Supervised machine learning is superior to indicator value inference in monitoring the environmental impacts of salmon aquaculture using eDNA metabarcodes

Insights into ecological effects of fish and shellfish mariculture on microeukaryotic community

To better understand the ecological effects of mariculture, the diversity distribution, determinant and interaction of microeukaryote communities from fish cage and suspended shellfish farming were investigated in three bays of South China Coast. Our alpha and beta diversity analyses showed that the difference of the microeukaryote community between fish and shellfish farming was more significant at local than regional scale, and microeukaryotes respond more to spatial effect than mariculture effect at regional scale. Mantel test, variation partitioning analysis and co-occurrence network analysis revealed that the environmental factors especially chemical and biotic factors contributed more to community assembly in fish than shellfish farming. Based on the comparisons of community composition and determinant between fish and shellfish farming, the effect mechanisms of the two farming types on microeukaryote community were proposed. Fish farming brings significant environmental variation and thus has strong bottom-up impacts on microeukaryotes, while shellfish farming exerts a grazing pressure on microeukaryotes by filter-feeding and has top-down control to them. Furthermore, the network stability analyses revealed weaker community stability in fish than shellfish farming, suggesting that the microeukaryote community was more sensitive to environmental change deduced by fish than shellfish farming. Overall, this study revealed the different influencing mechanisms of fish and shellfish mariculture on microeukaryotes, which will improve the understanding of the ecological effects of mariculture and provide guidance for the management of mariculture under future environmental pressures.

TICI: a taxon-independent community index for eDNA-based ecological health assessment

Global biodiversity is declining at an ever-increasing rate. Yet effective policies to mitigate or reverse these declines require ecosystem condition data that are rarely available. Morphology-based bioassessment methods are difficult to scale, limited in scope, suffer prohibitive costs, require skilled taxonomists, and can be applied inconsistently between practitioners. Environmental DNA (eDNA) metabarcoding offers a powerful, reproducible and scalable solution that can survey across the tree-of-life with relatively low cost and minimal expertise for sample collection. However, there remains a need to condense the complex, multidimensional community information into simple, interpretable metrics of ecological health for environmental management purposes. We developed a riverine taxon-independent community index (TICI) that objectively assigns indicator values to amplicon sequence variants (ASVs), and significantly improves the statistical power and utility of eDNA-based bioassessments. The TICI model training step uses the Chessman iterative learning algorithm to assign health indicator scores to a large number of ASVs that are commonly encountered across a wide geographic range. New sites can then be evaluated for ecological health by averaging the indicator value of the ASVs present at the site. We trained a TICI model on an eDNA dataset from 53 well-studied riverine monitoring sites across New Zealand, each sampled with a high level of biological replication (n = 16). Eight short-amplicon metabarcoding assays were used to generate data from a broad taxonomic range, including bacteria, microeukaryotes, fungi, plants, and animals. Site-specific TICI scores were strongly correlated with historical stream condition scores from macroinvertebrate assessments (macroinvertebrate community index or MCI; R2 = 0.82), and TICI variation between sample replicates was minimal (CV = 0.013). Taken together, this demonstrates the potential for taxon-independent eDNA analysis to provide a reliable, robust and low-cost assessment of ecological health that is accessible to environmental managers, decision makers, and the wider community.

Supervised machine learning improves general applicability of eDNA metabarcoding for reservoir health monitoring

Effective and standardized monitoring methodologies are vital for successful reservoir restoration and management. Environmental DNA (eDNA) metabarcoding sequencing offers a promising alternative for biomonitoring and can overcome many limitations of traditional morphological bioassessment. Recent attempts have even shown that supervised machine learning (SML) can directly infer biotic indices (BI) from eDNA metabarcoding data, bypassing the cumbersome calculation process of BI regardless of the taxonomic assignment of eDNA sequences. However, questions surrounding the general applicability of this taxonomy-free approach to monitoring reservoir health remain unclear, including model stability, feature selection, algorithm choice, and multi-season biomonitoring. Here, we firstly developed a novel biological integrity index (Me-IBI) that integrates multitrophic interactions and environmental information, based on taxonomy-assigned eDNA metabarcoding data. The Me-IBI can better distinguish the actual health status of the Three Gorges Reservoir (TGR) than physicochemical assessments and have a clear response to human activity. Then, taking this reliable Me-IBI as a supervised label, we compared the impact of selecting different numbers of features and SML algorithms on the stability and predictive performance of the model for predicting ecological conditions in multiple seasons using taxonomy-free eDNA metabarcoding data. We discovered that even with a small number of features, different SML algorithms can establish a stable model and obtain excellent predictive performance. Finally, we proposed a four-step strategy for standardized routine biomonitoring using SML tools. Our study firstly explores the general applicability problem of the taxonomy-free eDNA-SML approach and establishes a solid foundation for the large-scale and standardized biomonitoring application.

