Applying convolutional neural networks to speed up environmental DNA annotation in a highly diverse ecosystem

Flück, Benjamin; Mathon, Laëtitia; Manel, Stéphanie; Valentini, Alice; Déjean, Tony; Albouy, Camille; Mouillot, David; Thuiller, Wilfried; Murienne, Jérôme; Brosse, Sébastien; Pellissier, Loïc

doi:10.1038/s41598-022-13412-w

Cited by 8 publications

(6 citation statements)

References 77 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Instead of relying solely on the number of sequences identified per MOTU or the nucleotide sequences as in Cordier et al (2021) and Flück et al (2022), our methods combine both information (Table S1). VAESeq is based on VAE that is optimized to reconstruct the input data.…”

Section: Discussionmentioning

confidence: 99%

“…W and S). For example, A becomes [1, 0, 0, 0] or W becomes [0.5, 0, 0, 0.5] (Flück et al, 2022). The model consists of one autoencoder (AE) and one variational autoencoder (VAE).…”

Section: Vae-based Methods Applied To Edna Datamentioning

confidence: 99%

“…W and S). For example, A becomes [1, 0, 0, 0] or W becomes [0.5, 0, 0, 0.5] (Flück et al, 2022). Furthermore, N nucleotides added to equalize the sequence length become [0.25, 0.25, 0.25, 0.25].…”

Section: Methodsmentioning

confidence: 99%

See 2 more Smart Citations

New deep learning‐based methods for visualizing ecosystem properties using environmental DNA metabarcoding data

Lamperti,

Sanchez,

Si Moussi

et al. 2023

Molecular Ecology Resources

View full text Add to dashboard Cite

Environmental DNA (eDNA) metabarcoding provides an efficient approach for documenting biodiversity patterns in marine and terrestrial ecosystems. The complexity of these data prevents current methods from extracting and analyzing all the relevant ecological information they contain, and new methods may provide better dimensionality reduction and clustering. Here we present two new deep learning‐based methods that combine different types of neural networks (NNs) to ordinate eDNA samples and visualize ecosystem properties in a two‐dimensional space: the first is based on variational autoencoders and the second on deep metric learning. The strength of our new methods lies in the combination of two inputs: the number of sequences found for each molecular operational taxonomic unit (MOTU) detected and their corresponding nucleotide sequence. Using three different datasets, we show that our methods accurately represent several biodiversity indicators in a two‐dimensional latent space: MOTU richness per sample, sequence α‐diversity per sample, Jaccard's and sequence β‐diversity between samples. We show that our nonlinear methods are better at extracting features from eDNA datasets while avoiding the major biases associated with eDNA. Our methods outperform traditional dimension reduction methods such as Principal Component Analysis, t‐distributed Stochastic Neighbour Embedding, Nonmetric Multidimensional Scaling and Uniform Manifold Approximation and Projection for dimension reduction. Our results suggest that NNs provide a more efficient way of extracting structure from eDNA metabarcoding data, thereby improving their ecological interpretation and thus biodiversity monitoring.

show abstract

Section: Discussionmentioning

confidence: 99%

“…W and S). For example, A becomes [1, 0, 0, 0] or W becomes [0.5, 0, 0, 0.5] (Flück et al, 2022). The model consists of one autoencoder (AE) and one variational autoencoder (VAE).…”

Section: Vae-based Methods Applied To Edna Datamentioning

confidence: 99%

See 1 more Smart Citation

New deep learning‐based methods for visualizing ecosystem properties using environmental DNA metabarcoding data

Lamperti,

Sanchez,

Si Moussi

et al. 2023

Molecular Ecology Resources

View full text Add to dashboard Cite

show abstract

“…ASVs were defined as rare based on frequency plots, and ASVs with low read counts (<90 reads) or present in few replicates (<10 replicates) were discarded. Additionally, filtering according to ObiTools3 (Flück et al, 2022) was performed to remove ASVs with higher “internal” counts than head or singleton counts. The maximum read count of each ASV in the negative controls was subtracted from the dataset.…”

Section: Methodsmentioning

confidence: 99%

Decoding the Baltic Sea's Past and Present: A simple Molecular Index for Ecosystem Assessment

