Multi-species epigenetic modification prediction method and device

Abstract

The invention provides a multi-species epigenetic modification prediction method and device. The method comprises the following steps: acquiring a genome sequence and a pre-established multi-species epigenetic modification prediction model; wherein the multi-species epigenetic modification prediction model comprises a feature extraction module and a multi-task learning model; performing feature extraction on the genome sequence based on a feature extraction module to obtain feature representation; and inputting the feature representation into a multi-task learning model, and predicting to obtain modification level data of each epigenetic modification type. By combining the feature extraction module and the multi-task learning model, the cross-species prediction capability is realized, and the modification level data of various epigenetic modification types can be predicted more accurately. The method can be widely applied to the fields of plant epigenetics research, agricultural production, biodiversity protection and the like.

Description

Technical Field

[0001] The present invention relates to the technical fields of bioinformatics and plant epigenetics, and in particular to a method and device for predicting epigenetic modifications of multiple species. Background Art

[0002] Epigenetic modification refers to gene expression regulation mechanisms that do not involve changes in DNA sequences, including but not limited to DNA methylation and histone modifications. These modifications play a vital role in plant growth and development and response to environmental stress. Although the development of high-throughput sequencing technology has greatly promoted the study of plant epigenetic genetics, the limitations of experimental technology have led to limited understanding of the epigenetic modification maps of most plant species. Traditional machine learning methods are usually limited to a single species and have difficulty capturing complex cross-species epigenetic regulatory networks. Therefore, it is crucial to develop an epigenetic modification prediction method that can integrate multi-species data and has cross-species prediction capabilities. Summary of the invention

[0003] The present invention provides a multi-species epigenetic modification prediction method and device to solve the defects of the prior art that it is limited to a single species and difficult to capture complex cross-species epigenetic regulatory networks, and realize cross-species epigenetic modification prediction. The technical solution proposed by the present invention is as follows: In a first aspect, the present invention provides a method for predicting epigenetic modifications in multiple species, comprising: Obtaining a genome sequence and a pre-established multi-species epigenetic modification prediction model; wherein the multi-species epigenetic modification prediction model includes a feature extraction module and a multi-task learning model; Extracting features from the genome sequence based on a feature extraction module to obtain feature representation; The feature representation is input into the multi-task learning model to predict the modification level data of each epigenetic modification type.

[0004] Optionally, extracting features from the genome sequence based on a feature extraction module to obtain feature representation includes: Performing labeling processing on the genome sequence to obtain a forward sequence and a reverse complementary sequence respectively; The forward sequence and the reverse complementary sequence are input into the feature extraction module to perform feature extraction to obtain the feature representation.

[0005] Optionally, the feature extraction module includes a first convolution module, a plurality of residual modules, a position embedding layer, and a plurality of Transformer encoding modules; the step of inputting the forward sequence and the reverse complementary sequence into the feature extraction module for feature extraction to obtain the feature representation includes: Inputting the forward sequence and the reverse complementary sequence into the first convolution module respectively to obtain a first feature map; Inputting the first feature map into the plurality of residual modules, each residual module capturing local and global feature information through a skip connection mechanism, to obtain a second feature map; Inputting the second feature map into the position embedding layer for position encoding to obtain a third feature map; Input the third feature map into the Transformer encoding module, extract global contextual relationship features through position encoding and multi-head attention mechanism, and obtain a fourth feature map; The fourth feature graphs corresponding to the forward sequence and the reverse complementary sequence are superimposed to obtain the feature representation.

[0006] Optionally, the multi-species epigenetic modification prediction model is trained in the following manner: Obtaining genome sequences and corresponding epigenetic modification data of multiple species to obtain a data set; wherein the epigenetic modification data is the actual modification level of the genome sequence; Divide the dataset into training set, validation set and test set Based on the training set, the pre-established deep neural network model is trained, and the model parameters are updated using a back propagation algorithm to minimize the loss function to obtain a trained model; wherein the loss function is used to measure the difference between the model prediction output and the actual modification level; Using the validation set to evaluate the trained model under different hyperparameter combinations, and optimizing the hyperparameters by a preset optimization method to obtain an optimized model; The model performance of the optimized model is evaluated using a test set until the model performance reaches the preset requirements, thereby obtaining the multi-species epigenetic modification prediction model.

