Interim report on the performance of eDNA-based approaches and associated optimal diversity indices for biodiversity observations

Deliverable 2.1 Interim report on the performance of eDNA-based approaches and associated optimal diversity indices for biodiversity observations Version 1.0 19 Junel 2024 Aubrie Onoufriou 1 Iveta Matejusova 1 Daniel Kumazawa Morais 2 1 Marine Directorate, Scottish Government 2 UiT - The Arctic University of Norway MARK IF PUBLIC / SENSITIVE / CLASSIFIED Ref. Ares(2024)4567014 - 25/06/2024 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches Document Information Grant Agreement 101082021 Project Acronym MARCO-BOLO Project Title MARine COastal BiOdiversity Long-term Observations (MARCO-BOLO): Strengthening biodiversity observation in support of decision making Deliverable Number D2.1 Work Package Number WP2 Deliverable Title Interim report on the performance of eDNA-based approaches and associated optimal diversity indices for biodiversity observations Lead Beneficiary 26. MS Author(s) Aubrie Onoufriou (MS), Iveta Matejusova (MS), Daniel Kumazawa Morais (UiT) Due Date 17.05.2024 Submission Date 19.06.2024 Dissemination Level PU - Public 1 Type of Deliverable R - Document, report 2 Version N.N DD.MM.YYYY, Author Version N+1.+1 DD.MM.YYYY, Author Version N+2.N+2 DD.MM.YYYY, Author 2 Type of deliverable: R: Document, Report, DEM: Demonstration, pilot, prototype, DEC: Website, patent filing videos, DMP: Data Management Plan, Ethics: Ethics deliverable 1 Dissemination level : PU: Public, SEN: Sensitive, CL: EU Classified, information as referred to in European Commission Decision 2015/844 2 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches Contents Document Information 2 1. Objectives 6 2. Meta-Analysis 6 2.1 qPCR vs Abundance 6 2.2 Metabarcoding vs Abundance 6 2.3 Metabarcoding vs Species Richness 6 3. eDNA suitability for calculating EBVs 6 4. Statistical tools and analyses for eDNA data 6 1. Objectives Work Package 2 aims “to enable eDNA-based approaches (i.e. single-species detection, DNA metabarcoding and metagenomics) for biodiversity monitoring across trophic and functional groups (i.e. from microbes to vertebrates) and in marine, terrestrial and freshwater systems.” This interim report specifically addresses the progress of Task 2.1, “Meta-analysis of technical and practical aspects of current eDNA approaches.” This task consists of three components. In the first, meta-analyses will assess the performance of eDNA approaches compared to traditional methods. Specifically, eDNA-based methods to estimate species abundance will be evaluated using quantitative PCR (qPCR and ddPCR) in meta-analysis 1 and metabarcoding in meta-analyses 2, and species composition will be evaluated using metabarcoding in meta-analysis 3. In the second component, we will review the suitability of eDNA to calculate Essential Biodiversity Variables (EBVs) and Essential Ocean Variables (EOVs) . In the final component, we will review the diversity metrics and statistical tools that comply with eDNA data. 2. Meta-Analysis The prevalence of ecological studies using eDNA-based tools has risen dramatically, providing an opportunity to investigate the performance of eDNA compared to traditional methods in a statistically robust way. Meta-analyses provide a statistical framework where the results from multiple studies investigating the same question are pooled together to provide a consensus conclusion. The most common application of eDNA is to detect the presence of a species of interest in an aquatic environment. Recent publications however, have also used eDNA to quantify the abundance, biomass and composition of species in both terrestrial and aquatic environments. In response to these recent advances, Task 2.1 aims to assess the performance of eDNA-based approaches compared to more traditional techniques using meta-analyses. 3 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches To this end, we have defined following research questions: 1. How well can eDNA (using qPCR or ddPCR) quantify the abundance or biomass of a species compared to traditional methods? 2. How well can eDNA (using metabarcoding) quantify the abundance or biomass of a species compared to traditional methods? 3. How well can eDNA estimate species richness compared to traditional methods? Meta-analyses first require an exhaustive, systematic review of the relevant literature using relevant search terms for each question being asked. Here, The Lens (https://www.lens.org/) , a free and open-access literature search engine, was used. For the three meta-analyses, two literature searches were used, one for the abundance/biomass questions and another for species richness. Search terms were intentionally broad, to ensure the widest net could capture all relevant studies (see Appendix I for specific search terms). Resulting publications were exported to a reference manager (Zotero) and article titles and publishers were screened for inclusion based on set criteria (must be published in a