Planning and licensing for marine aquaculture

Marine aquaculture has the potential to increase its contribution to the global food system and provide valuable ecosystem services, but appropriate planning, licensing and regulation systems must be in place to enable sustainable development. At present, approaches vary considerably throughout the world, and several national and regional investigations have highlighted the need for reforms if marine aquaculture is to fulfil its potential. This article aims to map and evaluate the challenges of planning and licensing for growth of sustainable marine aquaculture. Despite the range of species, production systems and circumstances, this study found a number of common themes in the literature; complicated and fragmented approaches to planning and licensing, property rights and the licence to operate, competition for space and marine spatial planning, emerging species and diversifying marine aquaculture production (seaweed production, Integrated Multi-Trophic Aquaculture [IMTA], nutrient and carbon offsetting with aquaculture, offshore aquaculture and co-location and multiuse platforms), and the need to address knowledge gaps and use of decision-support tools. Planning and licensing can be highly complicated, so the UK is used as a case study to show more detailed examples that highlight the range of challenges and uncertainty that industry, regulators and policymakers face across interacting jurisdictions. There are many complexities, but this study shows that many countries have undergone, or are undergoing, similar challenges, suggesting that lessons can be learned by sharing knowledge and experiences, even across different species and production systems, rather than having a more insular focus.

A combination of machine-learning and eDNA reveals the genetic signature of environmental change at the landscape levels

The current advances of environmental DNA (eDNA) bring profound changes to ecological monitoring and provide unique insights on the biological diversity of ecosystems. The very nature of eDNA data is challenging yet also revolutionizing how biological monitoring information is analysed. In particular, new metrics and approaches should take full advantage of the extent and detail of molecular data produced by genetic methods. In this perspective, machine learning algorithms are particularly promising as they can capture complex relationships between the multiple environmental pressures and the diversity of biological communities. We investigated the potential of a new generation of biomonitoring tools that implement machine-learning techniques to fully exploit eDNA datasets. We trained a machine learning model to discriminate between reference and impacted communities of freshwater macroinvertebrates and assessed its performances using a large eDNA dataset collected at 64 standard federal monitoring sites across Switzerland. We show that a model trained on eDNA is significantly better than a naive model and performs similarly to a model trained on traditional data. Our proof-of-concept shows that such a combination of eDNA and machine learning approaches has the potential to complement or even replace traditional environmental monitoring, and could be scaled along temporal or spatial dimensions.

Assessing the effects of aquaculture on tidal flat ecological status using multi-metrics interaction-based index of biotic integrity (Mt-IBI)

Given tidal flat special environmental conditions and the degree of pollution caused by human activities, there is an urgent need to quantitatively assess their ecological status. Bioindication has become an indispensable part of environmental quality monitoring on account of its sensitivity to environmental disturbance. Thus, this study used bio-indicators to establish a multi-metrics-based index of biotic integrity (Mt-IBI) to evaluate the ecological status of the tidal flats with/without aquaculture through metagenomic sequencing. Four core indexes that were significantly correlated to other indexes with redundancy (p < 0.05), including Escherichia, beta-lactam antibiotic resistance genes, cellulase and xyloglucanases and the keystone species with 21° in the network, were selected after the screening processes. By implementing Mt-IBI in the tidal flats, the ecological health of the sampling sites was categorized into three levels, with Mt-IBI values of 2.01-2.63 (severe level), 2.81-2.93 (moderate level) and 3.23-4.18 (mild level), respectively. Through SEM analysis, water chemical oxygen demand and antibiotics were determined to be the primary controlling factors of the ecological status of tidal flat regions influenced by aquaculture, followed by salinity and total nitrogen. It is worth noting that the alteration of microbial communities impacted ecological status through the mediation of antibiotics. It is hoped that the results of our study will provide a theoretical basis for coastal environment restoration and that the use of Mt-IBI to assess ecosystem status in different aquatic environments will be further popularized in the future.

Comparing quantile regression spline analyses and supervised machine learning for environmental quality assessment at coastal marine aquaculture installations

Organic enrichment associated with marine finfish aquaculture is a local stressor of marine coastal ecosystems. To maintain ecosystem services, the implementation of biomonitoring programs focusing on benthic diversity is required. Traditionally, impact-indices are determined by extracting and identifying benthic macroinvertebrates from samples. However, this is a time-consuming and expensive method with low upscaling potential. A more rapid, inexpensive, and robust method to infer the environmental quality of marine environments is eDNA metabarcoding of bacterial communities. To infer the environmental quality of coastal habitats from metabarcoding data, two taxonomy-free approaches have been successfully applied for different geographical regions and monitoring goals, namely quantile regression splines (QRS) and supervised machine learning (SML). However, their comparative performance remains untested for monitoring the impact of organic enrichment introduced by aquaculture on marine coastal environments. We compared the performance of QRS and SML using bacterial metabarcoding data to infer the environmental quality of 230 aquaculture samples collected from seven farms in Norway and seven farms in Scotland along an organic enrichment gradient. As a measure of environmental quality, we used the Infaunal Quality Index (IQI) calculated from benthic macrofauna data (reference index). The QRS analysis plotted the abundance of amplicon sequence variants (ASVs) as a function to the IQI from which the ASVs with a defined abundance peak were assigned to eco-groups and a molecular IQI was subsequently calculated. In contrast, the SML approach built a random forest model to directly predict the macrofauna-based IQI. Our results show that both QRS and SML perform well in inferring the environmental quality with 89% and 90% accuracy, respectively. For both geographic regions, there was high correspondence between the reference IQI and both the inferred molecular IQIs (p < 0.001), with the SML model showing a higher coefficient of determination compared to QRS. Among the 20 most important ASVs identified by the SML approach, 15 were congruent with the good quality spline ASV indicators identified via QRS for both Norwegian and Scottish salmon farms. More research on the response of the ASVs to organic enrichment and the co-influence of other environmental parameters is necessary to eventually select the most powerful stressor-specific indicators. Even though both approaches are promising to infer environmental quality based on metabarcoding data, SML showed to be more powerful in handling the natural variability. For the improvement of the SML model, addition of new samples is still required, as background noise introduced by high spatio-temporal variability can be reduced. Overall, we recommend the development of a powerful SML approach that will be onwards applied for monitoring the impact of aquaculture on marine ecosystems based on eDNA metabarcoding data.