[0007] Optionally, the method further comprises: Convert the genome sequences of all species into a unified probabilistic coding format.

[0008] Optionally, the multi-task learning model includes a second convolution module, a global pooling layer, and a plurality of species prediction heads, each species prediction head corresponding to a prediction task of an epigenetic modification type; Inputting the feature representation into the multi-task learning model to predict modification level data of each epigenetic modification type includes: Inputting the feature representation into the second convolution module for feature extraction to obtain a convolution module output; The global pooling layer performs global pooling on the output of the convolution module, and inputs the output of the global pooling layer into the prediction head of each species to obtain modification level data of the corresponding epigenetic modification type.

[0009] In a second aspect, the present invention also provides a multi-species epigenetic modification prediction device, comprising the following modules: An acquisition module, used to acquire genome sequences and a pre-established multi-species epigenetic modification prediction model; wherein the multi-species epigenetic modification prediction model includes a feature extraction module and a multi-task learning model; An extraction module, used for extracting features from the genome sequence based on the feature extraction module to obtain feature representation; A prediction module is used to input the feature representation into the multi-task learning model to predict the modification level data of each epigenetic modification type.

[0010] In a third aspect, the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for predicting epigenetic modifications of multiple species as described in the first aspect above is implemented.

[0011] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting epigenetic modifications in multiple species as described in the first aspect above.

[0012] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method for predicting epigenetic modifications in multiple species as described in the first aspect above.

[0013] Based on the above technical solution, the beneficial effects of the present invention compared with the prior art are as follows: The multi-species epigenetic modification prediction method and device provided by the present invention extracts features from the genome sequence using a feature extraction module, learns and extracts feature representations that can represent epigenetic modification information from the genome sequence. Since the prediction model is trained on multi-species data, the feature extraction module can learn common features across species. These common features exist in the genome sequences of different species and can therefore be used for cross-species predictions. The feature representation is input into a multi-task learning model, which can simultaneously handle prediction tasks for multiple types of epigenetic modifications. The multi-task learning model predicts multiple types of epigenetic modifications based on the feature representation of the input genome sequence, and gives modification level data for each modification type. The present invention achieves cross-species prediction capabilities by combining a feature extraction module with a multi-task learning model, and can more accurately predict modification level data for multiple types of epigenetic modifications. It can be widely used in the fields of plant epigenetic research, agricultural production, and biodiversity conservation.

[0014] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0015] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 work.

[0017] Figure 1 It is a schematic diagram of the process of predicting the multi-species epigenetic modification provided by the present invention.

[0018] Figure 2 It is a schematic diagram of the architecture of the multi-species epigenetic modification prediction model provided by the present invention.

[0019] Figure 3 It is a schematic diagram of the structure of the multi-species epigenetic modification prediction device provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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 are within the scope of protection of the present invention.

[0022] Combine the following Figure 1-Figure 3 The present invention describes the method and device for predicting epigenetic modifications in multiple species.

[0023] Reference Figure 1 As shown, the multi-species epigenetic modification prediction method includes the following: Step S101, obtaining genome sequences and pre-established multi-species epigenetic modification prediction models.

[0024] First, the genome sequence to be predicted is obtained from the database or experimental data. The genome sequence can be obtained through high-throughput sequencing technology. At the same time, a pre-established multi-species epigenetic modification prediction model is obtained. The prediction model can be trained on a large amount of multi-species epigenetic modification data through machine learning methods, especially deep learning technology. The model consists of a feature extraction module and a multi-task learning model to effectively capture the characteristic differences between different plant species. The feature extraction module is used to extract useful feature information from the genome sequence, while the multi-task learning model is used to simultaneously handle the prediction tasks of multiple types of epigenetic modifications. The feature extraction module learns common epigenetic features, while the multi-task learning model captures the unique features of each species, enabling the model to make detailed distinctions between multi-species data. Through this architecture, the model can comprehensively analyze the input data and obtain the predicted modification levels of various epigenetic modifications, including DNA methylation and histone modification. The prediction results can provide accurate candidate modification sites and target regions for further epigenetic experimental research.

[0025] When training the above-mentioned multi-species epigenetic modification prediction model, it is necessary to integrate epigenetic modification data from multiple species. By pre-training on large-scale and diverse datasets, the model can learn common features across species and thus have the ability to predict across species.

