High heterogeneity of fish species detection among families
A total of 299,479,007 reads (length > 10 bp) were produced over the 92 eDNA samples corresponding to 14,423 unique sequences with a mean of 307 unique sequences per sample (± 134 SD). In a conservative approach, stringent bioinformatic filters retained 9,345 unique sequences so 65% of the total. These 9,345 unique sequences were then assigned to different taxonomic levels using the following genetic similarity thresholds: 100-98% for species, 90-98% for genus, 85-90% for family and 80-85% for order. This set of thresholds retained 7,389 unique sequences resulting in 678 taxonomic assignments (Suppl. Table S2).
A total of 310 species were detected, including 211 coastal fish species present in the checklist of the Bird’s Head Peninsula and 99 fish species present in other regions but absent from this checklist (Fig. 2a). Conversely, 183 sequenced fish species which are present in the Bird’s Head Peninsula were not detected in our eDNA samples using our stringent filters, representing 53.6% of the sequenced species present in the checklist. Since 75.5% of fish species in the checklist of the Bird’s Head Peninsula were not sequenced for the 12S rDNA, the largest part of fish species diversity remained hidden through direct assignment (Suppl. Table S1).
A total of 282 genera and 128 families of fish were detected compared to the regional checklist of 508 genera and 112 families out of which 46.1% and 72.3% are sequenced respectively (Suppl. Table S1). The number of fish species per family varied from 1 to 191 in the Bird’s Head checklist (Fig. 2b), the richest family being the Gobiidae. Only 12 species of Gobiidae were detected in our 92 samples. Meanwhile, the most represented family in the eDNA samples was the Labridae with 48 species (15.5% of the species found in the samples) out of 136 in the checklist (Fig. 2b).
The percentage of fish species sequenced per family varied between 0 and 100% with a mean of 40.3% (± 31% SD) in the Bird’s Head Peninsula checklist while the percentage of detected species per family varied between 0 and 100% with a mean of 27.1% (± 30.2% SD) in eDNA samples (Fig. 2b). These two percentages were significantly and strongly related (p < 0.001) with the percentage of species sequenced per family explaining 85% of variation in the percentage of detected species per family (Fig. 2c).