Management of marine fisheries requires a large amount of data collection. Such data include monitoring of fish stock abundance, biomass, and several life history parameters. Fisheries-independent research surveys based on capture or visual census of the fish species of interest provide a significant proportion of the data to assess commercial fish stocks. These surveys are complemented by studies determining maturity, fecundity, age and sex structure of the stocks. Moreover, fisheries-dependent data are also collected to monitor in situ fisheries accordingly to the authority’s regulation (Bradley et al., 2019; Urban et al., 2023). Altogether, these methods give essential information to assess the status of fisheries resources and to provide scientific advice on the best management strategies to achieve a sustainable exploitation.

At present, the effects of climate change (Pinsky et al., 2020) and the detected loss of marine biodiversity (Payne et al., 2016) impose new levels of complexities in the assessments of the sustainability of marine fisheries. Climate change is altering fish species distributions and likely accelerating adaptation events, potentially impacting species abundances and ecosystems structure (Perry et al., 2005; Gallego et al., 2020; Pinsky et al., 2020; Palacios-Abrantes et al., 2022). Moreover, the inception of ecosystem-based fisheries management (EBFM, (Pikitch et al., 2004)), highlights as crucial the evaluation of the effect of fishing operations on marine ecosystems. EBFM should ensure long-term productive, healthy, and resilient ecosystems. Thus, it is of maximum priority to characterize and monitor the ecosystem status at places where fishing operates, especially in ecosystems defined as vulnerable by FAO (2009). Hence, data collection to ensure sustainable management of fisheries is scaling up in complexity requiring specialized training and high budgets.

As fisheries science is urgently needing to collect an increasing amount of data to assess the long-term sustainability of fish resources, environmental DNA approaches (eDNA, Taberlet et al., 2012) have become a promising technology in simplifying survey tasks. Hence, studies using eDNA to investigate marine species richness, distribution, and abundance expanded in the very last years (reviewed Gilbey et al., 2021; Jo et al., 2022; Miya, 2022; Ramírez-Amaro et al., 2022; Rourke et al., 2022). eDNA refers to DNA that can be extracted from environmental samples (such as soil, water, or air), without first isolating any target organisms. It is characterized by a complex mixture of genomic DNA from many different organisms and by possible degradation (Taberlet et al., 2012). Environmental DNA can be classified into two types, organismal DNA, and extra-organismal DNA. The source of organismal DNA is from whole individuals most probably alive at the time of sampling, whereas extra-organismal DNA originates from a variety of sources such as part of tissue replacement, metabolic waste, gametes, etc. (Rodriguez-Ezpeleta et al., 2021). Thus, eDNA samples are composed of a complex mixture of both types of DNA (i.e., organismal and extra-organismal) from various sources and in varying proportions (Taberlet et al., 2012). This heterogeneity of eDNA samples provides for raw data to potentially answer wide-rank of scientific questions, though it also brings uncertainties at the time of interpreting and comparing results across studies.

Coupled with sequencing and laboratory technologies, the analysis of eDNA samples can be used to identify species and to characterize species communities associated with the environment from where the DNA was extracted. Basically, eDNA approaches are based on the extraction, amplification, and quantification of the DNA present in environmental samples. eDNA approaches can be divided into two main groups: (1) eDNA metabarcoding or community approach, and (2) species-specific or targeted approach (Seymour, 2019; Tsuji et al., 2019). Both strategies share the key characteristic of non-invasive indirect sampling but differ in their purpose, methodology, and interpretations. The current state of the art of sequencing and quantification technologies makes both eDNA approaches very powerful in helping sustainable marine fisheries surveys and monitoring (Gilbey et al., 2021; Ramírez-Amaro et al., 2022; Rourke et al., 2022). On the one hand, the metabarcoding or community eDNA approach is intended to characterize the species community living in a given environment (Alberdi et al., 2017). Therefore, it can greatly facilitate environmental monitoring surveys by providing data on species assemblages and changes in ecosystems where fishing operates. On the other hand, the targeted species-specific approach can inform on population dynamics such as geographical displacements and provide for quantitative estimates of abundance. In addition, intermediate approaches between targeted and metabarcoding eDNA approaches, are also being designed and implemented in aquatic ecosystems (e.g. Wilcox et al., 2020). Nevertheless, there are gaps in the knowledge of eDNA dynamics in the sea, which should be accounted for in a comprehensive implementation of eDNA approaches in fisheries data surveys.