[0026] Step S102: extract features from the genome sequence based on a feature extraction module to obtain feature representation.

[0027] The feature extraction module is used to extract features from the genome sequence to obtain feature representations that can represent the epigenetic modification information in the genome sequence. The feature extraction module can include deep learning models such as convolutional neural networks (CNN), recurrent neural networks (RNN), or Transformer to learn and extract useful feature information from the genome sequence. Through feature extraction, the original genome sequence can be converted into higher-level feature representations that can more accurately reflect the information of epigenetic modifications. Since the model is trained on multi-species data, the feature extraction module can learn common features across species. These common features exist in the genome sequences of different species and can therefore be used for cross-species predictions.

[0028] Step S103: input the feature representation into the multi-task learning model to predict the modification level data of each epigenetic modification type.

[0029] The feature representation is input into a multi-task learning model, which can handle the prediction tasks of multiple epigenetic modification types at the same time. The multi-task learning model uses the feature representation output by the feature extraction module to train multiple species prediction heads at the same time, each species prediction head corresponds to the prediction task of one type of epigenetic modification. In this way, the model can share information between multiple tasks and improve the accuracy of prediction. Multi-task learning can take advantage of the correlation between different tasks to improve the generalization ability and prediction accuracy of the model. At the same time, because the model is pre-trained on multi-species data, it has the ability to predict across species.

[0030] The above-mentioned epigenetic modification refers to a gene expression regulation mechanism that does not involve changes in the DNA sequence, including but not limited to DNA methylation and histone modification. Accordingly, the above-mentioned epigenetic modification types include DNA methylation, DNA modification, DNA binding proteins, histone marks, transcription factors, chromatin accessibility and RNA modification, etc. The modification level data of each type of epigenetic modification specifically refers to the degree or state of this type of modification in a specific tissue, cell or organism, and the specific description can be referred to in the prior art. For example, the modification level data of DNA methylation generally refers to the degree of addition of methyl groups on DNA molecules, that is, the methylation level. The modification level data of histone modification refers to the degree of modification of specific amino acid residues on histones, such as acetylation, methylation, etc. The modification level data of RNA modification refers to the degree of addition of specific chemical groups on RNA molecules, such as N6-methyladenine (m6A), etc.

[0031] The multi-species epigenetic modification prediction method provided by the present invention uses a feature extraction module to extract features from the genome sequence, learns and extracts feature representations that can represent epigenetic modification information from the genome sequence. Since the prediction model is trained on multi-species data, the feature extraction module can learn common features across species. These common features exist in the genome sequences of different species and can therefore be used for cross-species predictions. The feature representation is input into a multi-task learning model, which can simultaneously handle prediction tasks for multiple types of epigenetic modifications. The multi-task learning model predicts multiple types of epigenetic modifications based on the feature representation of the input genome sequence, and gives modification level data for each modification type. The present invention achieves cross-species prediction capabilities by combining a feature extraction module and a multi-task learning model, and can more accurately predict modification level data for multiple types of epigenetic modifications. It can be widely used in the fields of plant epigenetic research, agricultural production, and biodiversity conservation.

[0032] In an optional embodiment, the forward and reverse information of the genome sequence can be used to capture more biological features. The feature extraction module described in step S102 extracts features from the genome sequence to obtain feature representation, including: S1021. Perform labeling processing on the genome sequence to obtain a forward sequence and a reverse complementary sequence respectively.

[0033] The genome sequence is marked by the genome tagging module a, specifically the annotation of specific nucleotides or base sequences, to facilitate subsequent feature extraction and model understanding. The encoder module generates its forward sequence (i.e. the original sequence itself) and reverse complementary sequence based on the original genome sequence. The reverse complementary sequence is obtained by replacing the bases in the original sequence according to the complementary rules of AT and CG and reversing the sequence order.

[0034] S1022: Input the forward sequence and the reverse complementary sequence into the feature extraction module to perform feature extraction to obtain the feature representation.

[0035] The forward sequence and reverse complementary sequence are input into the feature extraction module to extract useful feature information from the sequence. The feature extraction module captures the pattern, structure and variation information in the genome sequence and converts it into a high-dimensional feature representation. These feature representations can more accurately reflect the epigenetic modification information in the genome sequence.

