Marine dumping area eukaryotic microorganism monitoring method based on environmental DNA technology

Abstract

The invention discloses a marine dumping area eukaryotic microorganism monitoring method based on an environmental DNA technology, and belongs to the technical field of environmental monitoring and marine ecology. The invention provides a microbe monitoring method which fully utilizes abundance data and DNA sequence data and improves the accuracy of subsequent analysis, and a marine dumping area eukaryotic microbe monitoring method which can accurately reveal the composition of eukaryotic microflora in different marine dumping areas and the diversity change of the eukaryotic microflora. And a flexible and reliable technical platform is provided for ecological monitoring of a marine pollution area. According to the method, the environment DNA data is divided into the environment DNA abundance data and the environment DNA sequence data, and the environment DNA data is denoised through the encoder in consideration of the characteristics that the environment DNA abundance data has high sparsity and has a large number of zero values, so that the data quality can be improved, and a reliable basis is provided for subsequent ecological inference. The structure information of the environmental DNA data can be better captured, and the high noise and sparsity influence of the environmental DNA data can be effectively reduced.

Description

Technical Field

[0001] This invention belongs to the field of environmental monitoring and marine ecology technology, specifically relating to a method for monitoring eukaryotic microorganisms in marine dumping areas based on environmental DNA technology. Background Technology

[0002] Marine dumping zones are areas in the ocean specifically designated for dumping waste, including but not limited to industrial wastewater, municipal sewage, silt, and other waste materials. These dumped materials typically contain large amounts of toxic and hazardous substances, posing a serious threat to marine ecosystems, particularly impacting benthic organisms and the microbial community structure. Microorganisms play a crucial role in marine ecosystems, participating not only in the decomposition of organic matter and nutrient cycling but also serving as indicators of ecosystem health. Therefore, understanding the microbial diversity of marine dumping zones is essential for assessing their ecological impacts and developing environmental management strategies.

[0003] Traditional methods for monitoring marine microorganisms primarily involve collecting samples by fishing combined with morphological identification using culture techniques. While these methods provide some biological information, they have significant limitations in detecting marine microorganisms. However, the rapid development of molecular biology techniques has led to environmental DNA (eDNA) technology becoming an important tool for studying and monitoring marine microbial communities. But due to limitations in sequencing technology and the complexity of environmental DNA data (characterized by high noise and high dimensionality), dimensionality reduction and other related processing are necessary to preserve the data structure and extract useful information from the environmental DNA data.

[0004] Environmental samples from marine dumping sites are complex and diverse. Traditional environmental DNA techniques used for species diversity analysis only reduce the dimensionality of abundance data, neglecting the species information hidden within environmental DNA sequence data and failing to fully utilize multi-source information. In contrast, the environment contains multiple species and sequences, and the interactions between their samples are often implicit in the data. Traditional dimensionality reduction methods alone are insufficient for effective analysis, leading to inaccuracies in subsequent analyses.

[0005] Therefore, developing a microbial monitoring method that can fully utilize abundance data and DNA sequence data to improve the accuracy of subsequent analyses has become an important issue that urgently needs to be addressed. Summary of the Invention

[0006] Therefore, the purpose of this invention is to provide a method for monitoring eukaryotic microorganisms in marine dumping areas based on environmental DNA technology, in order to solve the technical problem that existing microbial monitoring methods cannot fully utilize abundance data and DNA sequence data, resulting in inaccurate subsequent analyses.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for monitoring eukaryotic microorganisms in marine dumping areas based on environmental DNA technology includes the following steps: S101, Collect environmental DNA samples; S102, Sequencing analysis and preprocessing of the environmental DNA sample are performed to obtain environmental DNA abundance data and environmental DNA sequence data; S103, Input the environmental DNA sequence data into a CNN model to obtain the sequence features of the DNA sequence data; S104, the ASV sequence features in the sequence features are weighted and averaged according to the DNA abundance data to obtain fusion sequence feature data; S105, input the environmental DNA abundance data and the fusion sequence feature data into the multimodal variational autoencoder TotalVI to obtain environmental DNA data; S106, Perform species classification annotation and diversity analysis on the environmental DNA data.

[0008] Based on the above technical solution, the environmental DNA sample further includes water samples and sediment samples from the marine dumping area.

[0009] Based on the above technical solution, the sequencing analysis further includes the following steps: S102.1, Extract DNA from the environmental DNA sample; S102.2, PCR amplification of extracted DNA was performed using 18S rRNA V9 region-specific primers; S102.3, Perform high-throughput sequencing on the PCR amplification product; S102.4 uses bioinformatics analysis to perform quality control, sequence assembly, and removal of non-microbial sequences from sequencing data, and merges sequences based on 100% sequence similarity.

