Environmental DNA data dimension reduction method and system based on hierarchical auto-encoder
Through a hierarchical autoencoder-based method, including a combination of depth counting autoencoder and autoencoder, the problem of information loss in environmental DNA data dimensionality reduction is solved, and more accurate dimensionality reduction and visualization effects are achieved.
Patent Information
- Application Number
- CN202411092850.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-09
AI Technical Summary
The prior art is difficult to effectively reduce environmental DNA data in dimensionality, resulting in serious information loss, especially when the data dimension is large, the number of classification units is much larger than the number of samples, and there are a large number of false zero values.
The hierarchical autoencoder-based method is adopted, including data preprocessing, denoising, length equalization, encoding and differential autoencoder dimensionality reduction, and the environmental DNA data is denoised and compressed through the depth counting autoencoder and autoencoder. The zero-expansion negative binomial distribution model and differential autoencoder were used for dimensionality reduction.
It effectively reduces the high noise and sparsity effects of environmental DNA data, retains the structural information of the data, and obtains more accurate dimensionality reduction results and visualization effects.
Smart Images

Figure CN119003988A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of bioinformatics and data processing technology, and in particular to a method and system for reducing the dimension of environmental DNA data based on a hierarchical autoencoder. Background Art
[0002] The impact of human activities has led to the destruction of ecosystems, serious degradation of ecological functions and impacts on human health. Convenient, accurate and reliable ecological monitoring technology is essential to protecting the ecological environment. Environmental DNA technology is a molecular-based efficient biodiversity monitoring method with great application potential in ecological environmental restoration and natural resource protection. Environmental DNA is DNA extracted from environmental samples (water, soil, sediment, air, mixtures, etc.). It is a mixture of DNA from various organisms. By extracting DNA from the environment, biodiversity can be detected. However, due to the large dimension and complexity of environmental DNA data, the acquired environmental DNA data must be subjected to dimensionality reduction and other operations to obtain the useful information contained in the environmental DNA data.
[0003] At present, most of the methods for reducing the dimensionality of environmental DNA data are some traditional dimensionality reduction methods, such as principal component analysis (PCA), t-distributed random neighbor embedding (t-sne), non-metric multidimensional scaling (NMDS), and uniform manifold approximation (UMAP). However, unlike other data, the number of classification units in the spectrum matrix of environmental DNA data is much larger than the number of samples, and due to the technical limitations of sequencing, there is a high probability that non-zero values will be judged as zero, resulting in a large number of false zero values in the data. In view of the above characteristics of the data, traditional dimensionality reduction methods cannot well preserve the structure of the data, and the environmental DNA data loses information seriously after dimensionality reduction.
[0004] Therefore, there is a need for a data dimensionality reduction method for environmental DNA that can better preserve the information of environmental DNA data and obtain more accurate dimensionality reduction data. Summary of the invention
[0005] In view of this, the present invention provides an environmental DNA data dimensionality reduction method and system based on a hierarchical autoencoder to reduce the high noise and sparsity of environmental DNA data, thereby obtaining more accurate dimensionality reduction data.
[0006] To this end, the present invention provides the following technical solutions:
[0007] A method for reducing the dimension of environmental DNA data based on a hierarchical autoencoder, the method comprising the following steps:
[0008] S1. Data preprocessing, including:
[0009] Divide environmental DNA data into environmental DNA abundance data and environmental DNA sequence data;
[0010] Filter out all zero rows of the environmental DNA abundance data, and perform Min-Max normalization and log-2 logarithmic transformation to obtain pre-processed environmental DNA abundance data and environmental DNA sequence data;
[0011] S2, inputting the pre-processed environmental DNA abundance data into the deep counting autoencoder DCA for denoising to obtain denoised environmental DNA abundance data;
[0012] S3. Perform length equalization and translation on environmental DNA sequence data, including:
[0013] The environmental DNA sequence data were padded with the nucleotide code N in the IUPAC nucleotide code to equalize to the same length;
[0014] Each typical base of the environmental DNA sequence after equalization: A, C, T, G, and IUPAC ambiguous code is translated into four typical bases: A, T, C, G, and four-dimensional probability distribution, and the nucleotide sequence is represented as a two-dimensional array of sequence length multiplied by 4;
[0015] S4, representing the sequence information of the sample; combining the two-dimensional array representing the nucleotide sequence information obtained in S3 with the presence and absence of each operational classification unit in each sample to represent a three-dimensional array;
[0016] The presence of each operational taxonomic unit: nucleotide sequence information of the sequence translation of the detected OTU;
[0017] The missing of each operational taxonomic unit: if no OTU is detected, it is a zero matrix;
[0018] S5, inputting the three-dimensional array representing the sample sequence information into the autoencoder AE to compress the environmental DNA sequence data;
[0019] S6, using a differential autoencoder VAE to reduce dimensionality, using the denoised environmental DNA abundance data obtained in S2 and the compressed and encoded environmental DNA sequence data obtained in S5 as inputs of the differential autoencoder VAE, further reducing the dimension of the sample to a two-dimensional latent space;
[0020] S7. Draw a visualization result diagram in two-dimensional space.