[0036] The present invention can capture the forward and reverse information of the genome sequence by introducing the reverse complementary sequence, so as to more comprehensively understand the biological characteristics of the sequence. This helps to improve the prediction accuracy of the model for epigenetic modification. Moreover, the introduction of the reverse complementary sequence increases the diversity of the data set, so that the model can learn more features and patterns during the training process. This helps to enhance the generalization ability of the model, so that it can better adapt to different species and different types of genome sequences.

[0037] In an alternative embodiment, referring to Figure 2 As shown, the present invention designs a feature extraction module, including a convolution module (hereinafter referred to as the first convolution module), multiple residual blocks, a position embedding layer, and multiple Transformer encoder blocks. The convolution module captures key patterns from the genome sequence, and then applies batch normalization and activation functions to enhance the feature expression capability. The feature extraction module extracts features from the input sequence through the convolution module and the Transformer encoder block, thereby more effectively capturing the complex epigenetic regulation patterns in the plant genome and significantly improving the prediction accuracy.

[0038] The step S1022 of inputting the forward sequence and the reverse complementary sequence into the feature extraction module to extract features and obtain the feature representation includes: S201. Input the forward sequence and the reverse complementary sequence into the first convolution module respectively to obtain a first feature map.

[0039] Get the DNA sequence from the genome database and generate its forward sequence and reverse complementary sequence. These two sets of sequences represent the original direction and reverse complementary direction of the DNA sequence, respectively, which helps to capture more comprehensive biological information. The forward sequence and reverse complementary sequence are input into the first convolution module respectively.

[0040] The first convolution module performs a convolution operation on the sequence by sliding the convolution kernel on the sequence to capture key patterns from the genome sequence, and then applies batch normalization and activation functions to enhance the feature expression capability and generate the first feature map.

[0041] S202: Input the first feature map into the multiple residual modules, each residual module captures local and global feature information through a skip connection mechanism to obtain a second feature map.

[0042] Reference Figure 2 As shown in the figure, the structure of the residual module c consists of multiple convolutional layers and batch normalization layers, combined with residual connections, Swish activation functions and Dropout operations to achieve information transfer and feature retention in deep networks. The residual module alleviates the gradient vanishing problem by transferring information across layers and enhances the training effect of deep networks. The first feature map is input into multiple residual modules, which capture local and global feature information through a skip connection mechanism. Each residual module further extracts and refines features to generate a second feature map. The second feature map contains richer local and global feature information.

[0043] S203: Input the second feature map into the position embedding layer for position encoding to obtain a third feature map.

[0044] The position embedding layer is a mechanism for capturing position information in a sequence. The second feature map is input into the position embedding layer for position encoding. Through position encoding, each feature is given its position information in the sequence, which is crucial for understanding the contextual relationship in the sequence. The third feature map is generated, which contains the feature information after position encoding.

[0045] S204. Input the third feature map into the Transformer encoding module d, extract global contextual relationship features through position encoding and multi-head attention mechanism, and obtain a fourth feature map.

[0046] Reference Figure 2As shown in Figure 2, the Transformer encoding module d is a deep learning model based on the attention mechanism that can capture global contextual relationship features. The Transformer encoding module includes position encoding, multi-head attention mechanism, feedforward layer, layer normalization layer, and Dropout layer. These components work together to capture the dependencies and interactions of different positions in the genome sequence from a global perspective, which helps to extract more biologically meaningful features.

[0047] The third feature map is input into the Transformer encoding module. Through position encoding and multi-head attention mechanism, the Transformer encoding module can extract global features and information in the DNA sequence and generate the fourth feature map, which contains global contextual relationship features. The output of the Transformer encoding module corresponding to the above forward sequence is a forward tensor, and the output corresponding to the above reverse complementary sequence is a reverse tensor.

[0048] S205 , superimposing the fourth feature graphs corresponding to the forward sequence and the reverse complementary sequence to obtain the feature representation.

[0049] The forward tensor and reverse tensor corresponding to the forward sequence and the reverse complementary sequence are respectively superimposed with features. Through the feature superposition, a comprehensive feature representation containing the information of the forward sequence and the reverse complementary sequence can be obtained.

