Dietary DNA metabarcoding data analysis and comparison across
markers
A total of 2,197,057, 2,499,872 and 2,245,225 raw sequence reads were
obtained by HTS from MG2 , MiFish and Teleo ,
respectively. Information on the numbers of ASVs and reads retained
after the sequence-length filtering and denoising, and after the
assignment step is summarized in Table 2. A large proportion of reads
were assigned to fish (Actinopterygii ) for 12S rRNA Teleo and MiFish (74%, and 69% respectively), whereas the percentage
of fish reads assigned using MG2 (COI gene) was much lower (3.3
%). The three markers showed very strong taxonomic discriminatory
power, with the vast majority of taxa (>90%) identified at
the species level but Teleo and MiFish revealed
significantly more fish taxa (58 and 57 respectively) than MG2 (37).
Chao2 asymptotic richness estimate (63.68 ± 5.8 for MiFish , 73.9
± 10.9 for Teleo , 37.8 ± 1.3 for MG2 ) suggested the
presence of few undetected prey taxa compared to the actual number prey
detected (Figure S3).
The two 12S rRNA markers revealed similar fish assemblages, with more
than 71% species shared (48 out of 67) (Figure S4). However, 27%
(N=19) of the species were exclusivly detected by only one of the two
makers. Species detected by all three markers (N=32) were generally
those with the highest frequency of occurrence, with the top 13 most
frequently detected species being consistently identified across all
markers. The Frequency of Occurrences of fish in the crustacean samples
(FOO) was highly correlated between the MiFish and Teleo markers (r²=0.72). Furthermore, for both markers, Relative Read
Abundance (RRA) and FOO showed good correlation (r² = 0.50 and r² = 0.48
for MiFish and Teleo respectively), indicating that the species
frequently detected also tended to have higher sequencing coverage. TheMG2 marker revealed only one species,Saxatila aff. saxatilis sp2 , which was not detected byTeleo or MiFish , out of the 37 taxa identified (Figure
S4).
The three markers collectively detected 68 species belonging to 6
different orders, 21 families and 46 genera (Table S1). The vast
majority (N= 66; or 97%) were correctly assigned based on the reference
databases. For the remaining two species, local wildlife experts helped
in refining the taxonomic assignments.