https://orcid.org/0000-0003-3188-1747
Figures
Abstract
Selecting primers for DNA amplification poses a challenge in environmental DNA (eDNA) studies, particularly in biodiverse tropical ecosystems. Researchers must balance taxonomic coverage, primer specificity, and study feasibility when choosing between a single primer pair and multiple primer pairs. Empirical tests comparing these strategies are essential for improving eDNA protocols, enhancing the detection of aquatic biodiversity, and supporting conservation and environmental monitoring efforts. This study aimed to evaluate how different approaches affect the specificity and efficiency of detecting fish species in eDNA samples from the Teles Pires River basin. We (1) synthesized modified versions of primers commonly used in ichthyofaunal surveys; (2) compared the results generated by these modified primers with the standard primer versions; and (3) evaluated the species detection using two reference sequence databases: a global taxonomic database (GenBank) and a locally constructed taxonomic database containing samples of fish species previously collected in situ. The 12S-V5_mod primers detected the most fish species across the databases (67, 104, and 130 in the Midori2, Local, and joint databases, respectively), followed by the MiFish_mod primers, which detected 47, 101, and 114 species. Finally, the MiFish primers detected 39, 73, and 78 species. However, the 12S-V5_mod primers also had lower taxonomic specificity of fish species and amplified a broader range of vertebrate taxa than the MiFish_mod primers. The modified primers outperformed the MiFish primers, and detectability nearly doubled when the Local reference database was used. These results speak to the critical need for constructing reference databases of regional biodiversity, and the need to incorporate multiple primer pairs in eDNA studies.
Citation: Hashimoto S, Carvalho SV, Miranda SdSC, Machado VN, Carvalho LN, de Deus CP, et al. (2026) Assessing fish diversity in the Amazon: The impact of primers and reference libraries in eDNA surveys. PLoS One 21(8): e0336758. https://doi.org/10.1371/journal.pone.0336758
Editor: Mizanur Rahman, UTRGV: The University of Texas Rio Grande Valley, UNITED STATES OF AMERICA
Received: October 30, 2025; Accepted: June 30, 2026; Published: August 13, 2026
Copyright: © 2026 Hashimoto et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data underlying the results presented in the study are available from NCBI BioProject: PRJNA1346840, and at https://github.com/legalLab/publications.
Funding: This work was supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001, Additional federal support was provided by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) [grant numbers 316532/2023-1 and 421435/2025-9 to TH, and 316531/2023-5 to IPF]. SH received support from the Fundação de Amparo à Pesquisa do Estado do Amazonas (FAPEAM/POSGRAD) program. This research was also made possible by Sinop Energia/CES, which financed the ichthyofauna monitoring program during the pre-filling period of the Sinop Hydroelectric Power Plant (UHE-Sinop) reservoir, and the Genetic Monitoring Project of the Ichthyofauna of the Sinop HPP. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The Amazon basin is home to the world’s most diverse ichthyofauna, with estimates exceeding 3,000 species [1]. This estimate is growing rapidly. Of the 407 new species of fishes described in 2024, 260 were freshwater and 84 from South America, and most of them from the Amazon basin [2]. Of the 3,000 + species of Amazonian fishes, only 254 have DNA reference sequences for the Miya 12S region, and 324 have DNA reference sequences for the Riaz 12S region [3] in GenBank Release 259. The Miya 12S region taxa is a perfect subset of the Riaz 12S region taxa. The 324 species with 12S DNA reference sequences represent only 10% of the described fish species diversity of the Amazon basin, presenting a significant impediment in our ability to conduct biodiversity surveys and monitoring based on environmental DNA (eDNA) analyses. This is especially worrying as this rich aquatic biodiversity is increasingly threatened by human activities, including the intensive use of water resources for agricultural expansion, urbanization, and the construction of hydroelectric dams, all of which contribute to significant biodiversity loss [4–6].
The expansion of agriculture and deforestation has led to increased river silting, habitat degradation, and water contamination from pesticides, all of which directly impact fish populations [4]. For example, the Teles Pires River basin is located in a region characterized by intense agricultural activity [7,8], significant sport fishing [9], and several hydroelectric projects [10]. These activities have direct and negative effects on the local aquatic fauna.
The rapid transformation of aquatic environments due to anthropogenic pressures coupled with biodiversity knowledge gaps demands the adoption of more efficient and faster monitoring methods for monitoring and identifying fauna. Traditional methods, while essential, often struggle to keep pace with the speed of biodiversity loss, as they rely on physical sampling and detailed taxonomic identification by qualified experts, which is time-consuming and costly. In this context, the analysis of environmental DNA (eDNA) emerges as an innovative and highly sensitive approach for monitoring aquatic biodiversity [11]. Environmental DNA analyses enable the detection of species presence from traces of genetic material released into the environment, such as cell fragments, excrement, mucus, and decomposing matter [12], making identification faster, less invasive, and potentially more comprehensive than conventional methods. Therefore, eDNA represents a promising tool for the early detection of invasive species, monitoring threatened populations, and assessing the health of aquatic ecosystems [13], thereby aiding biodiversity conservation and management plans.
The analysis of eDNA from water has emerged as a powerful tool for conducting more detailed and efficient assessments of fish communities [14,15]. Additionally, eDNA allows for the analysis of the temporal dynamics of fish communities, including seasonal and long-term variations [16], and provides data on the relative abundance of different species [17]. This methodological advancement represents a significant step toward understanding aquatic biodiversity and enhancing conservation and environmental management strategies.
In eDNA studies, the mitochondrial 12S rDNA gene is commonly used for fish species identification due to the specificity of the primers used and the effectiveness of the amplified region in distinguishing among different taxa [18–20]. While the 12S gene is prevalent, some studies incorporate other gene regions, such as COI, 16S, and 18S [21–24]. The most frequently used primers for sampling aquatic environments and fish include MiFish-U [25] and 12S-V5 [26], each amplifying different but abutting regions within the 12S gene.
The choice of primers for DNA amplification poses a challenge in eDNA studies, particularly in diverse ecosystems like the Neotropics which are taxonomically very diverse. For example the 3’ ends of the universal primers developed by Miya et al. [25] for fishes have mismatches with characiform 12S rDNA. This will result in amplification biases that favor non-characiform lineages and will lead to an underestimation of actual biodiversity [27]. Specifically, while Characiformes represent about 43% of the Amazon basin’s total fish species diversity, in the eDNA survey of the lower Javari River basin, characiform fishes represented only 27.1% of the detected OTUs [28]. Some of these biases can be mitigated by the use of multiple primer sets which can enhance taxonomic coverage and improve the detection of species with varying target sequences [18], however, at the cost of additional time, laboratory and analytical effort and expense. Solutions can also be sought in redesigning primers to better match and amplify local biodiversity.