[0050] The present invention can extract local, global and contextual relationship features in DNA sequences through the combined use of convolution modules, residual modules, position embedding layers and Transformer encoding modules. These features help to improve the prediction accuracy of the model for epigenetic modification prediction tasks. Moreover, by simultaneously processing the forward sequence and the reverse complementary sequence, more comprehensive biological features and information in the DNA sequence can be captured. This helps to more accurately understand the biological functions and regulatory mechanisms of DNA sequences. The introduction of reverse complementary sequences increases the diversity of data sets, which helps the model learn more features and patterns. This can improve the generalization ability of the model under different data sets and scenarios.

[0051] In an optional embodiment, the multi-species epigenetic modification prediction model is trained in the following manner: S301. Obtain genome sequences and corresponding epigenetic modification data of multiple species to obtain a data set; wherein the epigenetic modification data is the actual modification level of the genome sequence.

[0052] Several plant species were selected as research objects, including model plants such as Arabidopsis thaliana, rice (Oryza sativa), corn (Zea mays), and other crops that play an important role in agriculture.

[0053] Get the genome sequence from a public database. Genome sequence data is usually stored in FASTA format, containing the chromosome or genome region sequence of the species. Epigenetic modification data is stored in BED, WIG or custom format, containing information such as modification location, type and modification level. Use bioinformatics tools or custom scripts to remove low-complexity regions (such as repetitive sequences, simple sequences, etc.) in the genome sequence. These regions may interfere with the prediction of epigenetic modifications.

[0054] Obtain epigenetic modification data from specific research literature or laboratory databases, which may include modification levels such as DNA methylation and histone modification. For epigenetic modification data, it is also necessary to remove noise data or unreliable modification sites to ensure the accuracy of the data.

[0055] Convert the genome sequences of all species into a unified probabilistic coding format. This coding method usually involves converting DNA sequences (A, T, C, G) into numerical vectors, or using more complex representation methods such as one-hot coding, k-mer coding, etc., to meet the input requirements of deep neural network models. For example, numpy can be used to convert the processed genome sequence into a probabilistic coding format for subsequent deep neural network model input. The specific process of probabilistic coding is: Create a numpy array with shape (sequence length, 4) to store the probability vector for each base. Iterate over each base in the genome sequence and set the corresponding probability value in the probability vector according to the base type (A, T, C, G) (for example, for base A, you can set the vector to [1, 0, 0, 0] or a more complex probability distribution). For example, it can be encoded as: 'A':[1, 0, 0, 0], 'C': [0, 1, 0, 0], 'G': [0, 0, 1, 0], 'T': [0, 0, 0, 1], 'M':[0.5, 0.5, 0, 0], 'R': [0.5, 0, 0.5, 0], 'W': [0.5, 0, 0, 0.5], 'S': [0, 0.5,0.5, 0], 'Y': [0, 0.5, 0, 0.5], 'K': [0, 0, 0.5, 0.5], 'V': [1 / 3, 1 / 3, 1 / 3,0], 'H': [1 / 3, 1 / 3, 0, 1 / 3], 'D': [1 / 3, 0, 1 / 3, 1 / 3], 'B': [0, 1 / 3, 1 / 3, 1 / 3], 'N': [0.25, 0.25, 0.25, 0.25].

[0056] Integrate the processed genome sequence and epigenetic modification data to form a comprehensive dataset containing multiple species and multiple epigenetic modification types. The dataset should contain the genome sequence of each species, the corresponding epigenetic modification data (including modification location, type and modification level), and any relevant metadata (such as species name, tissue type, treatment conditions, etc.).

[0057] S302, dividing the data set into a training set, a validation set and a test set, so as to train, validate and test the deep learning model in subsequent steps. When dividing the data, the independence between the training set, the validation set and the test set should be ensured to avoid the problem of overfitting of the model.

[0058] S303, training a pre-established deep neural network model based on the training set, and using a back propagation algorithm to update the model parameters to minimize the loss function to obtain a trained model; wherein the loss function is used to measure the difference between the model prediction output and the actual modification level.

[0059] A deep neural network model is designed in advance, which should be able to process genome sequence input and output corresponding epigenetic modification prediction results. This model is the initial form of the multi-species epigenetic modification prediction model.