[0010] Based on the above technical solution, the preprocessing further includes: denoising the DNA abundance data using the zero-inflated negative binomial distribution latent variable model ZINB-WaVE to obtain the environmental DNA abundance data; and numericalizing the DNA sequence data to obtain the environmental DNA sequence data.

[0011] Based on the above technical solution, the numericalization specifically involves setting the length of each sequence in the DNA sequence data to a fixed value of 200bp; using one-hot encoding to convert the DNA sequence into a numerical form, and obtaining a three-dimensional array of 2487 × 200 × 4 as the environmental DNA sequence data.

[0012] Based on the above technical solution, the dimension of the sequence feature is further 2487 × 64.

[0013] Based on the above technical solution, the weighted average is further described as follows: according to the abundance of the ASV sequence features in the environmental DNA sample, the ASV sequence features are weighted and averaged to obtain a fused sequence feature vector, and the sequence feature vector constitutes the fused sequence feature data of the environmental DNA sample.

[0014] Based on the above technical solution, the fused sequence feature data is further defined as a 27 × 64 fused sequence feature matrix.

[0015] Based on the above technical solution, TotalVI further reduces the sample dimension to 100 dimensions.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention provides a microbial monitoring method that fully utilizes abundance data and DNA sequence data to improve the accuracy of subsequent analysis. It can accurately reveal the composition and diversity changes of eukaryotic microbial communities in different marine dumping areas, providing a flexible and reliable technical platform for ecological monitoring of marine pollution areas.

[0017] 2. After acquiring environmental DNA data samples, this invention divides the environmental DNA data into environmental DNA abundance data and environmental DNA sequence data. Considering the high sparsity and large number of zero values ​​in environmental DNA abundance data, an encoder is used to denoise the environmental DNA data. This not only improves data quality but also provides a reliable foundation for subsequent ecological inference. It can better capture the structural information of environmental DNA data and effectively reduce the impact of high noise and sparsity.

[0018] 3. In addition to processing abundance data, this invention also considers that environmental DNA sequence data contains species information. It performs fixation, transformation, encoding, and fusion processing on environmental DNA sequence information, fixing sequence length to eliminate bias and adapting it to deep learning models, achieving ecological information integration from ASV to the sample level. While obtaining more accurate dimensionality reduction data, it better preserves the information of environmental DNA data and better represents the distribution of environmental DNA data, resulting in more accurate species information and thus enabling a more precise assessment of species diversity patterns. The overall workflow realizes a standardized pipeline from data preprocessing to biological applications, capable of efficiently processing large-scale datasets, especially suitable for complex ecosystems disturbed by human activities, such as marine dumping grounds, providing an efficient and accurate method for monitoring microbial diversity.

[0019] 4. This invention optimizes the detection process by measuring the physicochemical properties of water bodies and using environmental DNA technology and data processing techniques. It denoises the environmental DNA abundance data using a zero-expansion negative binomial distribution latent variable model and combines the abundance and sequence of the environmental DNA data as inputs. It then uses a multimodal variational autoencoder for dimensionality reduction, which better preserves the information in the environmental DNA data, improves the resolution of species diversity and community structure research, identifies rare species more quickly, and obtains more accurate visualization results. Attached Figure Description

[0020] To more clearly illustrate the embodiments of the present invention, the accompanying drawings involved in the embodiments will be briefly described below.

[0021] Figure 1 This is a schematic diagram of the eukaryotic microorganism monitoring process in marine dumping areas according to the present invention; Figure 2 This is a schematic diagram illustrating the steps of the method for monitoring and analyzing eukaryotic microorganisms in marine dumping areas according to the present invention; Figure 3 This is a diagram of the species composition of Krona before data processing in this invention; Figure 4 The species composition of Krona after treatment by the method of this invention; Figure 5 This is a species accumulation curve diagram in an embodiment of the present invention; Figure 6 This is an NMDS analysis diagram from an embodiment of the present invention. Detailed Implementation

[0022] This invention provides a method for monitoring eukaryotic microorganisms in marine dumping areas based on environmental DNA technology. It addresses the problems of high noise and sparsity in environmental DNA abundance data through optimized sampling strategies, high-throughput sequencing, and data processing and analysis techniques, obtaining more accurate dimensionality-reduced data and providing a more comprehensive, accurate, and efficient method for monitoring marine eukaryotic microbial communities. Figure 1 As shown, it includes the following steps: S101, Collect environmental DNA samples; S102, Sequencing analysis and preprocessing of environmental DNA samples to obtain environmental DNA abundance data and environmental DNA sequence data; S103, input environmental DNA sequence data into the CNN model to obtain the sequence features of the DNA sequence data; S104, the ASV sequence features in the sequence features are weighted and averaged according to the DNA abundance data to obtain the fusion sequence feature data; S105, input environmental DNA abundance data and fusion sequence feature data into the multimodal variational autoencoder TotalVI to obtain environmental DNA data; S106 involves species classification annotation and diversity analysis of environmental DNA data.