[0021] Furthermore, the pre-processed environmental DNA abundance data is input into a deep counting autoencoder DCA for denoising to obtain denoised environmental DNA abundance data, including:
[0022] The specific formula of the deep counting autoencoder DCA architecture is:
[0023] E = ReLU(XW E );
[0024] B = ReLU (EW B );
[0025] D = ReLU(BW E );
[0026] M=diag(s j )exp r(DW μ );
[0027] Π=sigmoid(DW π );
[0028] θ=exp(DW θ );
[0029] where E, B, and D represent the encoder layer, latent layer, and decoder layer, respectively; M, θ, and Π represent the matrix forms of mean, dispersion, and loss probability estimation, respectively; s j is the size factor; the activation function chosen for the mean and dispersion is exponential; the activation function for the loss probability estimate is sigmoid, which represents the dropout probability;
[0030] The zero-inflated negative binomial distribution ZINB is used to describe the distribution of environmental DNA data, where the zero-inflated negative binomial distribution ZINB is:
[0031]
[0032] ZINB(x;π,μ,θ)=πδ(x)+(1-π)NB(x;μ,θ);
[0033] where π, μ, and θ are the parameters of the ZINB distribution, representing the probability, mean, and dispersion of loss events, respectively; x represents the raw read counts, and the autoencoder based on the ZINB model estimates the parameters v, μ, and θ.
[0034] Furthermore, the loss function of the deep counting autoencoder DCA is the negative logarithm of the ZINB likelihood, and the specific formula is as follows:
[0035] L ZINB =-log(ZINB(X|π,μ,θ)).
[0036] Furthermore, the nucleotide code N padded to balance the sequence length is translated as [0.25, 0.25, 0.25, 0.25].
[0037] Furthermore, the three-dimensional array representing the sample sequence information is input into the autoencoder AE to compress the environmental DNA sequence data, specifically:
[0038] The autoencoder AE includes a fully connected layer and a potential layer;
[0039] Each fully connected layer uses ReLU activation and adopts a mirror architecture as a decoder;
[0040] The loss function of the autoencoder AE is binary cross entropy:
[0041] L AE =-y log(p(y))-(1-y)log(1-p(y));
[0042] Here, y is a binary label 0 or 1, and p(y) is the probability that the output belongs to the y label.
[0043] Furthermore, the differential autoencoder VAE includes:
[0044] Input layer, encoder network, latent sampling layer, decoder network, and output layer.
[0045] Furthermore, the potential sampling layer uses a reparameterization technique for sampling and a gradient descent method to optimize the likelihood function.
[0046] Furthermore, the loss function of the differential autoencoder VAE is composed of the VAE loss function: KL divergence, reconstruction error, and binary cross entropy of different combinations of potential encodings of nucleotide sequences:
[0047]
[0048] N(0,1)])+α(-y log(p(y))-(1-y)log(1-p(y)));
[0049] Where x is the abundance input matrix of the differential encoder, is the differential autoencoder abundance output matrix, N(u x , σ x ) represents the distribution of the potential layer z, y is the binary label 0 or 1, p(y) is the probability that the output belongs to the y label, and α is the weight of the loss function accounted for by the binary cross entropy.
[0050] Furthermore, the deep counting autoencoder DCA and the autoencoder AE are trained separately, and then the VAE is trained.
