduce these half-lives are time- and resource-intensive, our ability to weight the frequency of observations using their detectability has thus far been limited to systems with just a few prey. Here, we present a meta-analysis of 24 spider prey DNA half-lives and show that these half-lives are predictable given predator and prey mass, predator family, temperature, and DNA amplicon length. We further provide a new technique for weighting observations with half-lives, which allows not just for the ranking of prey in the diet, but reveals the proportion of the diet each prey comprises. Lastly, we apply this method to published dietary metabarcoding data to calculate half-lives and proportion of the predator’s diet for 35 prey families, demonstrating that this technique can generate improved understanding of diets in real, diverse systems.

See README file.