[0023] In some embodiments, environmental DNA samples include water samples and sediment samples from the marine dumping area.

[0024] The sampling scope covers the marine dumping area, with sampling points covering different regions of the marine dumping area; different regions include different horizontal regions and different vertical regions; vertical regions cover the areas where water and sediment are located; in order to comprehensively reflect the distribution of microorganisms in the water.

[0025] The water samples were filtered through a 0.45 μm mixed cellulose acetate membrane to obtain a microporous membrane containing aquatic environmental DNA and the filtered water sample. The physicochemical properties of the filtered water sample were then determined. The microporous membrane was stored in centrifuge tubes under refrigeration to prevent DNA degradation. After on-site collection, the samples were immediately transported to the laboratory for DNA extraction to ensure data reliability.

[0026] The environmental DNA extraction process for collecting environmental DNA samples includes: folding and cutting the filter membrane and placing it into a centrifuge tube; extracting DNA using the membrane filtration method; determining molecular size using 0.8% agarose gel electrophoresis; and quantifying the DNA using Nanodrop.

[0027] In some embodiments, sequencing analysis includes the following steps: S102.1, Extract DNA from environmental DNA samples; S102.2, PCR amplification of extracted DNA was performed using 18S rRNA V9 region-specific primers; S102.3, perform high-throughput sequencing on the products of PCR amplification; S102.4 uses bioinformatics analysis to perform quality control, sequence assembly, and removal of non-microbial sequences from sequencing data, and merges sequences based on 100% sequence similarity.

[0028] The 18S rRNA V9 region-specific primers include the following sequences: Euk1391f:GTACACACCGCCCGTC,EukBr,TGATCCTTCTGCAGGTTCACCTAC.

[0029] The quality control, sequence assembly, and removal of non-microbial sequences from sequencing data through bioinformatics analysis include: The amplification products were purified, recovered, and quantified. Final fragment selection and purification were performed on the DNA library. After library quality control, qualified libraries were subjected to high-throughput sequencing on the Illumina MiSeq platform, and the generated raw data was output in FASTQ format. The QIIME2 dada2 workflow was used to process the data, removing sequencing errors, chimeras, and non-microbial ASV sequences. Sequences were then merged based on 100% sequence similarity.

[0030] In some embodiments, preprocessing includes: denoising the DNA abundance data using a zero-inflated negative binomial distribution latent variable model ZINB-WaVE to obtain environmental DNA abundance data; and quantifying the DNA sequence data to obtain environmental DNA sequence data.

[0031] Specifically, the ZINB-WaVE model, a latent variable model with a zero-inflated negative binomial distribution, was used to denoise the DNA abundance data. 1. Set initialization potential variables Dimensions ( Follows a standard normal distribution Abundance matrix .

[0032] 2. The probability function of the ZINB distribution: .

[0033] In the formula: Let i be the abundance vector of sample i. The probability of zero inflation is zero. is the mean parameter of the negative binomial distribution. For discrete parameters, It is the negative binomial probability mass function.

[0034] Inflation probability: .

[0035] In the formula: The Logistic function is a non-linear transformation function. (x) = , For the zero-inflation model, T represents the total number of iterations. Zero inflation probability The regression coefficients.

[0036] Negative binomial distribution probability mass function: .

[0037] In the formula: For discrete parameters, The mean parameter, It is non-zero abundance.

[0038] Mean of a binomial distribution: Discrete parameters .

[0039] In the formula: Let T represent the known features of the sample, and T be the total number of iterations. It is the negative binomial mean parameter. For latent variable effects, Let i be the vector of latent variables for sample i.

[0040] Calculate the posterior expectation of the latent variables: Non-zero probability: .

[0041] Conditional probability of a negative binomial distribution: When y i >0, ,in .

[0042] In the formula: Let i be the abundance vector of sample i. Let Wi be the zero inflation probability, T be the total number of iterations, and β be the zero inflation probability. The regression coefficients, where NB is the negative binomial probability mass function. is the mean parameter of the negative binomial distribution. These are discrete parameters.