Since the sea is a highly variable environment, and applications of eDNA cover a wide spectrum of scientific questions (Barnes & Turner, 2016), it is noted disparity among results of eDNA analyses. Such incongruencies have raised concerns on the potential applicability of eDNA surveys to help fisheries sustainable management (Hansen et al., 2018; Ramírez-Amaro et al., 2022). It has been highlighted that part of the disparate results found among studies could be explained by the heterogeneity across pipelines and bioinformatics tools used for data processing and analyses (Creedy et al., 2022). eDNA data analyses entail the knowledge of the analyzing algorithms and prior assumptions made by these. For example, the choice of the algorithm to classify taxonomically the composition of an environmental DNA sample, as well as the completeness of the reference database used for it, can have a significant impact on diversity estimates and should be made according to the goals of the study (Mathon et al., 2021, Liu and Zhang, 2021, Miya, 2022). Additionally, it is imperative to understand what are the limitations of each particular eDNA study with respect to the power of detection (Burian et al., 2021), especially when reporting the presence/absence or abundance of target species.

Several software, pipelines, and scripts exist to manage and analyze the thousands of millions of sequences obtained in a typical eDNA study. User-friendly bioinformatics tools for the analysis of eDNA allow for obtaining results without bioinformatics or computational knowledge. Nevertheless, understanding of every step of the analytical pipeline is relevant for the reliable and efficient application of eDNA approaches. When correctly applied, algorithms and models included within eDNA data analysis pipelines allow infer and, many times solve common errors related to eDNA approaches. Therefore, to know the state of the art of data analysis for eDNA approaches is of great importance before planning an eDNA study. Thus, the aim of this review is to present and explain the main analytical approaches used in eDNA studies towards to facilitate its application in helping fisheries’ sustainable management. This review expands and complements the review made by Ramírez-Amaro et al. (2022) included in this Research Topic Issue. Nevertheless, whereas Ramírez-Amaro et al. (2022) center on sampling and laboratory steps, here we focus on the next step: the analyses of the eDNA data.

Several tasks related to fisheries sustainable management can be supported by eDNA approaches: (1) Surveillance of rare (i.e. endangered) and invasive species, (2) Describing and monitoring Vulnerable Marine Ecosystems, (3) Monitoring fish assemblages, and (4) Quantifying biomass and/or abundance of target economically important species. Case-study examples of the applicability of both eDNA approaches to these four tasks are presented in BOX 1 . eDNA metabarcoding can be used to describe species richness (e.g. in Leray and Knowlton, 2015; McClenaghan et al., 2020; Boulanger et al., 2021; Liu and Zhang, 2021; Sato et al., 2021; Yu et al., 2021; Fonseca et al., 2022; Good et al., 2022; Keck et al., 2022), surveillance of the presence/absence of target species such as invasive, or endangered (e.g. in Jerde et al., 2011; Boussarie et al., 2018; Holman et al., 2019; Bonfil et al., 2021; Manfrin et al., 2022), and monitor changes in species composition (e.g. in Stat et al., 2017; Jeunen et al., 2019; Gallego et al., 2020; Afzali et al., 2021; Gold et al., 2021; Russo et al., 2021; Stoeckle et al., 2021; Valdivia-Carrillo et al., 2021; West et al., 2021; Maiello et al., 2022). eDNA targeted approach is better suited for quantification of the abundance of target species (e.g. in Takahara et al., 2012; Lacoursière-Roussel et al., 2016; Yamamoto et al., 2016; Knudsen et al., 2019; Salter et al., 2019; Brys et al., 2021; Fukaya et al., 2021; Shelton et al., 2022; Urban et al., 2023), and therefore it could be applied in monitoring fish stocks and vulnerable species.

Data processing for eDNA metabarcoding requires several steps which include decision-making on the best algorithms to be used accordingly to the previous knowledge of the species community and the goal of the study, while for eDNA targeted approach it is simplified. Available, open-access bioinformatics tools with potential use for eDNA approaches in fisheries are listed in Supplementary Table 1 .