The success of any eDNA study is predicated on the availability of a locality informative reference database, a recurring issue in studies comparing primers where primer “success” is predicated on the reference database used [29–31]. With only about 10% of all Amazonian basin fishes having one or both of the commonly used 12S regions included in public databases—percentage calculated as the proportion of Amazonian taxa available in https://shiny.cefe.cnrs.fr/GAPeDNA/ [3] and assuming 3000 + fish species in the Amazon basin [1]—biodiversity monitoring via eDNA will always be lacking until more effort is put into generating curated reference databases. In an ichthyofaunal survey of the lower Javari River in the western Amazon basin, Santana et al. [28] sampled 443 species by traditional methods—of which 60 were new to science—and detected 222 operational taxonomic units of which 58 could be assigned to species using a 98.5% similarity criterion and the publicly available MiFish DB ver. 36. Furthermore, only 17 species were detected by both physical and eDNA surveys. This study clearly demonstrates that the standard MiFish primers amplify only a portion—no more than half—of the local fish diversity, and that 75% of this diversity cannot be assigned to a species. It is therefore clear that we need primers that more efficiently sample local diversity, and we need reference databases that are representative of local taxonomic diversity.
In this study, we synthesized two primer cocktails which also capture the taxonomic diversity of Amazonian fishes to evaluate and compare their performance with the MiFish primer, which is widely used in eDNA studies. Our goal was to assess the specificity of the sequences (% of taxa assigned to species level in Fig 3), informativeness (number of species detected in Fig 3) and efficiency (% on target sequences in Table 2) in detecting vertebrates in general and fish specifically in the Teles Pires River basin. Specifically, we aimed to: a) compare the original primer from Miya et al. [25] with our modified primers for Amazonian fishes; b) evaluate the efficiency of these primer sets using both the global taxonomic database, GenBank, and a locally constructed taxonomic database to determine how much GenBank can detect and how much the local reference enhances the accuracy of detecting local diversity; and c) make publically available a new set of primers which will make the study of Amazonian biodiversity more efficient and comprehensive.
Materials and methods
Ethics and permit approval statement
Environmental sample collection was authorized by the Ministério do Meio Ambiente, Instituto Chico Mendes de Conservação da Biodiversidade under SISBIO permit number 76359−1 granted to SH. Access to the genetic heritage associated with this study was registered in the National System for the Management of Genetic Heritage and Associated Traditional Knowledge (Sistema Nacional de Gestão do Patrimônio Genético e do Conhecimento Tradicional Associado – SISGEN) under accession number AB8314D, granted to IPF.
eDNA study area and sampling
The confluence of the Teles Pires and the Juruena rivers forms the Tapajós River, a major clear-water tributary on the right bank of the Amazon River [32]. The Teles Pires River is 1,457 km long, with the majority of its drainage area located in the state of Mato Grosso (1,370 km) and a smaller portion in the state of Pará. The Teles Pires River is characterized as a clear-water river due to the typical rock formations of the Brazilian Shield, which drain waters with minimal amounts of suspended sediment and organic matter, and very low electrical conductivity. The hydrological cycle exhibits a flood pulse pattern, with the rainy season occurring from November to February and the dry season from May to October, reflecting the region’s seasonality [33].
For this study, we sampled at 22 localities along the Teles Pires River (Fig 1) during the months of July and August, 2021. At each locality we collected three water replicates using a Van Dorn bottle and stored them in labeled 500 mL sterile bottles, kept away from light and heat. We then filtered each sample using a 50 mL disposable syringe fitted with a 0.45 μm Sterivex-GPTM PES filter (Merck Millipore, Inc., Darmstadt, Germany) attached to the tip. We completed all sample filtration within 24 hours of collection to prevent degradation of the DNA present in the samples. After each filtration, the equipment were sterilized with 10% sodium hypochlorite for 10 minutes and rinsed with distilled water, following the protocol of Handley et al. [16] to avoid the risk of contamination. The filters were stored in an ATL buffer (Qiagen), sealed, labeled, and stored away from light and heat until DNA extraction.
Laboratory processing
We extracted DNA in a dedicated eDNA extraction room, separate from the PCR room. Each extraction began with cleaning the room using 70% alcohol, followed by bleach for decontamination. We also disinfected all instruments, including pipettes, scalpels, and tweezers used in the DNA extraction steps with alcohol and bleach, and exposed them to UV light for 30 minutes as a final cleaning step.
We then extracted DNA from the filters using the DNeasy PowerSoil Kit (Qiagen), with a modification in the digestion step, which involved digestion at 60 °C for 2 hours on a shaker, followed by quantification using a NanoDrop 2000 spectrophotometer (Thermo Scientific).
We amplified each gene fragment using a two-step PCR protocol. The first step targeted the amplification of a mitochondrial 12S rDNA gene fragment, while the second step involved ligating sequencing adapters for sample identification during high-throughput sequencing. We carried out the first PCR in a final volume of 25 µL, which included: 17.45 µL of ultrapure water, 2.5 µL of 10X buffer (containing 500 mM Tris-Cl pH 9.2, 160 mM ammonium sulfate, 0.5% Brij 58, and 35 mM magnesium chloride), 2.5 µL of dNTPs (at a concentration of 2 mM), 0.75 µL of each 2µM primer (forward and reverse), 0.05 µL of OmniTaq polymerase (DNA Polymerase Technology, Inc.), and 1.0 µL of environmental DNA (eDNA). In addition to the original MiFish primers [25], we also amplified the eDNA using modified versions of the MiFish and 12S-V5 primers [26] (Table 1). These modifications were based on observed sequence variation in the 12S region of Amazonian fishes deposited in GenBank.
The PCR cycling conditions were as follows: initial denaturation at 94 °C for 1 minute; followed by 40 cycles of denaturation at 94 °C for 5 seconds, primer-specific annealing at a designated temperature (Table 1) for 30 seconds, and extension at 68 °C for 10 seconds. We performed each PCR reaction, including negative controls, in triplicates to minimize amplification biases caused by stochastic variation in early PCR cycles. We visualized each PCR product on a 1% agarose gel stained with GelRed (Biotium), and then pooled the triplicate reactions into a single composite sample and bead purified it [34] for subsequent indexing PCR.
We used a dual indexing strategy to individualize each sample. In the indexing PCR we used the same conditions as in the first PCR except we used one Q5 and one Q7 primer at 55 °C annealing temperature, 3 µL of the first PCR product, and 15 PCR cycles. We evaluated the PCR reactions on a 1% agarose gel, and then purified them using BOMB paramagnetic beads [34]. After quantifying each purified PCR product using the Qubit fluorometer, we pooled all PCR products equimolarly for sequencing. We sequenced the pooled library on the Illumina HiSeq 4000 platform (Illumina, Azenta Life Sciences, USA).