Unsupervised biological integrity assessment by eDNA biomonitoring of multi-trophic aquatic taxa

The biological integrity of global freshwater ecosystems is threatened by ever-increasing environmental stressors due to increased human activities, such as land-use change, eutrophication, toxic pollutants, overfishing, and exploitation. Traditional ecological assessments of lake or riverine ecosystems often require human supervision of a pre-selected reference area, using the current state of the reference area as the expected state. However, selecting an appropriate reference area has become increasingly difficult with the expansion of human activities. Here, an unsupervised biological integrity assessment framework based on environmental DNA metabarcoding without a prior reference area is proposed. Taxon richness, species dominance, co-occurrence network density, and phylogenetic distance were used to assess the aquatic communities in the Taihu Lake basin. Multi-gene metabarcoding revealed comprehensive biodiversity at multiple trophic levels including algae, protists, zooplankton, and fish. Fish sequences were mainly derived from 12S, zooplankton mainly from mitochondrial cytochrome C oxidase subunit I, and algae and protists mainly from 18S. There were significant differences in community composition among lakes, rivers, and reservoirs but no significant differences in the four fundamental biological indicators. The algal and zooplankton integrities were positively correlated with protist and fish integrities, respectively. Additionally, the algal integrity of lakes was found to be significantly lower than that of rivers. The unsupervised assessment framework proposed in this study allows different ecosystems, including the same ecosystem in different seasons, to adopt the same indicators and assessment methods, which is more convenient for environmental management and decision-making.

Sediment Microbiota as a Proxy of Environmental Health: Discovering Inter- and Intrakingdom Dynamics along the Eastern Mediterranean Continental Shelf

Sedimentary marine habitats are the largest ecosystem on our planet in terms of area. Marine sediment microbiota govern most of the benthic biological processes and therefore are responsible for much of the global biogeochemical activity. Sediment microbiota respond, even rapidly, to natural change in environmental conditions as well as disturbances of anthropogenic sources. The latter greatly impact the continental shelf. Characterization and monitoring of the sediment microbiota may serve as an important tool for assessing environmental health and indicate changes in the marine ecosystem. This study examined the suitability of marine sediment microbiota as a bioindicator for environmental health in the eastern Mediterranean Sea. Integration of information from Bacteria, Archaea, and Eukaryota enabled robust assessment of environmental factors controlling sediment microbiota composition: seafloor-depth (here representing sediment grain size and total organic carbon), core depth, and season (11%, 4.2%, and 2.5% of the variance, respectively). Furthermore, inter- and intrakingdom cooccurrence patterns indicate that ecological filtration as well as stochastic processes may control sediment microbiota assembly. The results show that the sediment microbiota was robust over 3 years of sampling, in terms of both representation of region (outside the model sites) and robustness of microbial markers. Furthermore, anthropogenic disturbance was reflected by significant transformations in sediment microbiota. We therefore propose sediment microbiota analysis as a sensitive approach to detect disturbances, which is applicable for long-term monitoring of marine environmental health. IMPORTANCE Analysis of data, curated over 3 years of sediment sampling, improves our understanding of microbiota assembly in marine sediment. Furthermore, we demonstrate the importance of cross-kingdom integration of information in the study of microbial community ecology. Finally, the urgent need to propose an applicable approach for environmental health monitoring is addressed here by establishment of sediment microbiota as a robust and sensitive model.

Maximising signal-to-noise ratios in environmental DNA-based monitoring

Man's impacts on global ecosystems are increasing and there is a growing demand that these activities be appropriately monitored. Monitoring requires measurement of a response metric ('signal') that changes maximally and consistently in response to the monitored activity irrespective of other factors ('noise'), thus maximising the signal-to-noise ratio. Indices derived from time-consuming morphology-based taxonomic identification of organisms are a core part of many monitoring programmes. Metabarcoding is an alternative to morphology-based identification and involves the sequencing of short fragments of DNA ('markers') from multiple taxa simultaneously. DNA suitable for metabarcoding includes that extracted from environmental samples (eDNA). Metabarcoding outputs DNA sequences that can be identified (annotated) by matching them against archived annotated sequences. However, sequences from most organisms are not archived - preventing annotation and potentially limiting metabarcoding in monitoring applications. Consequently, there is growing interest in using unannotated sequences as response metrics in monitoring programmes. We compared the sequences from three commonly used markers (16S (V3/V4 regions), 18S (V1/V2 regions) and COI) and, sampling along steep impact gradients, showed that the 16S and COI sequences were associated with the largest and smallest signal-to-noise ratio respectively. We trialled four separate, intuitive, noise-reduction approaches and demonstrated that removing less frequent sequences improved the signal-to-noise ratio, partitioning an additional 25 % from noise to explanatory factors in non-parametric ANOVA (NPA) and reducing dispersion in the data. For the 16S marker, retaining only the most frequently observed sequence, per sample, resulting in nine sequences across 150 samples, generated a near-maximal signal-to-noise ratio (95 % of the variance explained in NPA). We recommend that NPA, combined with rigorous elimination of less frequent sequences, be used to pre-filter sequences/taxa being used in monitoring applications. Our approach will simplify downstream analysis, for example the identification of key taxa and functional associations.