peer-reviewed journal, in english, and no conference proceedings). For the publications that passed the initial screening, full article PDFs were downloaded and all abstracts were screened for relevant studies. Each paper in the resulting collection of relevant, peer-reviewed journal articles was further screened for utility in each of the three meta-analyses, retaining studies that provided the required matched eDNA and traditional data. Data and meta-data were extracted from each article (Appendix II), and justification was documented for rejecting an article. In some cases, the necessary data was not presented either in the main body text of an article or supplemental materials, and paper authors were contacted directly. The extracted data were imported into the R programming software, where the package ‘metafor’ was used to conduct multi-level, random effects models. The basic data metric within a meta-analysis is the effect size, the nature of which will change depending on the question being asked. In the two abundance-related meta-analyses, the Pearson correlation coefficient (‘ r ’) was selected. This metric ranges from -1 to 1 and reflects the amount of covariation between two continuous variables and since the calculation standardizes the value by their standard deviations, the scale of the original metrics that are being compared do not need to be the same. Prior to inclusion in the meta-analysis, it is standard protocol to transform the ‘ r ’ values into Fisher’s ‘z’ values, to stabilize the variance when the correlation is close to 1 or -1. To compare species richness, two different effect sizes will be used due to differences in reporting within the individuals studies. The log risk ratio takes into account the number of species detected using only eDNA and traditional methods, and those that were shared between the two. The standardized mean difference takes into account the species detected across sites (average and standard deviation) using eDNA and traditional methods. Effect sizes were selected for their suitability to each data type as well as their use in published meta-analyses using eDNA data. The outcome of a meta-analysis is a single value of the overall effect with some measurement of uncertainty around the value. This result can be visualised using various plots, and can be further explored by including relevant covariates (‘moderators’) to the model. 4 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches 2.1 Meta-analysis 1: Quantification: qPCR/ddPCR For the quantification meta-analyses, the literature search yielded n =1095 publications. Following screening, n= 106 publications (including 335 experiments) contained data suitable to meta-analyse. These were divided between studies that estimated numerical abundance ( n = 154 experiments) and biomass ( n= 144 experiments). From each experiment, the DNA concentration (as estimated with qPCR or ddPCR), it’s paired abundance or biomass estimate calculated using a traditional metric and the sample size were extracted. If provided in the publication, the Pearson correlation coefficient ‘r’ was also recorded, but in most cases ‘r’ was calculated from the data provided in supplemental materials using the “correl” function in Microsoft Excel. The effect size was calculated from ‘r’ values and sample sizes using the “escalc” function in the R package ‘metafor’. A multi-level model with random effects was selected to account for the hierarchical nature of the data (multiple experiments nested within a study) and the effect sizes were weighted by the sample size of each experiment. Overall, there was a significant and positive correlation between abundance calculated using qPCR or ddPCR and traditional methods ( r= 0.458, 95% CI: 0.135-0.693; Figure 1). Due to residual heterogeneity in the data, covariates (‘moderators’) were included in the model. Based on moderators that have been used in other eDNA-based meta-analyses and considering the nature of the dataset itself, we chose the following to include in the model: taxa, molecular method (qPCR or ddPCR) and environment (field study vs laboratory setting). A significant omnibus test of moderators ( indicated that at least some 𝑄𝑀 13 = 60 . 5 , 𝑝 < 0 . 0001 ) moderators were affecting the overall effect size, and we were justified in including them in the analysis. To visualise the impact that the moderators had on the effect size, orchard plots were generated for each of the three moderators. In Figure 2, the correlation between eDNA and traditional data is higher in a laboratory setting compared to field settings, though the 95% confidence intervals overlap. When comparing the molecular method, studies using ddPCR had higher correlation values, though the confidence intervals also overlapped (Figure 3). Finally, when the correlation values are plotted by taxa, it’s clear that only studies that quantified raphidophytes, fish and amphibians had significant correlations (95% CI didn’t span 0; Figure 4). The data for studies estimating biomass have been extracted, and the meta-analysis will be completed by August 2024. 