Metabarcoding refers to the amplification and sequencing of a DNA region, known as barcoding, simultaneously for a species community or taxonomic group. Most of the widely used barcoding regions belong to the mitochondrial genome. It is because mitochondria are present in multiple copies within cells and therefore are most likely to be amplified from environmental samples which contain low DNA content by species. PCR products are then sequenced by High Throughput Sequencing (HTS), commonly in Illumina platforms, and the obtained sequences are subsequently clustered by similarity, and then taxonomically classified.

The selection of the barcoding region to be obtained from an eDNA sample depends on the goal of the study as well as on the group of species intended to be surveyed. Universal primers for barcoding regions are designed to amplify taxonomically related species. The taxonomical assignment is performed by sequence identity searches against reference sequence databases. The quality of the results of the metabarcoding approach depends on the correct taxonomical assignments of the species present in an eDNA sample. Therefore, the correct selection of the barcoding region to be used and the completeness of the reference database are the keys, and also the main limitations of the metabarcoding approach (Miya, 2022).

Bioinformatics pipelines are the result of executing linked instructions to process a large amount of sequence data (often several hundred million reads) using scripts, software, and databases. Computational requirements for an eDNA metabarcoding data analysis will depend on the sampling magnitude, number of replicates, and goals of the study, though hundreds of gigabytes of data are expected from metabarcoding sequencing. Data analyses are mostly performed on Unix platforms, and in high-performance computing systems. However, pipelines such as PEMA (Zafeiropoulos et al., 2020) can be installed from regular computers to cloud or HPC environments. Processing times of different pipelines have been compared in Mathon et al., 2021. Requirements of any particular pipeline used in eDNA metabarcoding can be found in the source webpage listed in Supplementary Table 1 .

Data analysis protocols should be adapted to the survey design and the ecological question under study (Alberdi et al., 2017; Zinger et al., 2019). However, independently of the particularities of each study, the metabarcoding data analyses pipeline has a number of common steps that are essential and summarized in Figure 1 .

Because the first application of eDNA metabarcoding was for the study of microorganisms, there are very well-established packages for the study of microbial communities based on the 16S metabarcoding region, such as MOTHUR (Schloss et al., 2009), Qiime2 (Bolyen et al., 2019), USEARCH (Edgar, 2010), and RDP pipeline (Cole et al., 2014). However, the extended use of eDNA in conservation and marine sciences has produced an increasing number of additional bioinformatics tools as well as reference databases for the analysis of eukaryote taxonomic groups. Therefore, the list of tools provided in Supplementary Table 1 does not intend to conduct a systematic review of all existing metabarcoding pipelines. Instead, we focus on those that have been proven useful for the study of marine biodiversity and fish communities, are open access, and are well-documented.

To date, for the study of fish communities, MitoFish, MitoAnnotator, and MiFish pipelines constitute a key platform (Sato et al., 2018), OBITools (Boyer et al., 2016) and ANACAPA (Curd et al., 2019) pipelines have been tested offering reliable results for fish communities, while mBrave (Ratnasingham, 2019) linked to BOLD system platform (Ratnasingham and Hebert, 2007), and Meta-Fish-Lib (Collins et al., 2021) are powerful tools to increase the knowledge in fish biodiversity surveys. Most of the pipelines presented in Supplementary Table 1 do not perform all steps needed to perform a complete eDNA metabarcoding analysis. However, most of them are modular allowing users to run different modules separately and choose better options according to their case study. Independently of the analytical protocol, the steps described below are essential in any eDNA metabarcoding analyses.

The second category of eDNA approaches involves species-specific techniques that use assays tailored to target DNA fragments of particular species in an environmental sample. When designed stringently and after thorough validation, these are highly reliable and often effectively linked to the biomass and abundance of the target organism (e.g. in Takahara et al., 2012; Doi et al., 2015; Lacoursière-Roussel et al., 2016; Yamamoto et al., 2016; Capo et al., 2019; Knudsen et al., 2019; Salter et al., 2019; LeBlanc et al., 2020; Brys et al., 2021; Fukaya et al., 2021; Shelton et al., 2022; Kasmi et al., 2023; Urban et al., 2023). Nowadays, the main technique used in species-specific detection of environmental samples is real-time quantitative PCR (qPCR). However, other PCR relate-techniques such as digital PCR can also be used. The workflow of data analyses for the targeted eDNA approach is presented in Figure 2 .