[0051] An environmental DNA data dimensionality reduction system based on a hierarchical autoencoder, the system comprising the following modules: a data preprocessing module, a denoising module, an encoding module, a dimensionality reduction module, and a visualization module;
[0052] The data preprocessing module is used to divide the environmental DNA data into environmental DNA abundance data and environmental DNA sequence data, filter out all zero rows of the abundance data, perform Min-Max normalization and log-2 logarithmic transformation on the abundance data, output the preprocessed environmental DNA abundance data and environmental DNA sequence data, and divide the environmental DNA data into environmental DNA abundance data and environmental DNA sequence data;
[0053] The denoising module is used to denoise the pre-processed environmental DNA abundance data and output the denoised environmental DNA abundance data;
[0054] The encoding module is used to equalize the length of the environmental DNA sequence data, and convert the sequence after equalization into the probability of four typical bases, and represent the nucleotide sequence as a two-dimensional array of the sequence length multiplied by 4; the sequence information is combined with the OTU presence / absence encoding into a three-dimensional array input to the autoencoder AE, and the environmental DNA sequence data is compressed and encoded, and the compressed sample sequence information data is output;
[0055] The dimension reduction module combines the two inputs of the denoised environmental DNA abundance data and the compressed and encoded environmental DNA sequence data to perform dimensionality reduction processing on the environmental DNA data, reducing the sample dimension to a two-dimensional space;
[0056] The visualization module visualizes the data reduced to two dimensions.
[0057] Advantages and positive effects of the present invention:
[0058] Since the distribution of environmental DNA abundance data is highly sparse, the present invention denoises the environmental DNA abundance data through a deep counting autoencoder (DCA), which can better capture the high-dimensional structural information of the environmental DNA data, thereby effectively reducing the impact of high noise and sparsity of the environmental DNA data.
[0059] The deep count autoencoder (DCA) in the present invention uses a loss function based on the zero-inflated negative binomial (ZINB) model to replace the traditional mean square error (MSE) loss function, which can better characterize the distribution of environmental DNA data, reduce the impact of "dropout" events on data dimensionality reduction, and effectively characterize discrete, over-dispersed and zero-inflated count data.
[0060] The present invention simultaneously adopts the abundance and sequence of environmental DNA data and uses a differential autoencoder to perform encoding and dimensionality reduction, which can better retain the information of the environmental DNA data and obtain more accurate dimensionality reduction and visualization results.
[0061] The deep counting autoencoder (DCA) and autoencoder (AE) of the present invention are trained separately, and then the VAE is trained, which reduces the running time of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0063] Figure 1 is a flow chart of the method of the present invention;
[0064] Figure 2 It is a system structure diagram of the present invention;
[0065] Figure 3 This is a dimensionality reduction result diagram of this embodiment. DETAILED DESCRIPTION
[0066] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0067] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0068] The present invention provides a dimensionality reduction method for environmental DNA data based on a hierarchical autoencoder. Considering the high sparsity of the distribution of environmental DNA abundance data, the environmental DNA abundance data is denoised by a deep count autoencoder (DCA), which can better capture the high-dimensional structural information of the environmental DNA data. And for DCA, this method uses a loss function based on a zero-inflated negative binomial (ZINB) model to replace the traditional mean square error (MSE) loss function, which can better characterize the distribution of environmental DNA data, reduce the impact of "dropout" events on data dimensionality reduction, and effectively characterize discrete, over-dispersed and zero-inflated count data. At the same time, the sequence information of the environmental DNA data is encoded by an autoencoder (AE), and the two inputs of the abundance and sequence of the environmental DNA data are combined, and the differential autoencoder (VAE) is used to encode and reduce the dimension. Compared with other methods that only use abundance data, the information of the environmental DNA data can be better retained, and more accurate dimensionality reduction and visualization results can be obtained.
[0069] Combination Figure 1 The process shown in the figure is used to illustrate this method, and the specific steps are described as follows:
[0070] Step 1: Divide the acquired marine environmental DNA dataset into environmental DNA abundance data and environmental DNA sequence data.
[0071] In this embodiment, the environmental DNA sequence data includes representative sequences of 237 OTU operational taxonomic units, and the environmental DNA abundance data includes abundance data of 334 samples × 237 OTU operational taxonomic units.
[0072] Step 2: Filter out all zero rows of environmental DNA abundance data and perform Min-Max normalization and log-2 logarithmic transformation to obtain a preprocessed environmental DNA abundance dataset containing 321 samples × 237 OTU operational taxonomic units.