Net overboard: Comparing marine eDNA sampling methodologies at sea to unravel marine biodiversity

Environmental DNA (eDNA) analyses are powerful for describing marine biodiversity but must be optimized for their effective use in routine monitoring. To maximize eDNA detection probabilities of sparsely distributed populations, water samples are usually concentrated from larger volumes and filtered using fine-pore membranes, often a significant cost-time bottleneck in the workflow. This study aimed to streamline eDNA sampling by investigating plankton net versus bucket sampling, direct versus sequential filtration including self-preserving filters. Biodiversity was assessed using metabarcoding of the small ribosomal subunit (18S rRNA) and mitochondrial cytochrome c oxidase I (COI) genes. Multispecies detection probabilities were estimated for each workflow using a probabilistic occupancy modelling approach. Significant workflow-related differences in biodiversity metrics were reported. Highest amplicon sequence variant (ASV) richness was attained by the bucket sampling combined with self-preserving filters, comprising a large portion of microplankton. Less diversity but more metazoan taxa were captured in the net samples combined with 5 μm pore size filters. Prefiltered 1.2 μm samples yielded few or no unique ASVs. The highest average (~32%) metazoan detection probabilities in the 5 μm pore size net samples confirmed the effectiveness of preconcentration plankton for biodiversity screening. These results contribute to streamlining eDNA sampling protocols for uptake and implementation in marine biodiversity research and surveillance.

Fine-scale differences in eukaryotic communities inside and outside salmon aquaculture cages revealed by eDNA metabarcoding

Aquaculture impacts on marine benthic ecosystems are widely recognized and monitored. However, little is known about the community changes occurring in the water masses surrounding aquaculture sites. In the present study, we studied the eukaryotic communities inside and outside salmonid aquaculture cages through time to assess the community changes in the neighbouring waters of the farm. Water samples were taken biweekly over five months during the production phase from inside the cages and from nearby points located North and South of the salmon farm. Eukaryotic communities were analyzed by eDNA metabarcoding of the partial COI Leray-XT fragment. The results showed that eukaryotic communities inside the cages were significantly different from those in the outside environment, with communities inside the cages having higher diversity values and more indicator species associated with them. This is likely explained by the appearance of fouling species that colonize the artificial structures, but also by other species that are attracted to the cages by other means. Moreover, these effects were highly localized inside the cages, as the communities identified outside the cages, both North and South, had very similar eukaryotic composition at each point in time. Overall, the eukaryotic communities, both inside and outside the cages, showed similar temporal fluctuations through the summer months, with diversity peaks occurring at the end of July, beginning of September, and in the beginning of November, with the latter showing the highest Shannon diversity and richness values. Hence, our study suggests that seasonality, together with salmonid aquaculture, are the main drivers of eukaryotic community structure in surface waters surrounding the farm.

Consumers' attitude toward participation in community-supported aquaculture: a case of Kurdistan province in the west of Iran

Considering the increasing importance of sustainable operations in the agricultural sector and the relationship between producers and consumers, the current study was to determine customers' attitudes on participation in community-supported aquaculture programs in Kurdistan province (in the west of Iran). The present study was a survey, non-experimental, applied, and descriptive-correlational research. Using a literature review and field studies, factors affecting consumers' attitudes have been extracted. Then, to determine validity, the questionnaire was given to a panel of subject matter experts. Furthermore, to assess the reliability of the research instrument, the Cronbach's alpha coefficient was calculated. The results showed the good validity and reliability of the research tool. Moreover, structural equation modeling was used to confirm the proposed model. The results showed that among the factors affecting the attitude of consumers, the price of aquatic products had the first place, which has been neglected in the previous studies. In this regard, it was suggested that by creating support funds and facilities for the participation of consumers in community-supported aquaculture programs, setting standards for healthy aquatic products, and producing programs related to the introduction of community-supported aquaculture programs on the radio and television, the attitude of the consumers can be improved.

Application of Omics Tools in Designing and Monitoring Marine Protected Areas For a Sustainable Blue Economy

A key component of the global blue economy strategy is the sustainable extraction of marine resources and conservation of marine environments through networks of marine protected areas (MPAs). Connectivity and representativity are essential factors that underlie successful implementation of MPA networks, which can safeguard biological diversity and ecosystem function, and ultimately support the blue economy strategy by balancing ocean use with conservation. New “big data” omics approaches, including genomics and transcriptomics, are becoming essential tools for the development and maintenance of MPA networks. Current molecular omics techniques, including population-scale genome sequencing, have direct applications for assessing population connectivity and for evaluating how genetic variation is represented within and among MPAs. Effective baseline characterization and long-term, scalable, and comprehensive monitoring are essential for successful MPA management, and omics approaches hold great promise to characterize the full range of marine life, spanning the microbiome to megafauna across a range of environmental conditions (shallow sea to the deep ocean). Omics tools, such as eDNA metabarcoding can provide a cost-effective basis for biodiversity monitoring in large and remote conservation areas. Here we provide an overview of current omics applications for conservation planning and monitoring, with a focus on metabarcoding, metagenomics, and population genomics. Emerging approaches, including whole-genome sequencing, characterization of genomic architecture, epigenomics, and genomic vulnerability to climate change are also reviewed. We demonstrate that the operationalization of omics tools can enhance the design, monitoring, and management of MPAs and thus will play an important role in a modern and comprehensive blue economy strategy.