Construction of the local reference library
We constructed the Local DNA reference library using samples of fish specimens collected between 2016 and 2019 as part of the ichthyofauna monitoring program associated with the Sinop Hydroelectric Power Plant (HPP). All specimens were identified to the species level, cataloged, and deposited as voucher specimens in the Ichthyological Collection of the Southern Amazon Biological Collection, housed at the Tropical Ichthyology Laboratory (LIT-ABAM) in Sinop, Mato Grosso, by the ichthyological staff of LIT-ABAM [35]. Tissue samples of these specimens were deposited in LIT-ABAM, CTGA tissue collection of UFAM or the genetic resources collection of INPA. Of the 354 species of fishes recorded by Carvalho et al. [35] for the middle and upper Teles Pires River, 173 species had tissues available and were included in the Local DNA reference library, representing 49% of the species collected in the faunal surveys prior to the construction of the Sinop Hydroelectric Power Plant [35].
To complement the Local database, we also included additional species from the Xingu River basin, some of which might be shared between the two basins—documented in the studies by Carvalho et al. [36] and Ximenes et al. [37]. Muscle tissue samples from these additional specimens are archived in the Animal Genetics Tissue Collection of the Laboratory of Animal Evolution and Genetics at the Federal University of Amazonas (CTGA/LEGAL/UFAM). On average, two individuals per species were included in the local database, resulting in a total of 390 species, representing 171 genera, 47 families, and 12 orders (S1 Table).
We used the Midori2 raw FASTA GB259 database [38] as our global reference library. We standardized the taxonomy of both the Local as well as the Midori2 reference database to a seven level taxonomy using TaxonKit [39] and validated the taxonomy of fish taxa using the Eschmeyer’s Catalog of Fishes [40]. Finally, we simplified the Midori2 database FASTA headers to Genbank IDs only, and using this simplified Midori2 FASTA, the Local database FASTA and the merged FASTA, we generated three local BLAST+ databases.
Bioinformatics processing
The raw data were processed using the package barcode_splitter v0.18.6 [41], which sorted the samples into folders based on their index sequences. From the raw reads, we extracted only those sequences which had both the forward and reverse primers on both ends, and then removed these primers with cutadapt v5.0 [42]. The paired-end reads were merged into consensus sequences using PEAR v0.9.11 [43], chimeric sequences were removed and then the merged reads were dereplicated using VSEARCH v2.21.1 [44] and clustered using SWARM v3.1.4 [45]. We then counted the number of reads for each primer, and normalized the read counts across primers by randomly subsetting the original reads using a custom Shell script. Finally we compared the clustered sequences against the Midori2 global reference library [38], a Local reference library, and a merged global/local library using BLAST+ [46].
Statistical analyses
We processed the BLAST+ output using a custom R [47] script, assigning sequences to taxa based on similarity thresholds following Beatty et al. [48]: ≥ 97% for species, ≥ 95% for genus, ≥ 90% for family, and ≥80% for order, removing all sequences shorter than 99 bp, and all taxa with less than four reads. We visualized the results using ggplot2 [49] and the VennDiagram package v1.7.3 [50]. To estimate species richness and visualize accumulation curves across primers, we used the iNEXT package v3.0.0 [51], based on species abundance and richness data.
Results
Primer set performance
From the 66 samples and nine negative controls, a total of 62,278,389 raw sequences were generated. Among the primers, MiFish produced the highest number of raw sequences (29,278,389), followed by 12S-V5_mod (19,685,401) and MiFish_mod (13,093,553). To enable fair comparison of the three primer pairs, we standardized the read counts to the number of reads in the MiFish_mod primer pair which corresponded to an average of 189,761 reads per sample.
Post-filtering retention (on target) rates were 1.06% for MiFish, 19.65% for MiFish_mod, and 20.46% for 12S-V5_mod primers. When using the Midori2 database for taxonomic assignment, 4.8% of MiFish on target reads (6,557 reads), 25.9% of MiFish_mod on target reads (665,184), and 41.9% of 12S-V5_mod on target reads (1,121,444) were successfully assigned as fish at any taxonomic level. Using the Local reference database, 6.9% of MiFish on target reads (9,524), 56.5% of MiFish_mod on target reads (1,450,684), and 60.1% of 12S-V5_mod on target reads (1,607,832) were classified as fish at any taxonomic level. Combining both databases, 7.1% of MiFish on target reads (9,757), 56.5% of MiFish_mod on target reads (1,452,107), and 62.4% of 12S-V5_mod on target read (1,667,143) were successfully assigned as fish at any taxonomic level (Table 2).
Amplification occurred in the negative controls, generating a total of 242,066 raw reads and 28,824 filtered reads for the MiFish primers; 745,462 raw and 56,933 filtered reads for MiFish_mod; and 543,065 raw and 33,157 filtered reads for 12S-V5_mod. In the negative controls we detected between two (MiFish x Midori2) and 12 (MiFish_mod x Local) species where presence of a species was supported by one or more sequences. Using a binomial function, we calculated the expected upper 95% CI of the number of reads in the negative control of each species, and removed this number from the total number of detected reads per species per sampling locality. Subsequently, any species that had less than four reads was removed from further analysis.
Primer set X reference database
The MiFish, MiFish_mod and 12S-V5_mod primer sets, in this order, consistently detect an increasing number of taxa across the three reference datasets. The MiFish primer set detects the smallest number of taxa most because it amplifies only a restricted subset of Amazonian fish species, performing especially poorly with Characiform taxa (Fig 7). When compared with the MiFish_mod primer pair, it detects only 50–60% of the taxa detected by the MiFish_mod primer pair (Fig 3). At the same time, a greater number of reads can be assigned to species and genus levels, reflecting the intersection of the greater specificity of this primer pair for the taxa that are deposited in public databases. The MiFish_mod primer set amplifies a much broader range of taxa, and this is also reflected in 18% of all taxonomic assignments being made to non-Amazonian taxa when the reference dataset is deficient in local taxa, such as the Midori2 reference dataset. However, when a local reference database is available, even if it contains only 50% of the expected local diversity, assignments to non Amazonian taxa drops to 4%. The MiFish_mod primer set also has high specificity, with 73.2% taxa being assigned to a species or a genus in the combined Midori2 and Local reference databases (Fig 3). The 12S-V5_mod primer pair detects the greatest number of taxa but at the costs of lowest specificity with 60.6% of taxa being assigned to a species or a genus in the combined Midori2 and Local reference databases (Fig 3). The number of detected non-Amazonian fishes was also slightly higher than in the MiFish_mod primer pair (Fig 3). The number of fish species detected by the 12S-V5_mod and MiFish_mod primer pairs against the Local reference database was effectively the same (104 vs 101 species, respectively, with 53 species being detected by both primers), while against the combined Midori2 and Local reference databases, 129 vs 113 species (53 species shared) were detected using the 12S-V5_mod and MiFish_mod primer pairs, respectively. This is probably explained, at least in part, by the GenBank Release 259—from which the Midori2 reference database was built—having 70 more Amazonian fish species for the Riaz 12S region which the 12S-V5_mod primer pair amplifies than for the Miya 12S region which the MiFish_mod primer pair amplifies. Finally, the 12S-V5_mod primer pair also amplifies more species of other vertebrate taxa than any other primer pair. It amplified species of 21 families of mammals and 22 families of birds, while the MiFish_mod primer pair amplified species of 12 families of mammals and seven families of birds, and the MiFish primer pair amplified species of eight families of mammals and no birds (S1 Fig). Both the 12S-V5_mod vs MiFish_mod primer pairs clearly perform better than the original MiFish primer pair, however, each performs better in a different way.