[0043] Posterior distribution of latent variables: Assume that zi (the latent variable vector of sample i, representing latent variables of technical noise) follows a normal distribution: Posterior mean: .

[0044] In the formula: Let i be the abundance vector of sample i. For the known features of the sample, For covariates in the zero-inflation model, For latent variable effects, For discrete parameters, Let be the mean parameter of the negative binomial distribution, and T be the total number of iterations. It is the negative binomial mean parameter. For latent variable z i The prior variance, Representing latent variables The reciprocal of the prior variance.

[0045] Posterior covariance: .

[0046] In the formula: , Let i be the abundance vector of sample i. For the known features of the sample, For the zero-inflation model, T represents the total number of iterations. For latent variable effects, This represents the covariance of the prior distribution of the latent variables. This is the transpose of the latent variable vector.

[0047] Update parameters: .

[0048] if =0, then =1, if >0, then =0. It is a zero-valued indicator variable.

[0049] In the formula: Let i be the abundance vector of sample i. The zero inflation probability is given by N, where N is the sample size.

[0050] Using the above-obtained and Construct observations: .

[0051] In the formula: is the mean parameter of the negative binomial distribution. Let T represent the known features of the sample, and T be the total number of iterations. It is the negative binomial mean parameter. For latent variable effects, the latent variable vector of sample i; Let i be the latent variable vector representing the latent variables of technical noise; This is the transpose of the latent variable vector.

[0052] Expected value: .

[0053] In the formula: For the observed values, Let i be the abundance vector of sample i. Let T represent the known features of the sample, and T be the total number of iterations. It is the negative binomial mean parameter. For latent variable effects, The latent variable of sample i The posterior mean vector, yes The transpose operation.

[0054] variance: .

[0055] In the formula: For the observed values, Let be the abundance vector of sample i, and T be the total number of iterations. For latent variable effects, This is the transpose of the latent variable vector.

[0056] Fitting using weighted least squares method: .

[0057] In the formula: The estimated value of the observed value. Given the known features of the sample, T is the total number of iterations, α is the negative binomial mean parameter, and γ is the effect of the latent variable. Iterative updates are performed. and Until convergence, The latent variable of sample i The posterior mean vector, yes The transpose operation.

[0058] .

[0059] In the formula: N is the sample size. For discrete parameters, Let i be the abundance vector of sample i. Let t be the observed value, t be the iteration number, and T be the total number of iterations.

[0060] 3. After model simulation, latent variables Characterization technique noise, Characterize the trend of true species abundance after removing technical noise; Based on the fitted model, the expected abundance value after denoising is calculated: .

[0061] In the formula: , For the known features of the sample, It is the negative binomial mean parameter. The probability of zero inflation is zero. For abundance vectors, Characterization technique noise.

[0062] The row and column dimensions of the denoised data matrix are the same as the original data, but each element is replaced with a predicted value after removing technical noise, thus reducing the impact of technical errors while preserving ecological signals. Finally, the denoised environmental DNA abundance data is obtained, and the size of the abundance data remains unchanged.

[0063] In some embodiments, numericalization specifically involves setting the length of each sequence in the DNA sequence data to a fixed value of 200 bp; using one-hot encoding to convert the DNA sequence into a numerical form, resulting in a three-dimensional array of 2487 × 200 × 4 as environmental DNA sequence data.

[0064] The environmental DNA sequence data obtained after quantifying the DNA sequence data includes: 1. Set the length of each sequence in the DNA sequence data to 200 bp: For sequences longer than 200bp, prioritize preserving high-quality regions and perform end trimming; for sequences shorter than 200bp, pad the sequence to 200bp and mark the padding region.

[0065] 2. One-hot encoding is used on the 200bp DNA sequence data to encode the sequence into a 200×4 matrix. Each row represents the position of a base, and each column represents the presence or absence of A / T / C / G. One-hot encoding rules: A→[1,0,0,0] T→[0,1,0,0] C→[0,0,1,0] G→[0,0,0,1] N (undefined base) → [0,0,0,0] or [0.25,0.25,0.25,0.25] Perform one-hot encoding on each sequence according to its position to obtain: ASV 1 →[[1,0,0,0] ← A [0,1,0,0] ← T [0,0,1,0] ← C ... ]∈ R^{200×4} After encoding, a three-dimensional array will be obtained: , where n is the number of ASVs, 200 is the length of each ASV sequence, and 4 is the coding dimension of the four bases.