Interfacing Machine Learning and Microbial Omics: A Promising Means to Address Environmental Challenges

Microbial communities are ubiquitous and carry an exceptionally broad metabolic capability. Upon environmental perturbation, microbes are also amongst the first natural responsive elements with perturbation-specific cues and markers. These communities are thereby uniquely positioned to inform on the status of environmental conditions. The advent of microbial omics has led to an unprecedented volume of complex microbiological data sets. Importantly, these data sets are rich in biological information with potential for predictive environmental classification and forecasting. However, the patterns in this information are often hidden amongst the inherent complexity of the data. There has been a continued rise in the development and adoption of machine learning (ML) and deep learning architectures for solving research challenges of this sort. Indeed, the interface between molecular microbial ecology and artificial intelligence (AI) appears to show considerable potential for significantly advancing environmental monitoring and management practices through their application. Here, we provide a primer for ML, highlight the notion of retaining biological sample information for supervised ML, discuss workflow considerations, and review the state of the art of the exciting, yet nascent, interdisciplinary field of ML-driven microbial ecology. Current limitations in this sphere of research are also addressed to frame a forward-looking perspective toward the realization of what we anticipate will become a pivotal toolkit for addressing environmental monitoring and management challenges in the years ahead.

Environmental DNA metabarcoding for benthic monitoring: A review of sediment sampling and DNA extraction methods

Environmental DNA (eDNA) metabarcoding (parallel sequencing of DNA/RNA for identification of whole communities within a targeted group) is revolutionizing the field of aquatic biomonitoring. To date, most metabarcoding studies aiming to assess the ecological status of aquatic ecosystems have focused on water eDNA and macroinvertebrate bulk samples. However, the eDNA metabarcoding has also been applied to soft sediment samples, mainly for assessing microbial or meiofaunal biota. Compared to classical methodologies based on manual sorting and morphological identification of benthic taxa, eDNA metabarcoding offers potentially important advantages for assessing the environmental quality of sediments. The methods and protocols utilized for sediment eDNA metabarcoding can vary considerably among studies, and standardization efforts are needed to improve their robustness, comparability and use within regulatory frameworks. Here, we review the available information on eDNA metabarcoding applied to sediment samples, with a focus on sampling, preservation, and DNA extraction steps. We discuss challenges specific to sediment eDNA analysis, including the variety of different sources and states of eDNA and its persistence in the sediment. This paper aims to identify good-practice strategies and facilitate method harmonization for routine use of sediment eDNA in future benthic monitoring.

LANDMark: an ensemble approach to the supervised selection of biomarkers in high-throughput sequencing data

Background

Identification of biomarkers, which are measurable characteristics of biological datasets, can be challenging. Although amplicon sequence variants (ASVs) can be considered potential biomarkers, identifying important ASVs in high-throughput sequencing datasets is challenging. Noise, algorithmic failures to account for specific distributional properties, and feature interactions can complicate the discovery of ASV biomarkers. In addition, these issues can impact the replicability of various models and elevate false-discovery rates. Contemporary machine learning approaches can be leveraged to address these issues. Ensembles of decision trees are particularly effective at classifying the types of data commonly generated in high-throughput sequencing (HTS) studies due to their robustness when the number of features in the training data is orders of magnitude larger than the number of samples. In addition, when combined with appropriate model introspection algorithms, machine learning algorithms can also be used to discover and select potential biomarkers. However, the construction of these models could introduce various biases which potentially obfuscate feature discovery.

Results

We developed a decision tree ensemble, LANDMark, which uses oblique and non-linear cuts at each node. In synthetic and toy tests LANDMark consistently ranked as the best classifier and often outperformed the Random Forest classifier. When trained on the full metabarcoding dataset obtained from Canada’s Wood Buffalo National Park, LANDMark was able to create highly predictive models and achieved an overall balanced accuracy score of 0.96 ± 0.06. The use of recursive feature elimination did not impact LANDMark’s generalization performance and, when trained on data from the BE amplicon, it was able to outperform the Linear Support Vector Machine, Logistic Regression models, and Stochastic Gradient Descent models (p ≤ 0.05). Finally, LANDMark distinguishes itself due to its ability to learn smoother non-linear decision boundaries.

Conclusions

Our work introduces LANDMark, a meta-classifier which blends the characteristics of several machine learning models into a decision tree and ensemble learning framework. To our knowledge, this is the first study to apply this type of ensemble approach to amplicon sequencing data and we have shown that analyzing these datasets using LANDMark can produce highly predictive and consistent models.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-022-04631-z.