[0073] Step 3: Input the preprocessed environmental DNA abundance data into the deep count autoencoder (DCA) for denoising to obtain denoised environmental DNA abundance data. The size of the abundance data does not change after denoising.
[0074] Step 4: fill each sequence in the environmental DNA sequence data with the nucleotide code N in the IUPAC nucleotide code to equalize them to the same length. The length of each sequence after equalization is unified to the length of the longest sequence, 151.
[0075] Step 5, translate each typical base (A, C, T, G) and IUPAC ambiguity code of the sequence after equalization into an appropriate four-dimensional probability distribution on four typical bases (A, T, C, G), including uncertain base sequences W and S, etc. For example, base A is translated into [1, 0, 0, 0], W is translated into [0.5, 0, 0, 0.5], and the nucleotide code N added to balance the sequence length is translated into [0.25, 0.25, 0.25, ,0.25]. Each sequence after translation is represented as a 151×4 two-dimensional array.
[0076] Step 6: Combine the two-dimensional array representing the sequence information with the presence / absence of each OTU operational taxonomic unit in each sample and represent it as a three-dimensional array. Each sample in this dataset contains 237 OTU operational taxonomic units, so the sequence information of each sample is represented as a three-dimensional array of size 237×151×4, and the size of the entire dataset is 321×237×151×4.
[0077] Step 7: Input the three-dimensional array representing the sample sequence information into the autoencoder (AE) and compress it to 100 dimensions. The size of the sequence encoding information of the entire data set is 321 (number of samples) × 100 (sequence information dimension).
[0078] The autoencoder (AE) consists of 6 fully connected layers and 1 potential layer. Each fully connected layer uses ReLU activation, adopts a mirror architecture as a decoder, and uses sigmoid in the output layer to make the output within [0,1].
[0079] In step 8, the denoised environmental DNA abundance data obtained in step 3 and the compressed and encoded environmental DNA sequence data obtained in step 7 are used as the input of the differential autoencoder (VAE) to further reduce the dimension of the sample to a two-dimensional latent space.
[0080] Among them, the differential autoencoder (VAE) can be divided into an input layer, an encoder network, a potential sampling layer, a decoder network, and an output layer. The differential autoencoder (VAE) has two input layers, which input abundance data and sequence encoding data respectively. A dropout layer is added after the input layer, and the dropout rate is set to 0.5. This layer sets some features to zero in the encoding stage to improve the performance of model learning. The encoder network has two layers of fully connected neural networks with dimensions of 150 and 50 respectively. No activation function is used for the first layer of fully connected neural network, but its weights are regularized by adding l1 norm, which reduces the sparsity of the model. The second layer is accompanied by ReLU activation, which makes the output of the model sparse and stable. The potential sampling layer has a dimension of 2. The data dimension is compressed to two dimensions through the encoder network, and it is sampled using the reparameterization technique. The sampling result z is used for the subsequent decoder network input. In this layer, the final required dimensionality reduction result of reducing the data to two dimensions can be obtained. The decoder network uses the sampling result of the reparameterization technique to restore the original expression matrix. It also has two layers of fully connected neural networks with hidden unit dimensions of 50 and 150, respectively, and uses ReLU activation. Finally, the output layer uses sigmoid to control the output within [0,1].
[0081] In order to reduce the running time of the model, the deep counting autoencoder (DCA) and autoencoder (AE) in this method are trained separately, and then the VAE is trained. The RMSprop method is selected for VAE training, and the learning rate is set to 0.0001 to ensure convergence of all test data sets. If the training loss does not decrease significantly within 50 times, the training is stopped.
[0082] Step 9: Draw a visualization result graph in two-dimensional space. The result graph after reducing this data set to two-dimensional space is as follows: Figure 3 As shown, the coordinate points are colored according to the logarithm of OTU abundance.
[0083] Corresponding to the environmental DNA data dimensionality reduction method based on hierarchical autoencoder of the present invention, the present invention also provides an environmental DNA data dimensionality reduction system based on hierarchical autoencoder for executing the aforementioned method.
[0084] Combination Figure 2 To illustrate, this system includes: data preprocessing module, denoising module, encoding module, dimensionality reduction module, and visualization module; the specific process is as follows:
[0085] A data preprocessing module is used to divide the environmental DNA data into environmental DNA abundance data and environmental DNA sequence data, filter out all zero rows of the abundance data, normalize and logarithmically transform the abundance data, and output the preprocessed environmental DNA abundance data and environmental DNA sequence data;
[0086] In this embodiment, the environmental DNA sequence data includes representative sequences of 237 OTU operational taxonomic units, and the environmental DNA abundance data includes abundance data of 321 samples×237 OTU operational taxonomic units.