[0066] In some embodiments, the dimension of the sequence feature is 2487 × 64.

[0067] Among them, the environmental DNA sequence matrix The sequence [n, 200, 4] is used as the input to the CNN model. After dimensional adjustment to [n, 4, 200], it passes through 4 input channels and 32 output channels, and the sequence length is halved to [n, 32, 100]. Then, it passes through 32 input channels and 64 output channels, and the sequence length is halved again to [n, 64, 50]. This results in a 64-dimensional feature vector. Finally, the shape of the environmental DNA sequence data feature matrix output by the CNN is [n, 64].

[0068] In some embodiments, the weighted average is specifically performed by weighting the ASV sequence features according to the abundance of ASV sequence features in the environmental DNA sample to obtain a fused sequence feature vector, which constitutes the fused sequence feature data of the environmental DNA sample.

[0069] In some embodiments, the fused sequence feature data is a 27 × 64 fused sequence feature matrix.

[0070] Specifically, the ASV sequence features in the sequence features are weighted and averaged according to DNA abundance data to obtain the fusion sequence feature data as follows: The fused sequence feature vector (1×64) for each sample is obtained by combining the ASV abundance vector (1×n) of the sample with the sequence features of all ASVs (n×64). That is, for the i-th sample, the abundance vector of that sample is combined with the sequence feature vector of the i-th sample. (Of length n), use this vector to compare the sequence matrix (n×64) weighted average , This represents the overall representation of all ASV sequence information contained in sample i, weighted by abundance. This process is then repeated for all N samples to obtain the fused sequence feature matrix. (N × 64).

[0071] Based on the above technical solution, TotalVI further reduces the sample dimension to 100 dimensions.

[0072] The environmental DNA abundance data and fusion sequence feature data are input into the multimodal variational autoencoder TotalVI to obtain environmental DNA data, including: Denoising environmental DNA abundance data Perform CPM standardization and logarithmic transformation.

[0073] CPM standardization: .

[0074] In the formula: This represents the count of the j-th ASV in the i-th sample, where n is the number of ASVs.

[0075] Logarithmic transformation: .

[0076] Environmental DNA sequence data after fusion Perform Z-score standardization: .

[0077] In the formula: Let the mean of the k-th dimension of the sequence features be denoted as . Let N be the standard deviation of the k-th dimension of the sequence features, where k is the feature and N is the sample size. Let represent the feature value of the k-th dimension sequence of the i-th sample.

[0078] Model architecture construction: Will Mapping to latent variable distribution Let the latent posterior distribution of the given abundance data y be: ;in .

[0079] In the formula: As a latent variable, The mean of the latent variables output by the abundance encoder. The standard deviation of the latent variable output by the abundance encoder. It is an abundance data encoder that compresses high-dimensional abundance data into a low-dimensional latent space through nonlinear transformation.

[0080] Will Mapping to the latent posterior distribution of a given sequence feature d : .

[0081] In the formula: The mean of the latent variables output by the sequence feature encoder. The standard deviation of the latent variable output by the sequence feature encoder. As a latent variable, For sequence feature vectors, It is a sequence feature encoder.

[0082] The prior hypothesis of the latent variable is that it follows a standard normal distribution. .

[0083] In the formula: As a latent variable, The symbol for the multivariate normal distribution (Gaussian distribution) is... It is an identity matrix.

[0084] The decoder uses a negative binomial distribution to model latent variables. The probability distributions are mapped to abundance data and sequence data, respectively.

[0085] Training process and loss function: Variational lower bound of abundance modes: .

[0086] In the formula: For variational posterior distribution Expectations Indicates that given a latent variable Under these conditions, abundance data The probability distribution, Given abundance data The latent variable posterior distribution, As a latent variable, Kullback-Leibler divergence measures the difference between two probability distributions.

[0087] Multimodal variational lower bound: .

[0088] In the formula: , This is the variational lower bound for the sequence feature modes.

[0089] Variational lower bound of characteristic modes: .

[0090] In the formula: For feature reconstruction values, These are priors for latent variables.

[0091] Total loss: .

[0092] In the formula: The value is linearly increased from 0 to 1 during the initial training phase. , For KL divergence annealing weights, This represents the current iteration number. This represents the total number of annealing cycles. For latent variables prior, This is the KL divergence term between the variational posterior distribution and the prior distribution.

[0093] Latent variable inference and dimensionality reduction: Output the hidden variable z: .

[0094] Where: μ and Here, represents the mean and variance of the encoder output, N is the sign of the multivariate normal distribution (Gaussian distribution), and I is the identity matrix. It is standard normal noise.