[0060] The model is trained using a training set that covers the genomic information of multiple model plants and important crops to ensure that the model can learn features that are generally adaptable. The model parameters are continuously adjusted through the back-propagation algorithm to minimize the loss function. The loss function is used to measure the difference between the model's predicted output and the actual modification level, and can use mean square error (MSE), cross entropy loss, etc.

[0061] The present invention uses mean square error (MSE) as a loss function to measure the difference between the model prediction output and the actual modification level. The MSE loss function can intuitively reflect the deviation between the predicted value and the true value, which helps to guide the optimization of model parameters. The AdamW optimizer is used for parameter update to effectively adjust the model weights during the training process and promote rapid convergence of the model.

[0062] The training process is as follows: input the training data into the model, calculate the predicted output through forward propagation, and calculate the loss value according to the MSE loss function. Use the AdamW optimizer to backpropagate and update the model parameters according to the loss value. Repeat the above steps until the preset number of training rounds is reached or the validation set loss no longer improves.

[0063] S304: Use the validation set to evaluate the trained model under different hyperparameter combinations, and optimize the hyperparameters using a preset optimization method to obtain an optimized model.

[0064] Use the validation set to evaluate the performance of the trained model under different hyperparameter combinations. Hyperparameters include learning rate, regularization coefficient, convolution kernel size, batch size, etc. Set a reasonable range of hyperparameters, traverse all possible combinations, and evaluate the performance of each combination through cross-validation. Find the optimal hyperparameter combination through preset optimization methods (such as grid search, random search, Bayesian optimization, etc.). Select the hyperparameter combination with the best performance as the configuration of the final model.

[0065] Monitor the changes in validation set loss during training. When validation set loss stops improving over time, stop the training process to prevent the model from overfitting. In addition to the weight decay provided by the AdamW optimizer, you can also consider introducing strategies such as dropout layers or L1 regularization in the model architecture to further reduce the risk of overfitting.

[0066] S305. Use a test set to evaluate the model performance of the optimized model until the model performance reaches a preset requirement, thereby obtaining the multi-species epigenetic modification prediction model.

[0067] Use a reserved independent test set to verify the performance of the model. The test set should contain plant species and epigenetic modification types that are different from the training set and validation set to evaluate the predictive ability of the model on unknown data. Evaluation indicators include accuracy, recall, F1 score, AUC-ROC curve, etc. If the model performance meets the preset requirements (such as accuracy, AUC value, etc. reach a certain threshold), the above-mentioned multi-species epigenetic modification prediction model is obtained. Evaluate the generalization ability of the model by testing on different plant species and epigenetic modification types. Ensure that the model can show good predictive performance in different application scenarios.

[0068] The present invention can accurately capture the complex relationship between genome sequence and epigenetic modification through the powerful feature extraction and learning ability of the deep neural network model, thereby improving the prediction accuracy. By training data from multiple species, the model can learn the common characteristics between different species, thereby enhancing the generalization ability of the model and enabling it to be applied to unknown species or newly discovered species. The model provides a powerful tool for epigenetic research. By predicting the epigenetic modification levels of different species, it can provide accurate candidate modification sites and target areas for further epigenetic experimental research.

[0069] In an alternative embodiment, referring to Figure 2 As shown, the multi-task learning model includes a convolution module (hereinafter referred to as the second convolution module), a global pooling layer, and multiple species prediction heads, each of which corresponds to a prediction task of an epigenetic modification type; The step S103 described above of inputting the feature representation into the multi-task learning model to predict the modification level data of each epigenetic modification type includes: S1031. Input the feature representation into the second convolution module for feature extraction to obtain a convolution module output.

[0070] The feature representation of the genomic sequence is input into the second convolutional module of the multi-task learning model. Figure 2 As shown, the convolution modules b in the present invention include a convolution layer, a batch normalization layer, a Swish activation function and a Dropout layer. The convolution layer is the core component of the second convolution module, which uses multiple convolution kernels to extract local features of the input data (i.e., the above-mentioned feature representation). Each convolution kernel slides on the input data, calculates the convolution result, and generates a feature map. These feature maps capture the local features of the input data at different scales and positions. Then batch normalization and Swish activation function are applied to enhance the feature expression capability.

[0071] After the above hierarchical processing, the second convolution module is able to extract complex feature representations in the input data and pass these features to the subsequent global pooling layer for further processing.