When assessments of species distribution are carried out, it is imperative to understand what are the limitations of the survey methods used. Similarly to all survey methods, detection (or lack of detection) of species by eDNA sampling is not free of errors (see Burian et al., 2021 for a review of potential sources of error of eDNA-based methods). A particular species might not be detected even if it is present in the environment (i.e. false negative), or it can be erroneously detected even when absent (i.e. false positive). Therefore, it is most important to understand the level of error that the particular eDNA survey presents, to make sound and consistent interpretations.

As commented in section 3.3.3 several bioinformatics tools help to filter out false positive errors in eDNA metabarcoding studies. However, it is important to consider the rate of false negative results too, especially when the goal of the study is to monitor a group of target species (i.e. by metabarcoding) or single target species. False negative detection can be produced because enzymatic inhibition, or eDNA for a species was not collected within the sample, or it is in such a low concentration that became undetectable or because of the sensitivity of PCR (Ficetola et al., 2015). To avoid missing detection of taxa that are actually present (false negatives), multiple extractions and amplifications of the same samples are often performed (Ficetola et al., 2015; McClenaghan et al., 2020). Moreover, increasing sequencing depth can also improve the rates of detection (McClenaghan et al., 2020; Fukaya et al., 2021).

Currently, it is becoming common and advisable to interpret surveys based on eDNA within an occupancy modelling framework. Occupancy modelling allows estimating the probability of detection of a given species in a given environment (Griffin et al., 2020). Occupancy modelling can be applied to eDNA metabarcoding (i.e. Doi et al., 2019; McClenaghan et al., 2020, Valdivia-Carrillo et al., 2021), and in targeted eDNA approach (Dorazio and Erickson, 2018, Fukaya et al., 2021; Buxton et al., 2022). Appropriate data for fitting an Occupancy Site model consist of a series of samples obtained from S sampling sites, with K replicate samples obtained per site (Dorazio and Erickson, 2018). Models have been developed allowing for multiscale occupancy models to be applied accounting for both positive and negative errors at the field and laboratory stages (Griffin et al., 2020). Occupancy models successfully estimated true prevalence, detection probability, and false-positive rates, and their performance increased with the number of replicates (Ficetola et al., 2015). Thus, the occupancy modelling framework can be also applied to study the optimal conditions for sampling and laboratory stages (Doi et al., 2019). Modelling is frequently performed by Bayesian inference, and modeled parameters can include environmental and experimental covariates (Guillera-Arroita et al., 2017; McClenaghan et al., 2020; Burian et al., 2021; Buxton et al., 2022).

Other origins of uncertainty in the results of surveys of eDNA, especially for quantitative results, come from the dynamics of concentrations of eDNA at the sampling places (Burian et al., 2021). The amount of DNA shed into the environment by a particular species can be different from another, the biomass of an adult fish can be the same as hundred juvenile individuals of the same species; while environmental characteristics can make DNA rapidly degraded or transported by oceanic currents (Hansen et al., 2018). Process-based models are based on a mechanistic understanding of the dynamics of eDNA concentrations in the environment and can help in accounting for these sources of false positive and false negative results (Burian et al., 2021). eDNA studies have been favourably augmented in precision when accounting for eDNA transport shed and decay (e.g. Collins et al., 2021, Murakami et al., 2019; Kirtane et al., 2021; Urban et al., 2023). The experimental output, informs on the expected rates of detection or lack of it in the eDNA survey and can be integrated as prior information within the occupancy modelling framework.

Occupancy modelling can be performed using the eDNA SHINY APP (Diana et al., 2021), or the R script EDNAOCCUPANCY (Dorazio and Erickson, 2018).