[0095] Weighted average fusion: , .

[0096] In the formula: and These are the sampling results from the abundance and feature encoders, respectively, with the weights being the reciprocal of the variance. The standard deviation of the latent variable output by the sequence feature encoder. The standard deviation is the latent variable output by the abundance encoder.

[0097] Finally, variational inference was used to analyze the high-dimensional data (abundance) of N samples. (+64-dimensional features) Dimensionality reduction to 100-dimensional latent space yields 100 dimensions of the sample: N is the number of samples.

[0098] The present invention will be described in detail below with reference to the embodiments. However, the implementation of the present invention is not limited thereto. Obviously, the embodiments described below are only some embodiments of the present invention. For those skilled in the art, other similar embodiments can be obtained without creative effort and all fall within the protection scope of the present invention.

[0099] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0100] It should be understood that although the terms first, second, third, etc., may be used to describe... in the embodiments of the present invention, these... should not be limited to these terms. These terms are only used to distinguish... For example, first... can also be referred to as second... without departing from the scope of the embodiments of the present invention, and similarly, second... can also be referred to as first...

[0101] Unless otherwise specified, the experimental methods described in the following examples are conventional methods; the reagents and materials used are commercially available unless otherwise specified.

[0102] Example 1 This embodiment focuses on a marine dumping site and collects environmental DNA samples.

[0103] The specific procedure for collecting environmental DNA samples from water and sediment in marine dumping areas and extracting DNA from these samples is as follows: Twenty-seven sampling points were established in the marine dumping area, covering different regions of the area. These included five sampling points at one coordinate in the port area, four sampling points at one coordinate in the control area, and 18 sampling points at three coordinates in the dumping area, to comprehensively reflect the distribution of microorganisms in the water. One liter of water sample was collected from the lower layers of the surface at each sampling point using a water sampler. Before use, the water sampler was disinfected with a 10% sodium hypochlorite solution and rinsed three times with in-situ seawater. All sampling personnel wore disposable rubber gloves, which were changed after each sampling. Samples were stored in brand-new, sterile, sealed wide-mouth bottles.

[0104] A 0.45μm mixed cellulose acetate membrane was used for on-site filtration with a vacuum pump or transported to the laboratory under refrigeration for filtration. The membrane was placed in a 10ml centrifuge tube and stored at -20℃. The physicochemical properties of the filtered water samples were analyzed, including: water temperature, salinity, pH, turbidity, conductivity, dissolved oxygen, heavy metals, nutrients, and soluble organic carbon.

[0105] After on-site collection and pretreatment, the DNA was immediately transported to the laboratory for extraction to ensure data reliability. The DNA extraction steps included: folding and cutting the microporous membrane and placing it into a 10 ml centrifuge tube, and then extracting the DNA using a kit method. The extracted environmental DNA was quantified using Nanodrop, and its molecular size was determined by 0.8% agarose gel electrophoresis.

[0106] Example 2 In this embodiment, the environmental DNA samples obtained in Example 1 were sequenced and preprocessed to obtain environmental DNA abundance data and environmental DNA sequence data.

[0107] The specific process is as follows: 1. Select primers specific to the 18S rRNA V9 region as templates for extracted DNA PCR amplification. These primers include the following sequences: Euk1391f: GTACACACCGCCCGTC, EukBr, TGATCCTTCTGCAGGTTCACCTAC. The PCR amplification program involves pre-denaturation, denaturation, annealing, extension, cycling, and final extension to exponentially amplify the DNA fragment.

[0108] 2. Perform high-throughput sequencing on the PCR amplification products, and use bioinformatics analysis to perform quality control, sequence assembly, and removal of non-microbial sequences on the sequencing data. The obtained sequences are then merged according to 100% sequence similarity.

[0109] High-throughput sequencing was performed using the Illumina MiSeq platform, and the raw data was output in FASTQ format. The dada2 workflow of QIIME2 was used to process the data, removing sequencing errors, chimeras, and non-microbial ASV sequences.

[0110] 3. Generate environmental DNA abundance data and environmental DNA sequence data: The environmental DNA sequence data contains 2487 ASV operational taxonomic units; the environmental DNA abundance data contains abundance data of 2487 ASV operational taxonomic units × 27 samples.

[0111] The abundance matrix is: The behavioral environment DNA sequence data contains representative sequences of 2487 ASV operational taxonomic units, listed in 27 samples; the sequence matrix is... .