Set A includes all detected taxa at all taxonomic levels. Set B includes only Amazonian taxa.
Species richness and the collector’s curve
When analyzing species richness, the 12S-V5_mod primers with the merged global/local reference database detected the largest number of species (130 species), and the MiFish primers with the Midori2 database detected the smallest number of species, only 39. The 12S-V5_mod primers were able to detect more species with all three databases: 67, 104 and 130 species corresponding to Midori2, Local and global/local databases, respectively. Next were the MiFish_mods primers with 47, 101 and 114, respectively. And finally, the MiFish primers detected 39, 73 and 78 species, respectively. The modified primers were better than MiFish and species detectability increased when the Midori2 and Local databases were used together, i.e., taxonomically more comprehensive reference databases are better at assigning more sequences to a more specific taxonomic level (Fig 4).
We generated a species accumulation curve to assess the performance of each primer in combination with the reference databases (Fig 5). The results indicate that the 12S-V5_mods primers used with the global and Local reference database were the most effective in detecting species, while the MiFish primers with the Midori2 database performed the worst.
Lines represent a sampling number of individuals and dotted lines represent extrapolated curves.
The Local reference database consistently outperformed Midori2, detecting approximately twice as many species. Moreover, combining both databases increased total species detection, as each database contributed unique taxa not found in the other. These findings highlight the critical role of the Local database in enhancing species detection and improving the overall assessment of ichthyofaunal diversity in the upper Teles Pires River basin specifically, and the Amazon basin in general.
Taxonomic identification
The combined use of all three primer sets and both reference databases (Midori2 and the Local dataset) enabled the identification of 40 families, 118 genera, and 256 species. Of these, 162 species were detected using the Local database, 109 species with the Midori2 database, and only two species—Hoplias malabaricus and Sternopygus macrurus—were detected in both databases. Within the Local database, the MiFish primer detected 73 species, 10 of which were unique to this primer; 12S-V5_mod detected 104 species, with 46 exclusive detections; and MiFish_mod detected 101 species, 26 of which were unique. A total of 36 species were shared among all three primers. Using the Midori2 database, 39 species were detected with MiFish (13 exclusive), 67 species with 12S-V5_mod (50 exclusive), and 47 species with MiFish_mod (18 exclusive). Considering the combined database, the MiFish primer detected 78 species, including 14 exclusive; 12S-V5_mod detected 129 species, with 71 exclusive detections; and MiFish_mod detected 113 species, 37 of which were exclusive. These results emphasize the importance of using multiple primers and complementary reference databases to maximize species detection (Fig 6).
When comparing taxonomic detection across databases, a total of 25 families were identified using the Midori2 database: 18 families with the MiFish primer pair, 19 with 12S-V5_mod, and 20 with MiFish_mod. In contrast, the Local database identified a broader taxonomic range, detecting 38 families overall—including Potamotrygonidae (Class Chondrichthyes), which was absent from Midori2. Specifically, the MiFish primer pair detected 26 families, 12S-V5_mod detected 31, and MiFish_mod detected 28 families.
When combining both databases, the total number of families detected increased to 40, with MiFish identifying 27 families, 12S-V5_mod identifying 33, and MiFish_mod identifying 29 families.
When compared to the Local and combined databases, the Midori2 database failed to detect any species in these 15 families: Acestrorhamphidae, Bryconidae, Chilodontidae, Ctenolucidae, Cynodontidae, Gasteropelecidae, Hypopomidae, Lebiasinidae, Parodontidae, Poecilidae, Potamotrygonidae, Rhamphichthyidae, Rivulidae, Sciaenidae, and Sternopygidae. All but Acestrorhamphidae have representatives in the Midori2 databases, however, these representatives dissimilar to the species occurring in the Teles Pires River to be detected. In contrast, the Local database did not detect, at the species level, only two families: Apteronotidae (because the Local database does not contain any Apteronotidae) and Synbranchidae (because the local database only contains a sequence of a Synbranchus royal from the middle Xingu River which showed 93% similarity to the Synbranchus species sequence in the eDNA sample).
Overall, the 12S-V5_mod primer demonstrated the highest family-level detection, while MiFish showed the lowest. These results underscore the importance of using a comprehensive reference database, particularly a local database, to enhance taxonomic resolution at the family level (Fig 7).
(A) using Midori2 database. (B) using a Local reference database. (C) using databases together.
It is important to note that the Midori2 database includes a wide range of Eukaryotic taxa, not just the target group (Actinopterygii). As a result, this database also yielded detections of non-target groups, such as birds and mammals. When analyzing the total reads against the complete Midori2 database, we detected a notable number of non-target taxa, including birds from 13 orders, 25 families, and 38 species, and mammals from nine orders, 19 families, and 46 species (see S1 Fig for details). These taxa are primarily detected by the 12S-V5_mod primer set, a modification of the 12S-V5 primer set which amplifies a broad-range of vertebrate taxa [26].
Discussion
Primer efficiency
The choice of metabarcoding primers is critical when sampling megadiverse regions such as the Amazon. Although COI reference databases are generally more comprehensive, their application in eDNA studies often results in the detection of relatively few fish sequences and a disproportionately high number of bacterial sequences. This limitation arises from the lack of conserved priming regions. A practical consequence of this is that effective universal primers cannot be designed, and highly degenerate primers result in amplification biases and non-target amplifications [18]. In contrast, the 12S rDNA gene has proven to be more suitable for eDNA studies, with much higher on-target amplification rates, thereby enhancing the likelihood of detection on-target taxa in environmental samples [19,20,25].
Although the MiFish primers have been used in eDNA studies in the Amazon [28,52,53], our findings indicate that they yield a low percentage of on-target reads (1.06%) compared to the modified primers: MiFish_mod (19.65%) and 12S-V5_mod (20.46%) (Table 2; Fig 2). The MiFish primers also detected only about 50% of taxa that the MiFish_mod and 12S-V5_mod primers detected, although a larger percentage of these taxa were identified to the species level (Fig 3).
The somewhat counterintuitive observation that the regions amplified by MiFish primers showed the greatest taxonomic resolution is explained by the specificity of the MiFish primers. The MiFish primers primarily amplify taxa which were deposited in public databases or generated by the Miya group—Paleotropical and Paleoarctic fish groups, and because of their lack of degeneracy do not efficiently amplify related taxa. This results in the preferential amplification of certain taxonomic groups and also the amplification of low taxonomic diversity within these groups. At the same time, only 1.8% of all the amplified fragments were on-target, and of those only 4.8% could be assigned to a vertebrate taxon (Table 2). This is a manifestation of low PCR efficiency, where despite standardized—across the three primer sets—paramagnetic bead cleaning and size selection of eDNA PCR products, the carry over of non-target PCR products is great in the MiFish PCR. This results in many fewer vertebrate taxa being detected—only about ½ to ⅗ of the taxa amplified by the MiFish_mod and 12S-V5_mod primers, but for those taxa greater proportion are detected at the species level.