[0072] The above Swish activation function is applied to the output of the batch normalization layer. The Swish function has the characteristic of linear growth when the input value is large, and tends to zero when the input value is small. This characteristic helps the model learn more complex feature representations. Dropout is a regularization technique used to prevent the model from overfitting during training. In the Dropout layer, each neuron has a certain probability (such as 0.5) of being randomly discarded during each forward propagation, that is, its output is set to zero.

[0073] S1032. The global pooling layer globally pools the output of the convolution module, and inputs the output of the global pooling layer into the prediction head of each species to obtain modification level data of the corresponding epigenetic modification type.

[0074] After the second convolution module, the global pooling layer performs a global pooling operation on the output of the convolution module. The output of the global pooling layer is input into multiple species prediction heads respectively. Each species prediction head corresponds to a prediction task for a specific species and epigenetic modification type. These prediction heads are fully connected layers, which learn how to predict the modification levels of the corresponding species and modification types based on the output of the global pooling layer. Each species prediction head outputs a prediction value, which represents the modification level data of the corresponding epigenetic modification type. These data can be continuous values ​​(such as methylation level, histone modification intensity, etc.) or classification labels (such as whether a certain modification exists).

[0075] The present invention performs feature extraction through the second convolution module, and the model can learn complex feature representations in the input data. These feature representations are shared in the prediction tasks of multiple species and epigenetic modification types, which helps the model learn more general and robust features. The global pooling layer retains the most important global feature information by performing a pooling operation on the entire feature map. This helps the model to better utilize these feature information in subsequent prediction tasks and improve the accuracy and stability of the prediction.

[0076] The multi-task learning model achieves information sharing and collaborative optimization between different tasks by processing multiple prediction tasks at the same time. This collaborative optimization helps the model better learn the relationship between different species and epigenetic modification types, further improving the accuracy and generalization of predictions. Since the model can handle multiple prediction tasks at the same time, it can significantly improve the prediction efficiency in practical applications. In addition, the model is also very scalable, and new species or epigenetic modification types can be easily added without retraining the entire model.

[0077] After obtaining the modification level data for each epigenetic modification type, the predicted modification level data can be compared with a preset threshold to determine the modification site.

[0078] The above thresholds are important criteria for determining whether a modification site exists. The setting of the threshold can be based on experimental validation data, literature reports, or the results of model performance evaluation. The selection of the threshold should ensure that the sites where the modification actually exists can be accurately identified while avoiding too many false positive results.

[0079] The multi-species epigenetic modification prediction model is used to predict the input genome sequence to obtain the epigenetic modification level data for each site or region. These modification levels can be continuous values, indicating the relative magnitude of the modification intensity.

[0080] The predicted modification level data of each site is compared with the preset threshold. If the predicted value is higher than the threshold, it is considered that the site has the corresponding epigenetic modification; if the predicted value is lower than the threshold, it is considered that there is no significant modification or the degree of modification is low. According to the comparison results, the sites with modifications are annotated. The annotation information can include the modification type (such as DNA methylation, histone modification, etc.), modification intensity, and site location (specific location on the genome). These annotation information provides important basis for subsequent bioinformatics analysis, experimental verification and biological discovery.

[0081] The multi-species plant epigenetic modification prediction method proposed in this invention has good cross-species applicability and high-precision prediction capabilities, and can be widely used in plant epigenetic research, agricultural production, and biodiversity conservation. Through the multi-species learning mechanism of shared weights, this method also shows excellent prediction results on species with scarce data.

[0082] The multi-species epigenetic modification prediction device provided by the present invention is described below. The multi-species epigenetic modification prediction device described below and the multi-species epigenetic modification prediction method described above can be referenced to each other.

[0083] The multi-species epigenetic modification prediction device provided by the present invention refers to Figure 3 As shown, including: An acquisition module 401 is used to acquire a genome sequence and a pre-established multi-species epigenetic modification prediction model; wherein the multi-species epigenetic modification prediction model includes a feature extraction module and a multi-task learning model; An extraction module 402, used for extracting features from the genome sequence based on a feature extraction module to obtain a feature representation; The prediction module 403 is used to input the feature representation into the multi-task learning model to predict the modification level data of each epigenetic modification type.