[0112] 4. Input environmental DNA abundance data into the single-cell variational inference model for denoising. The specific process is the same as above: The environmental DNA abundance matrix was obtained as follows .

[0113] 5. For sequences longer than 200bp, prioritize preserving high-quality regions and perform end trimming; for sequences shorter than 200bp, pad the sequence to 200bp to address potential issues caused by sequence length deviations.

[0114] 6. Encode the sequence obtained in S6 using one-hot encoding. One-hot encoding rules: The sequence is encoded into a 200×4 matrix, where each row represents the position of a base and each column represents the presence or absence of A / T / C / G. A→[1,0,0,0] T→[0,1,0,0] C→[0,0,1,0] G→[0,0,0,1] N (an undefined base) → [0,0,0,0] or [0.25,0.25,0.25,0.25] Perform one-hot encoding on each sequence according to its position to obtain: ASV 1 → [1,0,0,0] ← A [0,1,0,0] ← T [0,0,1,0] ← C ... 】∈ R^{200×4} After encoding, a three-dimensional array will be obtained: , where 2487 is the number of ASVs, 200 is the length of each ASV sequence, and 4 is the coding dimension of the four bases.

[0115] Example 3 This embodiment uses the environmental DNA sequence data obtained in Example 2 as the object, and obtains the sequence features of the DNA sequence data through a CNN model.

[0116] The specific process is as follows: The encoded environmental DNA sequence data in a three-dimensional array is used as input to a CNN model to extract sequence features from the environmental DNA sequence data, specifically including: Environmental DNA sequence matrix [2487, 200, 4] is used as the input to the CNN model. The dimensions are adjusted to [2487, 4, 200]. After passing through 4 input channels and 32 output channels, the sequence length is halved to obtain [2487, 32, 100]. After passing through 32 input channels and 64 output channels, the sequence length is halved again to obtain [2487, 64, 50]. This results in a 64-dimensional feature vector. Finally, the shape of the environmental DNA sequence data feature matrix output by the CNN is [2487, 64].

[0117] Example 4 This embodiment uses the sequence features obtained in Example 3 as the object, and performs a weighted average based on DNA abundance data to obtain fusion sequence feature data.

[0118] The specific process is as follows: By combining environmental DNA abundance data and environmental DNA sequence data, the sequence features of all ASVs in each sample are weighted and averaged according to their abundance in the sample to obtain a fused sequence feature vector, which specifically includes: The ASV abundance vector (1×2487) of the sample is combined with the sequence features of all ASVs (2487×64) to obtain the fused sequence feature vector (1×64) of each sample.

[0119] That is, for the i-th sample, combine the abundance vector of that sample. (Length is 2487), use this vector to pair the sequence matrix (2487×64) weighted average , This represents the overall representation of all ASV sequence information contained in sample i, weighted by abundance. This process is then repeated for all N samples to obtain the fused sequence feature matrix. (27 × 64).

[0120] Example 5 This embodiment uses the environmental DNA abundance data and fusion sequence feature data obtained in the above embodiments as objects, and obtains environmental DNA data through the multimodal variational autoencoder TotalVI.

[0121] The specific process is as follows: The denoised environmental DNA abundance data and the fused sequence feature data are used together as input to the multimodal variational autoencoder (TotalVI) to reduce the dimension of the samples to 100 dimensions. The specific process is the same as above.

[0122] The difference lies in the denoised environmental DNA abundance data. Standardize CPM: (2487 ​​represents the number of ASVs) and logarithmic transformation Processing; Performing... on the fused environmental DNA sequence data Perform Z-score standardization: .

[0123] Finally, variational inference was used to analyze the high-dimensional data (abundance) of the 27 samples. (+ 64-dimensional features) Dimensionality reduction to a 100-dimensional latent space yields the 100-dimensional sample: ,27 is the sample size.

[0124] Example 6 This embodiment uses the environmental DNA data obtained in the above embodiments as the object, compares the processed ASV representative sequences with the reference database to determine their taxonomic position (kingdom → phylum → class → order → family → genus → species), and performs species classification annotation. It also conducts diversity analysis by revealing the core differences in microbial community structure, including two dimensions: α diversity (intra-sample diversity) and β diversity (inter-sample diversity), to reveal the composition and diversity of eukaryotic microorganisms in marine dumping areas.

[0125] Environmental samples from marine dumping areas are complex and diverse. Traditional environmental DNA techniques used for species diversity analysis rely solely on environmental DNA abundance data to analyze community composition and structure, neglecting the species information hidden within the environmental DNA sequence data. In particular, species with low abundance, such as Platyhelminthes, Ctenophora, Evosea, Endomyxa, and Tubulinea, are often undetectable, failing to fully utilize multi-source information and leading to inaccuracies in subsequent analyses.