When comparing both the MiFish and MiFish_mod primers to the 12S-V5_mod primers, a larger proportion of reads can be assigned to lower taxonomic ranks due to the greater taxonomic informativeness of the MiFish region (170 bp) vs. the 12S-V5 region (100 bp) (Fig 3). However, the total number of species detected by the MiFish_mod and 12S-V5_mod primers is similar (Fig 6).
Although we did not include the original 12S-V5 primer from Riaz et al. [26] in our study, our analyses with the modified 12S-V5 primers support the superior performance of the 12S-V5 region over MiFish region amplified with the MiFish primers [25] in detecting species [54–56]. However, our MiFish_mod primers which were modified to better capture Neotropical fish diversity detect approximately the same number of species as the 12S-V5_mod primers (Fig 6). These findings collectively underscore the importance of primer selection in maximizing detection rates and for improving biodiversity assessments in complex tropical ecosystems.
Importance of a Local database
The comparison between the global reference database—which has the Miya 12S region for 254 species and the Riaz 12S region for 324 species of Amazonian fishes, and the local reference database which has sequences of the Miya 12S and Riaz 12S region for 392 Amazonian fish species, with a focus on the Teles Pires River basin—showed that only two species were shared between them, highlighting the critical role of a local reference library in eDNA-based monitoring of Amazonian ichthyofauna. The Local database permitted the detection of 162 species, compared to the 109 species detected using only the Midori2 database. When both databases were combined, species detection increased substantially to a total of 265 species.
In a previous survey of the ichthyofauna of the Teles Pires River using traditional sampling methods, Ohara et al. [32] recorded 11 orders, 42 families, 198 genera, and 355 species of fishes, using a variety of fishing gear such as gillnets, purse seines, longlines, sieves. This survey was carried out over a five year period, four times a year (high, receding, low, rising waters) and at 20 sampling sites distributed along a 220 km long section of the lower portion of the main channel of the Teles Pires River and its tributaries. Similarly, Carvalho et al. [35] registered 354 species in 38 families and nine orders in the area of the present study based on four years of sampling at 15 sites, each with four sampling periods (high, receding, low, and rising waters) per year.
In contrast, for our eDNA study we sampled 22 sites along a non-overlapping 210 km stretch of the Teles Pires River in one year over the four water seasons, and we detected 10 orders, 40 families, 118 genera, and 265 species of fish. Of the 256 detected species, 108 were from the Teles Pires fishes represented in the Local reference library. For the remaining 65 species of Teles Pires fishes represented in the Local reference library, we detected 28 sister species instead. This represents 72.3% and 74.6% of the species diversity detected by Carvalho et al. [35] and Ohara et al. [32], respectively, when using both the Midori2 and Local reference databases.
The 72.3–74.6% species detection level is the conservative lower limit, since only 49% of the species registered by Carvalho et al. [35] and Ohara et al. [32] were included in the Local reference database. It is also possible that some of the species detected by Ohara et al. [32] may not occur in our sampled area due to regionalization of ichthyofauna within Amazonian River basins [57]. Furthermore, it is clear that no more than 50% of the ichthyofaunal diversity could be assigned to species level due to the deficiency of the reference database (Fig 3). Therefore, conservatively some 500 fish species are likely to occur in the area we surveyed using eDNA.
While 500 fish species may seem a large number, major Amazonian tributaries harbor anywhere from 502 to 1062 fish species [57]. The Tapajos River, which is formed by the confluence of the Juruena and Teles Pires rivers, has 522 registered species [57]. Additionally, a large number of species from the Amazon basin await description, and professionally conducted surveys almost invariably result in the discovery of new species, both previously described as well as new to science. For example, the multi-year ichthyological survey of the Madeira River prior to the construction of the Santo Antônio and Jirau hydroelectric dams resulted in a nearly 100% increase of species diversity—439 species listed in Torrente-Vilara [58] prior to the survey to more than 1000 species [57,59] following the survey.
The large number of detected species in our eDNA study is the consequence of the primers used, and the availability of a comprehensive local reference database, whose development was possible due to the availability of genetic resources obtained through prior ichthyofaunal monitoring efforts using conventional sampling techniques. As expected, the use of a region-specific reference database significantly improves both detection efficiency and taxonomic resolution of eDNA-based studies. A customized reference library enhances the resolution of taxonomic assignments and increases overall detection efficiency [60]. Our results strongly support the effectiveness of eDNA metabarcoding for biodiversity monitoring in tropical freshwater systems, especially when paired with a well-curated local database.
Numerous studies have emphasized the scarcity of comprehensive 12S reference libraries, which results in few species-level identifications in eDNA studies [61,62]. These and our study highlight the urgent need for expanded efforts to enrich reference sequence databases, thereby improving the performance and reliability of eDNA-based surveys. One practical starting point for building a local reference library is to consult regional fauna checklists for the target area. In the absence of such lists, researchers can compile expected species inventories using resources like the Eschmeyer’s Catalog of Fishes [40] or from species distribution models such as in Map of Life (https://mol.org/). Subsequently, available sequences can be retrieved from GenBank or other platforms [63], or tissue samples from zoological collections can be obtained to generate new reference sequences [64].
Additionally, conducting traditional sampling in parallel with eDNA surveys provides valuable reference material and helps validate molecular detections [56]. Such integrative approaches are essential for the development of robust and taxonomically comprehensive databases that support accurate biodiversity assessments in megadiverse ecosystems like the Amazon.