A bacterial index to estimate lake trophic level: National scale validation

Lakes and their catchments have been subjected to centuries to millennia of exploitation by humans. Efficient monitoring methods are required to promote proactive protection and management. Traditional monitoring is time consuming and expensive, which limits the number of lakes monitored. Lake surface sediments provide a temporally integrated representation of environmental conditions and contain high microbial biomass. Based on these attributes, we hypothesized that bacteria associated with lake trophic states could be identified and used to develop an index that would not be confounded by non-nutrient stressor gradients. Metabarcoding (16S rRNA gene) was used to assess bacterial communities present in surface sediments from 259 non-saline lakes in New Zealand encompassing a range of trophic states from alpine microtrophic lakes to lowland hypertrophic lakes. A subset of lakes (n = 96) with monitoring data was used to identify indicator amplicon sequence variants (ASVs) associated with different trophic states. A total of 10,888 indicator taxa were identified and used to develop a Sediment Bacterial Trophic Index (SBTI), which signficantly correlated (r² = 0.842, P < 0.001) with the Trophic Lake Index. The SBTI was then derived for the remaining 163 lakes, providing new knowledge of the trophic state of these unmonitored lakes. This new, robust DNA-based tool provides a rapid and cost-effective method that will allow a greater number of lakes to be monitored and more effectively managed in New Zealand and globally. The SBTI could also be applied in a paleolimnological context to investigate changes in trophic status over centuries to millennia.

Revision and annotation of DNA barcode records for marine invertebrates: report of the 8th iBOL conference hackathon

The accuracy of specimen identification through DNA barcoding and metabarcoding relies on reference libraries containing records with reliable taxonomy and sequence quality. The considerable growth in barcode data requires stringent data curation, especially in taxonomically difficult groups such as marine invertebrates. A major effort in curating marine barcode data in the Barcode of Life Data Systems (BOLD) was undertaken during the 8th International Barcode of Life Conference (Trondheim, Norway, 2019). Major taxonomic groups (crustaceans, echinoderms, molluscs, and polychaetes) were reviewed to identify those which had disagreement between Linnaean names and Barcode Index Numbers (BINs). The records with disagreement were annotated with four tags: a) MIS-ID (misidentified, mislabeled, or contaminated records), b) AMBIG (ambiguous records unresolved with the existing data), c) COMPLEX (species names occurring in multiple BINs), and d) SHARE (barcodes shared between species). A total of 83,712 specimen records corresponding to 7,576 species were reviewed and 39% of the species were tagged (7% MIS-ID, 17% AMBIG, 14% COMPLEX, and 1% SHARE). High percentages (&gt;50%) of AMBIG tags were recorded in gastropods, whereas COMPLEX tags dominated in crustaceans and polychaetes. The high proportion of tagged species reflects either flaws in the barcoding workflow (e.g., misidentification, cross-contamination) or taxonomic difficulties (e.g., synonyms, undescribed species). Although data curation is essential for barcode applications, such manual attempts to examine large datasets are unsustainable and automated solutions are extremely desirable.

Comparing sediment preservation methods for genomic biomonitoring of coastal marine ecosystems

To avoid loss of genetic information in environmental DNA (eDNA) field samples, the preservation of nucleic acids during field sampling is a critical step. In the development of standard operating procedures (SOPs) for eDNA-based compliance monitoring, the effect of different routinely used sediment preservations on biological community structures serving as bioindicators has gone untested. We compared eDNA metabarcoding results of marine bacterial communities from sample aliquots that were treated with a nucleic acid preservation solution (treated samples) and aliquots that were frozen without further treatment (non-treated samples). Sediment samples were obtained from coastal locations subjected to different stressors (aquaculture, urbanization, industry). DNA extraction efficiency, bacterial community profiles, and measures of alpha- and beta-diversity were highly congruent between treated and non-treated samples. As both preservation methods provide the same relevant information to environmental managers and regulators, we recommend the inclusion of both methods into SOPs for biomonitoring in marine coastal environments.

Environmental DNA gives comparable results to morphology-based indices of macroinvertebrates in a large-scale ecological assessment

Anthropogenic activities are changing the state of ecosystems worldwide, affecting community composition and often resulting in loss of biodiversity. Rivers are among the most impacted ecosystems. Recording their current state with regular biomonitoring is important to assess the future trajectory of biodiversity. Traditional monitoring methods for ecological assessments are costly and time-intensive. Here, we compared monitoring of macroinvertebrates based on environmental DNA (eDNA) sampling with monitoring based on traditional kick-net sampling to assess biodiversity patterns at 92 river sites covering all major Swiss river catchments. From the kick-net community data, a biotic index (IBCH) based on 145 indicator taxa had been established. The index was matched by the taxonomically annotated eDNA data by using a machine learning approach. Our comparison of diversity patterns only uses the zero-radius Operational Taxonomic Units assigned to the indicator taxa. Overall, we found a strong congruence between both methods for the assessment of the total indicator community composition (gamma diversity). However, when assessing biodiversity at the site level (alpha diversity), the methods were less consistent and gave complementary data on composition. Specifically, environmental DNA retrieved significantly fewer indicator taxa per site than the kick-net approach. Importantly, however, the subsequent ecological classification of rivers based on the detected indicators resulted in similar biotic index scores for the kick-net and the eDNA data that was classified using a random forest approach. The majority of the predictions (72%) from the random forest classification resulted in the same river status categories as the kick-net approach. Thus, environmental DNA validly detected indicator communities and, combined with machine learning, provided reliable classifications of the ecological state of rivers. Overall, while environmental DNA gives complementary data on the macroinvertebrate community composition compared to the kick-net approach, the subsequently calculated indices for the ecological classification of river sites are nevertheless directly comparable and consistent.