As sequencing and quantitative technologies grow, the applications of eDNA approaches also do. To date, automatic sequencing robots are already available to process eDNA in situ within aquatic environments (Sepulveda et al., 2020), and long-read sequencing of eDNA enhances taxonomic assignments opening also the possibility to estimate population genetics parameters at the same time (Sigsgaard et al., 2020; Tsuji et al., 2020). Bioinformatics tools to analyze the different data from eDNA sampling are growing along with the sequencing technologies, and are not a limiting issue in the application of eDNA approaches. The increasing number of eDNA studies published in the very last few years reflects both the simplicity of the method to potentially answer different scientific questions, and the need to simplify costly surveys traditionally used in ecological sciences. However, this diversity of studies with different analytical approaches makes it difficult to navigate the bibliography for a systematic implementation of eDNA approaches in the sustainable management of fisheries. Here, we show the state of the art of data analyses with the available bioinformatics tools proven to be useful for avoiding common mistakes in processing, analyzing, and interpreting results based on eDNA samples.

Across different studies referenced here, it seems clear that both eDNA approaches (metabarcoding and targeted) suffer uncertainty regarding detection probabilities. This uncertainty can be at least partially improved by: (i) designing eDNA sampling with enough numbers of sample replications, and sampling sites, and covering seasonal variance; (ii) getting PCR and amplicon sequencing replications, (iii) getting enough amplicon coverage, and (iv) by using occupancy and process-based modelling. This is particularly important in the application of eDNA approaches to surveillance of rare species (e.g. endangered), which are prone to suffer from uncertainties for the low probability of detection (Ficetola et al., 2015, but see Boussarie et al., 2018 in BOX 1-TASK 1 ). Major complexities are found in the application of eDNA approaches to surveillance of Marine Vulnerable Ecosystems, as many of them are poorly described, and therefore, reference databases are incomplete (e.g. Good et al., 2022, see BOX 1-TASK 2 ). In this regard, complementing and comparing eDNA sampling with traditional surveys such as visual census and video recording is necessary (e.g. Boussarie et al., 2018; Good et al., 2022). Nevertheless, the molecular characterization of the relevant invertebrate species of Vulnerable Ecosystems will be necessary for reference databases to be complete before monitoring can be systematically implemented. On this point, the use of different primer sets and different genomics regions gives more complete results (Good et al., 2022; Kumar et al., 2022). However, the use of eDNA approaches are giving promising results when monitoring Marine Protected Areas (MPA) (eg. Gold et al., 2021). Ecosystems within MPA are generally well described making the implementation of eDNA approaches augmented by prior knowledge. Similarly, barcoding regions have been obtained for thousands of fish species, and especially for economically important ones making that monitoring of fish communities does not present obvious methodological or analytical difficulties. Thus, the research on the limitations of the eDNA approaches for the study of target fish species (i.e. commercial) and fish communities are less challenging to be accomplished (Keck et al., 2022). Reference databases enriched with sequences of fish species locally known increase the likelihood of detection (e.g. Stoeckle et al., 2021, see BOX 1- TASK 3 ) and it is advisable when monitoring fish communities.Moreover, uncertainties about detection probabilities are decreased when DNA studies are performed in fisheries-related conditions (e.g. Russo et al., 2021; Stoeckle et al., 2021; Maiello et al., 2022; Kasmi et al., 2023, and Urban et al., 2023 presented in BOX 1- TASK 4 ).

In summary, eDNA approaches are showing realistic results for tasks related to fish abundance quantification and monitoring. Nonetheless, its use in ecosystem-based fisheries management could be delayed until reference databases of vulnerable marine ecosystems are completed. Several studies referenced here have demonstrated the feasibility of applying eDNA approaches to support tasks related to sustainable fisheries management. Given the urgent need of promoting sustainability in fisheries management, the extensive amount of data collection required for this, and the rapid scientific advances improving eDNA data analysis, it is likely that eDNA approaches can be incorporated within fisheries-related tasks in the nearby future.

The information and views set out in this publication are those of the authors and do not necessarily reflect the official opinion of CINEA or of the European Commission. Neither CINEA nor the European Commission can guarantee the accuracy of the scientific data/information collected under the above Specific Contract or the data/information included in this publication. Neither CINEA nor the European Commission or any person acting on their behalf may be held responsible for the use which may be made of the information contained therein.

NP-M, LC and FS-R wrote, edited and reviewed the manuscript. All authors approved the submitted version.