Schmidt,

Roman,

Andren

et al. 2024

Preprint

View full text Add to dashboard Cite

Marginal sea ecosystems, such as the Baltic Sea, are severely affected by anthropogenic pressures, such as climate warming, pollution, and eutrophication, which increased in the course of the past century. Biodiversity monitoring data and assessment of environmental status in such systems have typically been carried out only for the past few decades, if at all, and knowledge on pre-impact stability and good ecological status is limited. An extension of monitoring time series can potentially be achieved through analyses of paleoecological records, e.g. for phytoplankton, which form the base of the food web and are highly susceptible to environmental changes. Within the phytoplankton community, dinoflagellates and diatoms play a significant role as primary producers, and their relative dominance in the spring bloom, calculated as Dia/Dino index, is used as an indicator for the environmental status of the Baltic Sea. To extend time series on the dominance patterns and include non-fossilized dinoflagellates, we here establish a simple droplet digital PCR (ddPCR) reaction on ancient DNA from sediment cores that decodes phytoplankton dynamics. We focus on two common spring bloom species, the diatom Skeletonema marinoi and the dinoflagellate Apocalathium malmogiense, for which we evaluate a DNA based dominance index. It performs very well in comparison to DNA metabarcoding and modern monitoring and can elucidate past species dominance across the past century in three basins of the Baltic and across millennia in two of these basins. For the past century, we see a dominance shift already starting before the mid-20th century in two of the Baltic Sea basins, thus substantially predating current monitoring programs. Shifts are only partly coeval among the cores and the index shows different degrees of stability. This pattern is confirmed across millennia, where a long-term stable relationship between the diatom and the dinoflagellate is observed in the Eastern Gotland Basin, while data from the Gulf of Finland bear testimony to a much more unstable relationship. This confirms that good ecological status based on the dominance pattern of diatoms and dinoflagellates must be established locally and exemplifies how sediment core DNA can be employed to extend monitoring data.

show abstract

“…Assessing the vitality of habitats and identifying areas in need of restoration demands a level of insight that transcends human capacity alone. Artificial Intelligence (AI) offers an aerial lens that unravels the secrets of Earth's diverse ecosystems through its unparalleled image analysis capabilities, often propelled by sophisticated convolutional neural networks (CNNs) (Flück et al, 2022, Perry et al, 2022). AI's ability to process enormous datasets swiftly and accurately is a game-changer in habitat assessment, transcending the limitations of manual surveys and traditional monitoring methods.…”

Section: Habitat Assessment and Resource Conservationmentioning

confidence: 99%

Reviewing the role of AI in environmental monitoring and conservation: A data-driven revolution for our planet

Onyebuchi Nneamaka Chisom,

Preye Winston Biu,

Aniekan Akpan Umoh

et al. 2024

World J. Adv. Res. Rev.

View full text Add to dashboard Cite

The rapid increase in human activities is causing significant damage to our planet's ecosystems, necessitating innovative solutions to preserve biodiversity and counteract ecological threats. Artificial Intelligence (AI) has emerged as a transformative force, providing unparalleled capabilities for environmental monitoring and conservation. This research paper explores the applications of AI in ecosystem management, including wildlife tracking, habitat assessment, biodiversity analysis, and natural disaster prediction. AI's role in environmental monitoring and conservation includes wildlife tracking, habitat assessment, resource conservation, biodiversity analysis, and species identification. AI algorithms analyze camera trap footage, drone imagery, and GPS data to identify and estimate population sizes, leading to improved anti-poaching efforts and enhanced protection of diverse species. Habitat assessment and resource conservation involve AI-powered image analysis, which aids in assessing forest health, detecting deforestation, and identifying areas in need of restoration. Biodiversity analysis and species identification are achieved through AI algorithms that analyze acoustic recordings, environmental DNA (eDNA), and camera trap footage. These innovations identify different species, assess biodiversity levels, and even discover new or endangered species. AI-powered flood prediction systems provide early warnings, empowering communities with better preparedness and evacuation efforts. Challenges, such as data quality and availability, algorithmic bias, and infrastructure limitations, are acknowledged as opportunities for growth and improvement. In policy and regulation, the paper advocates for clear frameworks prioritizing data privacy and security, algorithmic transparency, and equitable access. Responsible development and ethical use of AI are emphasized as foundational pillars, ensuring that the integration of AI into environmental conservation aligns with principles of fairness, transparency, and societal benefit.

show abstract

Applying convolutional neural networks to speed up environmental DNA annotation in a highly diverse ecosystem

Cited by 8 publications

References 77 publications

New deep learning‐based methods for visualizing ecosystem properties using environmental DNA metabarcoding data

New deep learning‐based methods for visualizing ecosystem properties using environmental DNA metabarcoding data

Decoding the Baltic Sea's Past and Present: A simple Molecular Index for Ecosystem Assessment

Reviewing the role of AI in environmental monitoring and conservation: A data-driven revolution for our planet

Contact Info

Product

Resources

About