Ecosystems monitoring powered by environmental genomics: A review of current strategies with an implementation roadmap

A decade after environmental scientists integrated high-throughput sequencing technologies in their toolbox, the genomics-based monitoring of anthropogenic impacts on the biodiversity and functioning of ecosystems is yet to be implemented by regulatory frameworks. Despite the broadly acknowledged potential of environmental genomics to this end, technical limitations and conceptual issues still stand in the way of its broad application by end-users. In addition, the multiplicity of potential implementation strategies may contribute to a perception that the routine application of this methodology is premature or "in development", hence restraining regulators from binding these tools into legal frameworks. Here, we review recent implementations of environmental genomics-based methods, applied to the biomonitoring of ecosystems. By taking a general overview, without narrowing our perspective to particular habitats or groups of organisms, this paper aims to compare, review and discuss the strengths and limitations of four general implementation strategies of environmental genomics for monitoring: (a) Taxonomy-based analyses focused on identification of known bioindicators or described taxa; (b) De novo bioindicator analyses; (c) Structural community metrics including inferred ecological networks; and (d) Functional community metrics (metagenomics or metatranscriptomics). We emphasise the utility of the three latter strategies to integrate meiofauna and microorganisms that are not traditionally utilised in biomonitoring because of difficult taxonomic identification. Finally, we propose a roadmap for the implementation of environmental genomics into routine monitoring programmes that leverage recent analytical advancements, while pointing out current limitations and future research needs.

Global Trends of Benthic Bacterial Diversity and Community Composition Along Organic Enrichment Gradients of Salmon Farms

The analysis of benthic bacterial community structure has emerged as a powerful alternative to traditional microscopy-based taxonomic approaches to monitor aquaculture disturbance in coastal environments. However, local bacterial diversity and community composition vary with season, biogeographic region, hydrology, sediment texture, and aquafarm-specific parameters. Therefore, without an understanding of the inherent variation contained within community complexes, bacterial diversity surveys conducted at individual farms, countries, or specific seasons may not be able to infer global universal pictures of bacterial community diversity and composition at different degrees of aquaculture disturbance. We have analyzed environmental DNA (eDNA) metabarcodes (V3-V4 region of the hypervariable SSU rRNA gene) of 138 samples of different farms located in different major salmon-producing countries. For these samples, we identified universal bacterial core taxa that indicate high, moderate, and low aquaculture impact, regardless of sampling season, sampled country, seafloor substrate type, or local farming and environmental conditions. We also discuss bacterial taxon groups that are specific for individual local conditions. We then link the metabolic properties of the identified bacterial taxon groups to benthic processes, which provides a better understanding of universal benthic ecosystem function(ing) of coastal aquaculture sites. Our results may further guide the continuing development of a practical and generic bacterial eDNA-based environmental monitoring approach.

Identifying the minimum amplicon sequence depth to adequately predict classes in eDNA-based marine biomonitoring using supervised machine learning

Environmental DNA metabarcoding is a powerful approach for use in biomonitoring and impact assessments. Amplicon-based eDNA sequence data are characteristically highly divergent in sequencing depth (total reads per sample) as influenced inter alia by the number of samples simultaneously analyzed per sequencing run. The random forest (RF) machine learning algorithm has been successfully employed to accurately classify unknown samples into monitoring categories. To employ RF to eDNA data, and avoid sequencing-depth artifacts, sequence data across samples are normalized using rarefaction, a process that inherently loses information. The aim of this study was to inform future sampling designs in terms of the relationship between sampling depth and RF accuracy. We analyzed three published and one new bacterial amplicon datasets, using a RF, based initially on the maximal rarefied data available (minimum mean of > 30,000 reads across all datasets) to give our baseline performance. We then evaluated the RF classification success based on increasingly rarefied datasets. We found that extreme to moderate rarefaction (50-5000 sequences per sample) was sufficient to achieve prediction performance commensurate to the full data, depending on the classification task. We did not find that the number of classification classes, data balance across classes, or the total number of sequences or samples, were associated with predictive accuracy. We identified the ability of the training data to adequately characterize the classes being mapped as the most important criterion and discuss how this finding can inform future sampling design for eDNA based biomonitoring to reduce costs and computation time.

Machine learning applications in microbial ecology, human microbiome studies, and environmental monitoring

Advances in nucleic acid sequencing technology have enabled expansion of our ability to profile microbial diversity. These large datasets of taxonomic and functional diversity are key to better understanding microbial ecology. Machine learning has proven to be a useful approach for analyzing microbial community data and making predictions about outcomes including human and environmental health. Machine learning applied to microbial community profiles has been used to predict disease states in human health, environmental quality and presence of contamination in the environment, and as trace evidence in forensics. Machine learning has appeal as a powerful tool that can provide deep insights into microbial communities and identify patterns in microbial community data. However, often machine learning models can be used as black boxes to predict a specific outcome, with little understanding of how the models arrived at predictions. Complex machine learning algorithms often may value higher accuracy and performance at the sacrifice of interpretability. In order to leverage machine learning into more translational research related to the microbiome and strengthen our ability to extract meaningful biological information, it is important for models to be interpretable. Here we review current trends in machine learning applications in microbial ecology as well as some of the important challenges and opportunities for more broad application of machine learning to understanding microbial communities.