[0084] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the multi-species epigenetic modification prediction method.

[0085] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0086] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the multi-species epigenetic modification prediction method provided by the above-mentioned methods.

[0087] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the multi-species epigenetic modification prediction method provided by the above-mentioned methods.

[0088] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0089] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0090] 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 make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting epigenetic modifications in multiple species, characterized in that: include: Obtaining a genome sequence and a pre-established multi-species epigenetic modification prediction model; wherein the multi-species epigenetic modification prediction model includes a feature extraction module and a multi-task learning model; Extracting features from the genome sequence based on a feature extraction module to obtain feature representation; The feature representation is input into the multi-task learning model to predict the modification level data of each epigenetic modification type.

2. The multi-species epigenetic modification prediction method according to claim 1, characterized in that: The step of extracting features from the genome sequence based on a feature extraction module to obtain a feature representation includes: Performing labeling processing on the genome sequence to obtain a forward sequence and a reverse complementary sequence respectively; The forward sequence and the reverse complementary sequence are input into the feature extraction module to perform feature extraction to obtain the feature representation.

3. The multi-species epigenetic modification prediction method according to claim 2, characterized in that: The feature extraction module includes a first convolution module, multiple residual modules, a position embedding layer, and multiple Transformer encoding modules; the forward sequence and the reverse complementary sequence are input into the feature extraction module for feature extraction to obtain the feature representation, including: Inputting the forward sequence and the reverse complementary sequence into the first convolution module respectively to obtain a first feature map; Inputting the first feature map into the plurality of residual modules, each residual module capturing local and global feature information through a skip connection mechanism, to obtain a second feature map; Inputting the second feature map into the position embedding layer for position encoding to obtain a third feature map; Input the third feature map into the Transformer encoding module, extract global contextual relationship features through position encoding and multi-head attention mechanism, and obtain a fourth feature map; The fourth feature graphs corresponding to the forward sequence and the reverse complementary sequence are superimposed to obtain the feature representation.

4. The multi-species epigenetic modification prediction method according to claim 1, characterized in that: The multi-species epigenetic modification prediction model is trained in the following manner: Obtaining genome sequences and corresponding epigenetic modification data of multiple species to obtain a data set; wherein the epigenetic modification data is the actual modification level of the genome sequence; Dividing the data set into a training set, a validation set, and a test set; Based on the training set, the pre-established deep neural network model is trained, and the model parameters are updated using a back propagation algorithm to minimize the loss function to obtain a trained model; wherein the loss function is used to measure the difference between the model prediction output and the actual modification level; Using the validation set to evaluate the trained model under different hyperparameter combinations, and optimizing the hyperparameters by a preset optimization method to obtain an optimized model; The model performance of the optimized model is evaluated using a test set until the model performance reaches the preset requirements, thereby obtaining the multi-species epigenetic modification prediction model.

5. The multi-species epigenetic modification prediction method according to claim 4, characterized in that: The method further comprises: Convert the genome sequences of all species into a unified probabilistic coding format.

6. The multi-species epigenetic modification prediction method according to claim 1, characterized in that: The multi-task learning model includes a second convolution module, a global pooling layer, and multiple species prediction heads, each species prediction head corresponding to a prediction task of an epigenetic modification type; Inputting the feature representation into the multi-task learning model to predict modification level data of each epigenetic modification type includes: Inputting the feature representation into the second convolution module for feature extraction to obtain a convolution module output; The global pooling layer performs global pooling on the output of the convolution module, and inputs the output of the global pooling layer into the prediction head of each species to obtain modification level data of the corresponding epigenetic modification type.

7. A method for predicting epigenetic modifications in multiple species, characterized in that: include: An acquisition module, used to acquire genome sequences and a pre-established multi-species epigenetic modification prediction model; wherein the multi-species epigenetic modification prediction model includes a feature extraction module and a multi-task learning model; An extraction module, used for extracting features from the genome sequence based on the feature extraction module to obtain feature representation; A prediction module is used to input the feature representation into the multi-task learning model to predict the modification level data of each epigenetic modification type.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the multi-species epigenetic modification prediction method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting multi-species epigenetic modifications as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for predicting multi-species epigenetic modifications as claimed in any one of claims 1 to 6 is implemented.

Images