5 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches Figure 1. Orchard plot showing the effect sizes from each study that estimated abundance using qPCR or ddPCR. Each blue dot represents the experiment-specific effect size and the size of each dot represents the precision. The pooled correlation coefficient (r=0.458) is presented with the 95% confidence intervals in heavy black lines. Figure 2. Orchard plot showing the effect sizes from each study that estimated abundance using qPCR and ddPCR grouped by each experimental setting moderator (Lab and Field). 6 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches Figure 3. Orchard plot showing the effect size from each study that estimated abundance using qPCR and ddPCR grouped by each molecular method (qPCR and ddPCR). Figure 4. Orchard plot showing the effect size from each study that estimated abundance using qPCR and ddPCR grouped by each taxon. 7 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches 2.2 Meta-analysis 2: Quantification: Metabarcoding From the same n=1095 publications that were screened in the previous section, n= 122 were screened for appropriate metabarcoding data. In total, n=29 publications contained adequate data from 362 experiments. For each study, either the number of metabarcoding reads or relative read abundance were extracted. The actual or relative abundance or biomass of each species calculated using a traditional method was also extracted. In some studies, the ‘r’ value was presented, but for many it was calculated using the “correl” function in Excel. Most of the data from these studies have been extracted and coded, and the meta-analysis will be completed by August 2024. 2.3 Meta-analysis 3: Species Richness The literature search resulted in n= 328 publications. After filtering out papers that didn’t meet the basic criteria ( n= 205 remaining), screening and data extraction are ongoing for the n= 165 publications that are left. As of 21/05/2024, data has been extracted from n= 23 publications ( n =38 experiments). A further n =80 papers will be screened by July 2024 and the meta-analysis will be completed by August 2024. 3. eDNA suitability for calculating EBVs With biodiversity decreasing globally, indicators such as EBVs and EOVs have been proposed to track changes over time. The Europa Biodiversity Observation Network (EuropaBON) project, to provide a framework to monitor EU biodiversity and ecosystem services, has developed a list of 84 EBVs that are divided into six classes (Species populations, genetic composition, species traits, community composition, ecosystem structure, and ecosystem function), span three realms (terrestrial, freshwater, marine) and all major taxa. Novel technology, including eDNA, has been highlighted as a source of data that is ready to be implemented in the calculation of EBVs. Similarly, EOVs have been developed to monitor and forecast the health of our oceans and climate, including many that relate to biology and ecosystems. Within these EOVs, the use of eDNA-based methods may be highly appropriate as a suitable source of data for calculations. In this deliverable, we will aim to comprehensively review each of the relevant EBVs and EOVs to identify the suitability, relevance and history of eDNA data to calculate each one. To this end, we have set up a document to assess the suitability of each EBV using the ranking order described by Zinger et al. (2020): Unknown, Useless, Poorly Useful, Fairly Useful, Useful, Very Useful. So far, two EBVs have been documented fully with ongoing background reading to populate the rest. The final deliverable will contain the finished table. 4. Statistical tools and analyses for eDNA data The final component of this deliverable is to review the diversity metrics and statistical tools that are currently being used with eDNA-based data, with a specific aim to identify those that best comply with the unique characteristics of this data type. To that end, we will review the methods currently 8 Deliverable 2.1 - Interim report on meta-analysis of eDNA-based approaches being implemented, and report on the optimal diversity indices for observing and measuring biodiversity. 5. Proposed outcomes for final deliverable The final deliverable for Task 2.1 will summarize the performance of eDNA compared to traditional methods by answering the question: Does eDNA perform better, the same or worse? We will additionally share the analytical methods that are most suitable for eDNA data and give examples of EBVs/EOVs that may benefit from the use of this type of data. Furthermore, we anticipate up to three peer-reviewed publications from the work done in this task based on the following components of the task: 1. The two quantification meta-analyses 2. The species richness meta-analysis and summary of statistical tools for biodiversity analyses using eDNA data 3. The table of eDNA data suitability to measure EBVs and EOVs, providing examples when possible. 9