Morphological vs. DNA metabarcoding approaches for the evaluation of stream ecological status with benthic invertebrates: Testing different combinations of markers and strategies of data filtering

Macroinvertebrate assemblages are the most common bioindicators used for stream biomonitoring, yet the standard approach exhibits several time-consuming steps, including the sorting and identification of organisms based on morphological criteria. In this study, we examined if DNA metabarcoding could be used as an efficient molecular-based alternative to the morphology-based monitoring of streams using macroinvertebrates. We compared results achieved with the standard morphological identification of organisms sampled in 18 sites located on 15 French wadeable streams to results obtained with the DNA metabarcoding identification of sorted bulk material of the same macroinvertebrate samples, using read numbers (expressed as relative frequencies) as a proxy for abundances. In particular, we evaluated how combining and filtering metabarcoding data obtained from three different markers (COI: BF1-BR2, 18S: Euka02 and 16S: Inse01) could improve the efficiency of bioassessment. In total, 140 taxa were identified based on morphological criteria, and 127 were identified based on DNA metabarcoding using the three markers, with an overlap of 99 taxa. The threshold values used for sequence filtering based on the "best identity" criterion and the number of reads had an effect on the assessment efficiency of data obtained with each marker. Compared to single marker results, combining data from different markers allowed us to improve the match between biotic index values obtained with the bulk DNA versus morphology-based approaches. Both approaches assigned the same ecological quality class to a majority (86%) of the site sampling events, highlighting both the efficiency of metabarcoding as a biomonitoring tool but also the need for further research to improve this efficiency.

Monitoring the ecological status of rivers with diatom eDNA metabarcoding: A comparison of taxonomic markers and analytical approaches for the inference of a molecular diatom index

Recently, several studies demonstrated the usefulness of diatom eDNA metabarcoding as an alternative to assess the ecological quality of rivers and streams. However, the choice of the taxonomic marker as well as the methodology for data analysis differ between these studies, hampering the comparison of their results and effectiveness. The aim of this study was to compare two taxonomic markers commonly used in diatom metabarcoding and three distinct analytical approaches to infer a molecular diatom index. We used the values of classical morphological diatom index as a benchmark for this comparison. We amplified and sequenced both a fragment of the rbcL gene and the V4 region of the 18S rRNA gene for 112 epilithic samples from Swiss and French rivers. We inferred index values using three analytical approaches: by computing it directly from taxonomically assigned sequences, by calibrating de novo the ecovalues of all metabarcodes, and by using a supervised machine learning algorithm to train predictive models. In general, the values of index obtained using the two "taxonomy-free" approaches, encompassing molecular assignment and machine learning, were closer correlated to the values of the morphological index than the values based on taxonomically assigned sequences. The correlations of the three analytical approaches were higher in the case of rbcL compared to the 18S marker, highlighting the importance of the reference database which is more complete for the rbcL marker. Our study confirms the effectiveness of diatom metabarcoding as an operational tool for rivers ecological quality assessment and shows that the analytical approaches by-passing the taxonomic assignments are particularly efficient when reference databases are incomplete.

Comparing sediment DNA extraction methods for assessing organic enrichment associated with marine aquaculture

Marine sediments contain a high diversity of micro- and macro-organisms which are important in the functioning of biogeochemical cycles. Traditionally, anthropogenic perturbation has been investigated by identifying macro-organism responses along gradients. Environmental DNA (eDNA) analyses have recently been advocated as a rapid and cost-effective approach to measuring ecological impacts and efforts are underway to incorporate eDNA tools into monitoring. Before these methods can replace or complement existing methods, robustness and repeatability of each analytical step has to be demonstrated. One area that requires further investigation is the selection of sediment DNA extraction method. Environmental DNA sediment samples were obtained along a disturbance gradient adjacent to a Chinook (Oncorhynchus tshawytscha) salmon farm in Otanerau Bay, New Zealand. DNA was extracted using four extraction kits (Qiagen DNeasy PowerSoil, Qiagen DNeasy PowerSoil Pro, Qiagen RNeasy PowerSoil Total RNA/DNA extraction/elution and Favorgen FavorPrep Soil DNA Isolation Midi Kit) and three sediment volumes (0.25, 2, and 5 g). Prokaryotic and eukaryotic communities were amplified using primers targeting the 16S and 18S ribosomal RNA genes, respectively, and were sequenced on an Illumina MiSeq. Diversity and community composition estimates were obtained from each extraction kit, as well as their relative performance in established metabarcoding biotic indices. Differences were observed in the quality and quantity of the extracted DNA amongst kits with the two Qiagen DNeasy PowerSoil kits performing best. Significant differences were observed in both prokaryotes and eukaryotes (p < 0.001) richness among kits. A small proportion of amplicon sequence variants (ASVs) were shared amongst the kits (~3%) although these shared ASVs accounted for the majority of sequence reads (prokaryotes: 59.9%, eukaryotes: 67.2%). Differences were observed in the richness and relative abundance of taxonomic classes revealed with each kit. Multivariate analysis showed that there was a significant interaction between "distance" from the farm and "kit" in explaining the composition of the communities, with the distance from the farm being a stronger determinant of community composition. Comparison of the kits against the bacterial and eukaryotic metabarcoding biotic index suggested that all kits showed similar patterns along the environmental gradient. Overall, we advocate for the use of Qiagen DNeasy PowerSoil kits for use when characterizing prokaryotic and eukaryotic eDNA from marine farm sediments. We base this conclusion on the higher DNA quality values and richness achieved with these kits compared to the other kits/amounts investigated in this study. The additional advantage of the PowerSoil Kits is that DNA extractions can be performed using an extractor robot, offering additional standardization and reproducibility of results.