[0087] The denoising module is used to denoise the preprocessed environmental DNA abundance data and output the denoised environmental DNA abundance data. The size of the abundance data does not change after denoising.
[0088] An encoding module, used to equalize the length of environmental DNA sequence data;
[0089] In this embodiment, the length of each sequence is unified to the length of the longest sequence 151, and the sequence after equalization is converted into the probability of four typical bases. Finally, the sequence information is combined with the OTU presence / absence encoding into a three-dimensional array and input into the autoencoder (AE) to achieve compression encoding of environmental DNA sequence data, and output the compressed sample sequence information data, the size of which is 321 (number of samples) × 100 (sequence information dimension);
[0090] The dimension reduction module combines the two inputs of environmental DNA abundance data and environmental DNA sequence data to reduce the dimension of the environmental DNA data to two-dimensional space;
[0091] The visualization module visualizes the data after reducing it to two dimensions.
[0092] Compared with the prior art, the present invention has the following advantages:
[0093] The method and system provided by the present invention take into account the high sparsity of the distribution of environmental DNA abundance data, and denoise the environmental DNA abundance data through a deep counting autoencoder (DCA), which can better capture the high-dimensional structural information of the environmental DNA data and effectively reduce the impact of high noise and sparsity of the environmental DNA data. In addition to using the abundance information of the environmental DNA data, the sequence information of the environmental DNA data is also encoded, and the two inputs of the abundance and sequence of the environmental DNA data are combined, and the differential autoencoder is used to encode and reduce the dimension. Compared with other methods that only use abundance data, it can better retain the information of the environmental DNA data and obtain more accurate dimensionality reduction and visualization results.
[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements 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 reducing the dimension of environmental DNA data based on hierarchical autoencoders, characterized in that: The method comprises the following steps: S1. Data preprocessing, including: Divide environmental DNA data into environmental DNA abundance data and environmental DNA sequence data; Filter out all zero rows of the environmental DNA abundance data, and perform Min-Max normalization and log-2 logarithmic transformation to obtain pre-processed environmental DNA abundance data and environmental DNA sequence data; S2, inputting the pre-processed environmental DNA abundance data into the deep counting autoencoder DCA for denoising to obtain denoised environmental DNA abundance data; S3. Perform length equalization and translation on environmental DNA sequence data, including: The environmental DNA sequence data were padded with the nucleotide code N in the IUPAC nucleotide code to balance them to the same length; Each typical base of the environmental DNA sequence after equalization: A, C, T, G, and IUPAC ambiguous code is translated into four typical bases: A, T, C, G, and four-dimensional probability distribution, and the nucleotide sequence is represented as a two-dimensional array of sequence length multiplied by 4; S4, representing the sequence information of the sample; combining the two-dimensional array representing the nucleotide sequence information obtained in S3 with the presence and absence of each operational classification unit in each sample to represent a three-dimensional array; The presence of each operational taxonomic unit: nucleotide sequence information of the sequence translation of the detected OTU; The missing of each operational taxonomic unit: if no OTU is detected, it is a zero matrix; S5, inputting the three-dimensional array representing the sample sequence information into the autoencoder AE to compress the environmental DNA sequence data; S6, using a differential autoencoder VAE to reduce dimensionality, using the denoised environmental DNA abundance data obtained in S2 and the compressed and encoded environmental DNA sequence data obtained in S5 as inputs of the differential autoencoder VAE, further reducing the dimension of the sample to a two-dimensional latent space; S7. Draw a visualization result graph in two-dimensional space.
2. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 1 is characterized in that: The pre-processed environmental DNA abundance data is input into a deep counting autoencoder DCA for denoising to obtain denoised environmental DNA abundance data, including: The specific formula of the deep counting autoencoder DCA architecture is: E=ReLU(XW E ); B=ReLU(EW B ): D=ReLU(BW E ); M=diag(s j )exp r(DW μ ); Π=sigmoid(DW π ); θ=exp(DW θ ); where E, B, and D represent the encoder layer, latent layer, and decoder layer, respectively; M, θ, and Π represent the matrix forms of mean, dispersion, and loss probability estimation, respectively; s j is the size factor; the activation function chosen for the mean and dispersion is exponential; the activation function for the loss probability estimate is sigmoid, which represents the dropout probability; The zero-inflated negative binomial distribution ZINB is used to describe the distribution of environmental DNA data, where the zero-inflated negative binomial distribution ZINB is: ZINB(x;π,μ,θ)=πδ(x)+(1-π)NB(x;μ,θ); where π, μ, and θ are the parameters of the ZINB distribution, representing the probability, mean, and dispersion of loss events, respectively; x represents the raw read counts, and the autoencoder based on the ZINB model estimates the parameters π, μ, and θ.
3. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 2 is characterized in that: The loss function of the deep counting autoencoder DCA is the negative logarithm of the ZINB likelihood, and the specific formula is as follows: L ZINB =-log(ZINB(X|v, μ, θ)).
4. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 1 is characterized in that: The nucleotide code N padded to balance the sequence length is translated as [0.25, 0.25, 0.25, 0.25].
5. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 1, characterized in that: The three-dimensional array representing the sample sequence information is input into the autoencoder AE to compress the environmental DNA sequence data, specifically: The autoencoder AE includes a fully connected layer and a potential layer; Each fully connected layer uses ReLU activation and adopts a mirror architecture as a decoder; The loss function of the autoencoder AE is binary cross entropy: L AE (-ylog(p(y))-(1-y)log(1-p(y))); Here, y is a binary label 0 or 1, and p(y) is the probability that the output belongs to the y label.
6. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 1, characterized in that: The differential autoencoder VAE includes: Input layer, encoder network, latent sampling layer, decoder network, and output layer.
7. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 8, characterized in that: The potential sampling layer uses the reparameterization technique for sampling and the gradient descent method to optimize the likelihood function.
8. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 1, characterized in that: The loss function of the differential autoencoder VAE is composed of the VAE loss function: KL divergence, reconstruction error, and binary cross entropy of different combinations of potential encodings of nucleotide sequences: N(0,1)])+α(-ylog(p(y))-(1-y)log(1-p(y))); Where x is the abundance input matrix of the differential encoder, is the differential autoencoder abundance output matrix, N(u x , σ x ) represents the distribution of the potential layer z, y is the binary label 0 or 1, p(y) is the probability that the output belongs to the y label, and α is the weight of the loss function accounted for by the binary cross entropy.
9. The environmental DNA data dimensionality reduction method based on hierarchical autoencoder according to claim 1, characterized in that: The deep counting autoencoder DCA and the autoencoder AE are trained separately, and then the VAE is trained.
10. An environmental DNA data dimensionality reduction system based on hierarchical autoencoders, characterized in that: The system includes the following modules: data preprocessing module, denoising module, encoding module, dimensionality reduction module, and visualization module; The data preprocessing module is used to divide the environmental DNA data into environmental DNA abundance data and environmental DNA sequence data, filter out all zero rows of the abundance data, perform Min-Max normalization and log-2 logarithmic transformation on the abundance data, output the preprocessed environmental DNA abundance data and environmental DNA sequence data, and divide the environmental DNA data into environmental DNA abundance data and environmental DNA sequence data; The denoising module is used to denoise the pre-processed environmental DNA abundance data and output the denoised environmental DNA abundance data; The encoding module is used to equalize the length of the environmental DNA sequence data, and convert the sequence after equalization into the probability of four typical bases, and represent the nucleotide sequence as a two-dimensional array of the sequence length multiplied by 4; the sequence information is combined with the OTU presence / absence encoding into a three-dimensional array input to the autoencoder AE, and the environmental DNA sequence data is compressed and encoded, and the compressed sample sequence information data is output; The dimension reduction module combines the two inputs of the denoised environmental DNA abundance data and the compressed and encoded environmental DNA sequence data to perform dimensionality reduction processing on the environmental DNA data, reducing the sample dimension to a two-dimensional space; The visualization module visualizes the data reduced to two dimensions.
Citation Information
Patent Citations
ScRNA-seq data dimension reduction method based on deep adversarial variational auto-encoder
CN114067915A
Cited By
Deep sea mining area bacteria monitoring data analysis method
CN120510914A
EDNA data dimension reduction method
CN120510925A