Understanding “false positive” detections
The power of eDNA studies comes from being able to detect species from very small quantities of environmental DNA [11]. This can lead to what appear to be false positive detections. Biologically diverse regions are relatively poorly sampled, and faunal inventories are often incipient [65,66]. Although two fish species lists have been published for our study region, these species lists are far from definite or complete. For example, in our eDNA study we detected Semaprochilodus insignis, and while no species of Semaprochilodus is listed neither in Ohara et al. [32] nor in Carvalho et al. [35], this genus occurs in the Tapajos River—which is formed by the confluence of the Teles Pires and Juruena rivers; and further more, Semaprochilodus insignis from the lower Teles Pires River—the region sampled by Ohara et al. [32]—is deposited in the ichthyological collection of INPA (cat # INPA-ICT 057619). The same situation applies to Pygocentrus nattereri (INPA-ICT 057620). Similarly, while no species of Retroculus is listed in Ohara et al. [32] or Carvalho et al. [35], species of this reophilic specialist genus occur in the Tapajos River, and have recently (2023-10-18 and 2023-10-15) been collected in the Juruena River (MUZUSP 130087, MUZUSP 130198), and therefore it is not surprising that a species of this genus would also occur in the Teles Pires River and that we would detect it. We also detected Mylossoma albiscopum, Potamorhina latior and other characins, all of which occur in the Tapajos River although they have not been recorded by traditional sampling methods in the Teles Pires River. Finally, we also detected 12 additional species which occur in other river basins and have sequences in the reference databases, while species of the same genus occurring in the Teles Pires River are not represented in the reference databases. If these species are closely related, we will be “detecting” species from the neighboring basins instead of species occurring in the Teles Pires River. A point in case, we detected Cichlasoma dimerus which occurs in the Paraguai River basin and is in the reference database, while a morphologically and genetically very similar species (Cichlasoma amazonarum) has been collected in the Teles Pires River (INPA-ICT 057623) but is not in the reference database. Therefore, although Cichlasoma dimerus is in our list of species “detected” by eDNA, the species that most likely occurs in our eDNA sample is Cichlasoma amazonarum.
These examples clearly illustrate the complexity of analyzing eDNA samples from biologically diverse regions plagued by Linnean and Wallacean shortfalls. In these regions, which are primarily in the Global South, species lists generated from eDNA need an additional level of interpretation that is predicated on a certain level of expert knowledge of the focal study group. However, it is exactly these regions where Linnean and Wallacean shortfalls are the most acute [67], and therefore, even with expert input, species lists generated via eDNA analysis will be incomplete, and may contain “false positives” if not subjected to secondary verifications by qualified professionals.
Conclusion
Environmental DNA studies are predicated on the specificity of primers, the taxonomic resolution of the amplified target region and reference databases. Studies focusing on DNA barcoding and early eDNA studies have proposed taxonomically informative target regions [25,26]. Subsequent studies investigated specificity of primers [18], recommending 12S and 16S primer sets. While specific, at least in the case of the standard 12S-V5 and MiFish primer sets, much of Neotropical fish diversity is not sampled since the primers do not incorporate priming site variation of the local ichthyofauna dominated by Characiform fishes. The standard 12S-V5 and MiFish primer sets were developed for Paleoarctic fauna in general (12S-V5) and Paleoarctic and Asian fishes (MiFish). Numerous groups of Neotropical fishes have base pair mismatches at the 3’ end of these standard primers, and thus do not amplify, or do not amplify efficiently components of Neotropical ichthyofaunal. Consequently this generates false absences in local fish communities. A final key component of any eDNA study is the reference database. The Midori2 database includes a wide range of taxa from multiple vertebrate classes, making it a valuable tool when the objective is to assess broader vertebrate biodiversity. In our analysis, three vertebrate classes were detected using Midori2: Actinopterygii, Aves, and Mammalia (see S1 Fig and S3 Table). Similar patterns were also reported by Ritter et al. [68], confirming that universal 12S primers—in our case the 12S-V5_mod primers, paired with broad databases can capture a wide range of vertebrate diversity. However, as emphasized earlier, the development and use of a Local reference database is critical for achieving accurate species-level identification—especially in highly diverse and understudied regions like the Amazon. When relying solely on Midori2, a high proportion of sequences are classified only at the Class level, due to the generic and incomplete nature of many entries (see Fig 3). In contrast, the combination of target-specific primers and a regionally curated database enables more precise taxonomic resolution, enhancing our ability to detect and monitor the rich Amazonian fish fauna with greater confidence and detail. This approach will be especially important for long-term monitoring and conservation of Amazonian aquatic biodiversity, which faces growing environmental pressures.
Supporting information
S1 Table. List of species were identified to species level and deposited at the Instituto Nacional de Pesquisas da Amazônia (INPA) and Coleção Ictiológica do Acervo Biológico da Amazônia Meridional (ABAM-I), at Laboratório de Ictiologia Tropical (UFMT).
https://doi.org/10.1371/journal.pone.0336758.s001
(XLSX)
S2 Table. Summary of Illumina reads and bioinformatic filtering.
https://doi.org/10.1371/journal.pone.0336758.s002
(XLSX)
S1 Fig. Stacked bar plots showing the relative read abundance of each primer pair in each Class.
Relative read abundance of A) Fish, B) Mammal, and C) Bird families.
https://doi.org/10.1371/journal.pone.0336758.s004
(TIFF)
Acknowledgments
We acknowledge Jaine Sousa and Gustavo Wolf from the LIT-ABAM lab for their support with eDNA sample collection.
References
- 1.
Sleen P van der, Albert JS. Field guide to the fishes of the Amazon, Orinoco & Guianas. Princeton, New Jersey: Princeton University Press. 2018.
- 2.
Edmondstone MRJ, Bull G, Nathaniel SRN, Chouly O. New Species 2024: The Freshwater Fish Species Described in 2024 (Report 4). SHOAL. 2025. https://shoalconservation.org/wp-content/uploads/2025/03/New-Species-2024.pdf
- 3. Marques V, Milhau T, Albouy C, Dejean T, Manel S, Mouillot D, et al. GAPeDNA: Assessing and mapping global species gaps in genetic databases for eDNA metabarcoding. Diversity and Distributions. 2021;27:1880–92.
- 4. Castello L, Macedo MN. Large-scale degradation of Amazonian freshwater ecosystems. Glob Chang Biol. 2016;22(3):990–1007. pmid:26700407
- 5. Fearnside PM. Environmental and Social Impacts of Hydroelectric Dams in Brazilian Amazonia: Implications for the Aluminum Industry. World Development. 2016;77:48–65.
- 6. Winemiller KO, McIntyre PB, Castello L, Fluet-Chouinard E, Giarrizzo T, Nam S, et al. Balancing hydropower and biodiversity in the Amazon, Congo, and Mekong. Science. 2016;351(6269):128–9.
- 7. Zaiatz APSR, Zolin CA, Vendrusculo LG, Lopes TR, Paulino J. Agricultural land use and cover change in the Cerrado/Amazon ecotone: A case study of the upper Teles Pires River basin. Acta Amaz. 2018;48(2):168–77.
- 8. Kraeski A, De Almeida FT, De Souza AP, De Carvalho TM, De Abreu DC, Hoshide AK. Land use changes in the Teles Pires River Basin’s Amazon and Cerrado biomes, Brazil, 1986–2020. Sustainability. 2023;15:4611.
- 9. Freire KMF, Tubino RA, Monteiro‐Neto C, Andrade‐Tubino MF, Belruss CG, Tomás ARG. Brazilian recreational fisheries: current status, challenges and future direction. Fish Management Ecology. 2016;23:276–90.