[0126] The results obtained using the method of this invention are as follows: Figure 3 and 4 As shown, low-abundance species that are easily missed were detected, such as *Platyhelminthes*, *Ctenophora*, *Evosea*, *Endomyxa*, and *Tubulinea*, which accounted for only 0.009%, 0.0007%, 0.0005%, 0.0004%, and 0.0001% of the abundance of species detected in environmental DNA, respectively. This indicates that for some low-abundance species, the method of this invention is more sensitive than traditional methods, better preserves data, and thus makes the analysis results more accurate.

[0127] like Figure 3 and 4 The diagram shows the species composition of Krona. A curve is constructed by repeatedly sampling from a large number of samples and counting the number of observed species (ASVs) in each sample. This curve is used to measure and predict the increase in species richness in a community as the sample size increases, and is widely used to determine whether the sample size is sufficient and to estimate community richness.

[0128] like Figure 5 As shown in the species accumulation curve, when the sample size is already large, the total number of ASVs in the community will no longer increase significantly with the addition of new samples, and the curve will tend to flatten out, indicating that the sample size is sufficient to reflect the species composition of the community. This demonstrates that the sequencing data obtained using the method of this invention is sufficient to represent the microbial community of the sample and is reliable for subsequent α-diversity comparisons and interpretations.

[0129] like Figure 6The NMDS analysis plot shown illustrates that a lower stress value is generally better; a value less than 0.2 is considered more reliable. In this study, the stress value was 0.0293, indicating reliable results that accurately reflect the true relationships between samples. This plot also visualizes the overall similarity and differences between samples. Clustering of points within the same group indicates similar characteristics within that group, while significant separation and differences exist between groups. Therefore, the method described in this invention provides a clearer and more reliable representation of similarities and differences between species, making it suitable for visual analysis of differences in data related to ecological communities and microbiomes.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring eukaryotic microorganisms in a marine dumping area based on environmental DNA technology, characterized by, The method comprises the following steps: S101, collecting an environmental DNA sample; S102, performing sequencing analysis and preprocessing on the environmental DNA sample to obtain environmental DNA abundance data and environmental DNA sequence data; S103, inputting the environmental DNA sequence data into a CNN model to obtain sequence features of the DNA sequence data; S104, weighting and averaging ASV sequence features in the sequence features according to the DNA abundance data to obtain fused sequence feature data; S105, inputting the environmental DNA abundance data and the fused sequence feature data into a multi-modal variational autoencoder TotalVI to obtain environmental DNA data; S106, performing species classification annotation and diversity analysis on the environmental DNA data.

2. The method according to claim 1, wherein, The environmental DNA sample comprises a water sample and a sediment sample of a marine dumping area.

3. The method according to claim 1, wherein, The sequencing analysis comprises the following steps: S102.1, extracting DNA in the environmental DNA sample; S102.2, performing PCR amplification on the extracted DNA using 18S rRNA V9 region-specific primers; S102.3, performing high-throughput sequencing on the PCR amplification product; S102.4, performing quality control, sequence splicing and removing non-microbial sequences on the sequencing data through bioinformatics analysis, and merging sequences according to 100% sequence similarity.

4. The method according to claim 1, wherein, The preprocessing comprises: denoising the DNA abundance data using a latent variable model ZINB-WaVE of zero-inflated negative binomial distribution to obtain the environmental DNA abundance data; and obtaining the environmental DNA sequence data after numerically processing the DNA sequence data.

5. The method according to claim 4, wherein, The numerical processing is specifically: setting the length of each sequence in the DNA sequence data as a fixed value of 200 bp; and converting the DNA sequence into a numerical form using One-hot encoding, so that a three-dimensional array of 2487 × 200 × 4 is obtained as the environmental DNA sequence data after encoding.

6. The method according to claim 1, wherein, The dimension of the sequence features is 2487 × 64.

7. The method according to claim 1, wherein the method is characterized by, The weighting and averaging is specifically: weighting and averaging the ASV sequence features according to the abundance of the ASV sequence features in the environmental DNA sample to obtain a fused sequence feature vector, and the sequence feature vector constitutes the fused sequence feature data of the environmental DNA sample.

8. The method according to claim 1, wherein, The fused sequence feature data is a fused sequence feature matrix of 27 × 64.

9. The method according to claim 1, wherein the method is characterized by, The TotalVI reduces the dimension of the sample to 100.

Images