- 10. Oliveira E, Kasper D, da Silva SAA, Lázaro WL, Muniz CC, de Carvalho GS, et al. Cascade reservoirs affect mercury concentrations in fish from Teles Pires river, Brazilian Amazon. Ecotoxicology. 2025;34(3):444–55. pmid:39779615
- 11. Ficetola GF, Miaud C, Pompanon F, Taberlet P. Species detection using environmental DNA from water samples. Biol Lett. 2008;4(4):423–5. pmid:18400683
- 12. Taberlet P, Coissac E, Pompanon F, Brochmann C, Willerslev E. Towards next-generation biodiversity assessment using DNA metabarcoding. Mol Ecol. 2012;21(8):2045–50. pmid:22486824
- 13. Ray M, Umapathy G. Environmental DNA as a tool for biodiversity monitoring in aquatic ecosystems – a review. J Threat Taxa. 2022;14(5):21102–16.
- 14. Seymour M, Edwards FK, Cosby BJ, Bista I, Scarlett PM, Brailsford FL, et al. Environmental DNA provides higher resolution assessment of riverine biodiversity and ecosystem function via spatio-temporal nestedness and turnover partitioning. Commun Biol. 2021;4(1):512. pmid:33941836
- 15. Li Z, Li F, Qin S, Guo F, Wang S, Zhang Y. Environmental DNA biomonitoring reveals the human impacts on native and non-native fish communities in subtropical river systems. J Environ Manage. 2024;349:119595. pmid:37979384
- 16. Lawson Handley L, Read DS, Winfield IJ, Kimbell H, Johnson H, Li J, et al. Temporal and spatial variation in distribution of fish environmental DNA in England’s largest lake. Environmental DNA. 2019;1(1):26–39.
- 17. O’Donnell JL, Kelly RP, Lowell NC, Port JA. Indexed PCR Primers Induce Template-Specific Bias in Large-Scale DNA Sequencing Studies. PLoS One. 2016;11(3):e0148698. pmid:26950069
- 18. Collins RA, Bakker J, Wangensteen OS, Soto AZ, Corrigan L, Sims DW, et al. Non‐specific amplification compromises environmental DNA metabarcoding with COI. Methods Ecol Evol. 2019;10(11):1985–2001.
- 19. Zhang S, Zhao J, Yao M. A comprehensive and comparative evaluation of primers for metabarcoding eDNA from fish. Methods Ecol Evol. 2020;11(12):1609–25.
- 20. Yang J, Zhang L, Mu Y, Zhang X. Small changes make big progress: A more efficient eDNA monitoring method for freshwater fish. Environmental DNA. 2023;5(2):363–74.
- 21. van den Heuvel-Greve MJ, van den Brink AM, Glorius ST, de Groot GA, Laros I, Renaud PE, et al. Early detection of marine non-indigenous species on Svalbard by DNA metabarcoding of sediment. Polar Biol. 2021;44(4):653–65.
- 22. Schwentner M, Zahiri R, Yamamoto S, Husemann M, Kullmann B, Thiel R. eDNA as a tool for non-invasive monitoring of the fauna of a turbid, well-mixed system, the Elbe estuary in Germany. PLoS One. 2021;16(4):e0250452. pmid:33861810
- 23. He X, Jeffery NW, Stanley RRE, Hamilton LC, Rubidge EM, Abbott CL. eDNA metabarcoding enriches traditional trawl survey data for monitoring biodiversity in the marine environment. ICES Journal of Marine Science. 2023;80(5):1529–38.
- 24. Lines R, Juggernauth M, Peverley G, Keating J, Simpson T, Mousavi-Derazmahalleh M, et al. A large scale temporal and spatial environmental DNA biodiversity survey of marine vertebrates in Brazil following the Fundão tailings dam failure. Mar Environ Res. 2023;192:106239. pmid:37926039
- 25. Miya M, Sato Y, Fukunaga T, Sado T, Poulsen JY, Sato K, et al. MiFish, a set of universal PCR primers for metabarcoding environmental DNA from fishes: detection of more than 230 subtropical marine species. R Soc Open Sci. 2015;2(7):150088. pmid:26587265
- 26. Riaz T, Shehzad W, Viari A, Pompanon F, Taberlet P, Coissac E. ecoPrimers: inference of new DNA barcode markers from whole genome sequence analysis. Nucleic Acids Res. 2011;39(21):e145. pmid:21930509
- 27. Elbrecht V, Leese F. Can DNA-Based Ecosystem Assessments Quantify Species Abundance? Testing Primer Bias and Biomass--Sequence Relationships with an Innovative Metabarcoding Protocol. PLoS One. 2015;10(7):e0130324. pmid:26154168
- 28. de Santana CD, Parenti LR, Dillman CB, Coddington JA, Bastos DA, Baldwin CC, et al. The critical role of natural history museums in advancing eDNA for biodiversity studies: a case study with Amazonian fishes. Sci Rep. 2021;11(1):18159. pmid:34518574
- 29. Mathon L, Marques V, Mouillot D, Albouy C, Andrello M, Baletaud F, et al. Cross-ocean patterns and processes in fish biodiversity on coral reefs through the lens of eDNA metabarcoding. Proc Biol Sci. 2022;289(1973):20220162. pmid:35440210
- 30. Gómez-Buckley MC, Gallego R, Arranz V, Halafihi T, Stone K, Erdmann M, et al. Comparing anesthetic stations and environmental DNA sampling to determine community composition of cryptobenthic coral reef fishes of Vava’u, Kingdom of Tonga. Coral Reefs. 2023;42(3):785–97.
- 31. Corral-Lou A, Doadrio I. Environmental DNA metabarcoding of water samples as a tool for monitoring Iberian freshwater fish composition. PLoS One. 2023;18(10):e0283088. pmid:37903086
- 32. Ohara W, Lima FCT, Salvador GN, Andrade MC. Peixes do Rio Teles Pires: diversidade e guia de identificação. Gráfica Amazonas e Editora LTDA - EPP. 2017.
- 33.
Goulding M, Barthem RB, Ferreira EJG. The Smithsonian Atlas of the Amazon. Washington, DC: Smithsonian Institution Press. 2003.
- 34. Oberacker P, Stepper P, Bond DM, Höhn S, Focken J, Meyer V, et al. Bio-On-Magnetic-Beads (BOMB): Open platform for high-throughput nucleic acid extraction and manipulation. PLoS Biol. 2019;17(1):e3000107. pmid:30629605
- 35. Carvalho LN, Hashimoto S, Cabeceira FG, Carvalho FRD. The coleção ictiológica do acervo biológico da Amazônia Meridional, Laboratório de Ictiologia Tropical, Sinop, MT. Bol Soc Bras Ictiol. 2019;129:45–8.
- 36. Carvalho APC, Collins RA, Martínez JG, Farias IP, Hrbek T. From shallow to deep divergences: mixed messages from Amazon Basin cichlids. Hydrobiologia. 2018;832(1):317–29.
- 37. Ximenes AM, Bittencourt PS, Machado VN, Hrbek T, Farias IP. Mapping the hidden diversity of the Geophagus sensu stricto species group (Cichlidae: Geophagini) from the Amazon basin. PeerJ. 2021;9:e12443. pmid:34909270
- 38. Leray M, Knowlton N, Machida RJ. MIDORI2: A collection of quality controlled, preformatted, and regularly updated reference databases for taxonomic assignment of eukaryotic mitochondrial sequences. Environmental DNA. 2022;4(4):894–907.
- 39. Shen W, Ren H. TaxonKit: A practical and efficient NCBI taxonomy toolkit. J Genet Genomics. 2021;48(9):844–50. pmid:34001434
- 40.
Fricke R, Eschmeyer WN, van der Laan R. Eschmeyer’s Catalog of Fishes: Genera, Species, Reference. 2025. Available: http://researcharchive.calacademy.org/research/ichthyology/catalog/fishcatmain.asp
- 41.
Leach R, Parsons L. Barcode Splitter, version 0.18.6. Princeton, NJ, USA. 2019. Available: https://bitbucket.org/princeton_genomics/barcode_splitter
- 42. Martin M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal. 2011;17:10.
- 43. Zhang J, Kobert K, Flouri T, Stamatakis A. PEAR: a fast and accurate Illumina Paired-End reAd mergeR. Bioinformatics. 2014;30(5):614–20. pmid:24142950
- 44. Rognes T, Flouri T, Nichols B, Quince C, Mahé F. VSEARCH: a versatile open source tool for metagenomics. PeerJ. 2016;4:e2584. pmid:27781170
- 45. Mahé F, Rognes T, Quince C, de Vargas C, Dunthorn M. Swarm: robust and fast clustering method for amplicon-based studies. PeerJ. 2014;2:e593. pmid:25276506
- 46. Camacho C, Coulouris G, Avagyan V, Ma N, Papadopoulos J, Bealer K, et al. BLAST+: architecture and applications. BMC Bioinformatics. 2009;10:421. pmid:20003500
- 47.
R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. 2025.
- 48. Beatty DS, Deen E, Gross C, Stachowicz JJ. Northeast Pacific eelgrass fish communities characterized by environmental DNA represent local diversity and show habitat specificity. Environmental DNA. 2023;5(4):706–22.
- 49. Wickham H. ggplot2. Cham: Springer International Publishing. 2016.
- 50. Chen H, Boutros PC. VennDiagram: a package for the generation of highly-customizable Venn and Euler diagrams in R. BMC Bioinformatics. 2011;12:35. pmid:21269502
- 51. Hsieh TC, Ma KH, Chao A. iNEXT: an R package for rarefaction and extrapolation of species diversity (Hill numbers). Methods Ecol Evol. 2016;7(12):1451–6.
- 52. Jackman JM, Benvenuto C, Coscia I, Oliveira Carvalho C, Ready JS, Boubli JP, et al. eDNA in a bottleneck: Obstacles to fish metabarcoding studies in megadiverse freshwater systems. Environmental DNA. 2021;3(4):837–49.
- 53. Batista LM, De Sá-Leitão CS, De Souza ÉMS, Dos Anjos-Santos CH, De Almeida-Val VMF. Addressing Amazonian fish diversity using environmental DNA (eDNA): a first glance. Eur J Aquat Sci. 2022;1:9–17.
- 54. Kumar G, Reaume AM, Farrell E, Gaither MR. Comparing eDNA metabarcoding primers for assessing fish communities in a biodiverse estuary. PLoS One. 2022;17(6):e0266720. pmid:35714082
- 55. How CM, Ip JC, Deconinck D, Zhao M, Yan M, Cheng J, et al. Refining Sampling Efforts for Fish Diversity Assessment in Subtropical Urban Estuarine and Oceanic Waters Using Environmental DNA With Multiple Primers. Environmental DNA. 2024;6(5).
- 56. Oliveira Carvalho C, Pazirgiannidi M, Ravelomanana T, Andriambelomanana F, Schrøder-Nielsen A, Stuart Ready J, et al. Multi-method survey rediscovers critically endangered species and strengthens Madagascar’s freshwater fish conservation. Sci Rep. 2024;14(1):20427. pmid:39227484
- 57. Dagosta FCP, de Pinna MCC. The fishes of the Amazon: Distribution and biogeographical patterns, with a comprehensive list of species. Bull Am Mus Nat Hist. 2019;431:1–168.
- 58.
Torrente-Vilara G. Heterogeneidade ambiental e diversidade ictiofaunística do trecho de corredeiras do rio Madeira, Rondônia, Brasil. Instituto Nacional de Pesquisas da Amazônia. 2009.
- 59.
Queiroz LJ de, Torrente-Vilara G, Ohara WM, Pires THS, Zuanon J, Doria CRC. Peixes do rio Madeira. São Paulo, SP, Brazil: Editora Dialeto. 2013.
- 60. Valdivia-Carrillo T, Rocha-Olivares A, Reyes-Bonilla H, Domínguez-Contreras JF, Munguia-Vega A. Integrating eDNA metabarcoding and simultaneous underwater visual surveys to describe complex fish communities in a marine biodiversity hotspot. Mol Ecol Resour. 2021;21(5):1558–74. pmid:33683812
- 61. Antil S, Abraham JS, Sripoorna S, Maurya S, Dagar J, Makhija S, et al. DNA barcoding, an effective tool for species identification: a review. Mol Biol Rep. 2023;50(1):761–75. pmid:36308581
- 62. Blackman R, Couton M, Keck F, Kirschner D, Carraro L, Cereghetti E, et al. Environmental DNA: The next chapter. Mol Ecol. 2024;33(11):e17355. pmid:38624076
- 63. Jeunen G-J, Dowle E, Edgecombe J, von Ammon U, Gemmell NJ, Cross H. crabs-A software program to generate curated reference databases for metabarcoding sequencing data. Mol Ecol Resour. 2023;23(3):725–38. pmid:36437603
- 64. Juhel J, Marques V, Utama RS, Vimono IB, Sugeha HY, Kadarusman K, et al. Estimating the extended and hidden species diversity from environmental DNA in hyper‐diverse regions. Ecography. 2022;2022(10).
- 65. Almeida TC de, Tessarolo G, Nabout JC, Teresa FB. Non‐stationary drivers on fish sampling efforts in Brazilian freshwaters. Diversity and Distributions. 2021;27:1224–34.
- 66. Whittaker RJ, Araújo MB, Jepson P, Ladle RJ, Watson JEM, Willis KJ. Conservation biogeography: assessment and prospect. Diversity and Distributions. 2005;11:3–23.
- 67. Guedes JJM, Moura MR, Jardim L, Diniz-Filho JAF. Global Patterns of Taxonomic Uncertainty and its Impacts on Biodiversity Research. Syst Biol. 2025;74(5):834–49. pmid:39953959
- 68. Ritter CD, Pont GD, Stica PV, Horodesky A, Cozer N, Netto OSM. Wanted not, wasted not: Searching for non-target taxa in environmental DNA metabarcoding by-catch. Environ Adv. 2022;7:100169.