Enzyme substrate interaction prediction method, device, equipment, medium and product

Abstract

The invention discloses an enzyme-substrate interaction prediction method, device, equipment, medium and product, and relates to the technical field of protease catalysis, the enzyme-substrate interaction prediction method comprises the following steps: respectively extracting enzyme characteristics of a target enzyme and substrate characteristics of a target substrate through a pre-trained encoder; adjusting weight parameters of a preset protein prediction model according to enzyme characteristics and the substrate characteristics to obtain a target protein prediction model; inputting the enzyme characteristics and the substrate characteristics into a target protein prediction model, and outputting interaction characteristics of the target enzyme and the target substrate by the target protein prediction model; and inputting the interaction characteristics into a preset decoder, decoding the interaction characteristics by the decoder, and outputting an interaction prediction result of the target enzyme and the target substrate. The technical problem that a traditional enzyme substrate pair interaction prediction scheme is poor in interaction prediction effect on rare enzyme substrate pairs is solved.

Description

Technical Field

[0001] The present application relates to the field of protein enzyme catalysis technology, and in particular to an enzyme-substrate interaction prediction method, apparatus, device, computer-readable storage medium, and computer program product. Background Art

[0002] Predicting enzyme-substrate interactions is crucial for understanding biological processes, accelerating drug development, and driving innovation in biomedical research. However, traditional wet lab methods are limited in both time and cost. With the rapid development of deep learning technology, a growing number of studies are attempting to leverage deep learning computational methods to predict enzyme-substrate interactions.

[0003] However, current deep learning computational methods ignore the severe imbalance at the data level and directly fit neural networks to enzyme-substrate interaction data. These methods range from training neural network models from scratch to using features extracted from pre-trained models. While these methods demonstrate good predictive performance on test sets, they often rely on fitting in-distribution data and have poor predictive ability for out-of-distribution data. This limitation is particularly severe in long-tail data distributions, as the model tends to overfit the head data during training and ignore the tail data, resulting in poor prediction results for rare or small numbers of enzyme-substrate pairs.

[0004] The above information disclosed in this Background section is only for understanding the background of the present invention and therefore it may contain information that does not constitute prior art. Summary of the Invention

[0005] The main purpose of this application is to provide an enzyme-substrate interaction prediction method, device, equipment, computer-readable storage medium and computer program product, aiming to solve the technical problem that traditional enzyme-substrate pair interaction prediction schemes have poor interaction prediction effect on rare enzyme-substrate pairs.

[0006] To achieve the above objectives, the present application provides an enzyme-substrate interaction prediction method, which comprises:

[0007] The enzyme features of the target enzyme and the substrate features of the target substrate are extracted respectively through the pre-trained encoder;

[0008] Adjusting the weight parameters of a preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model;

[0009] Inputting the enzyme characteristics and the substrate characteristics into the target protein prediction model, and outputting the interaction characteristics of the target enzyme and the target substrate from the target protein prediction model;

[0010] The interaction characteristics are input into a preset decoder, which decodes the interaction characteristics and outputs a predicted result of the interaction between the target enzyme and the target substrate.

[0011] In one embodiment, the encoder includes at least a protein language model and a graph neural network;

[0012] The steps of respectively extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate by the pre-trained encoder include:

[0013] Obtaining enzyme sequence information of the target enzyme and substrate molecule information of the target substrate;

[0014] Inputting the enzyme sequence information into a preset protein language model to obtain enzyme features of the target enzyme, wherein the enzyme features include hierarchical features of the enzyme;

[0015] The substrate molecule information is input into a preset graph neural network to obtain the substrate characteristics of the target substrate, wherein the substrate characteristics include topological structure and chemical properties.

[0016] In one embodiment, after the step of inputting the substrate molecule information into a preset graph neural network to obtain substrate characteristics of the target substrate, the method further includes:

[0017] The enzyme features and the substrate features are mapped to a unified latent space, wherein the distance between the enzyme features and the substrate features in the latent space is used to characterize the interaction strength between the target enzyme and the target substrate.

[0018] In one embodiment, before the step of extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate respectively by using a pre-trained encoder, the method further comprises:

[0019] Inputting the enzyme sequence information and the substrate molecule information into an initial protein language model and an initial graph neural network, respectively, and then the initial protein language model and the initial graph neural network respectively output prediction results corresponding to the enzyme sequence information and the substrate molecule information;

[0020] Based on a preset contrast loss function, the true labels corresponding to the enzyme sequence information and the substrate molecule information, and the prediction results, the initial protein language model and the initial graph neural network are pre-trained in a preset shared embedding space to obtain a pre-trained protein language model and graph neural network.

[0021] In one embodiment, the preset protein prediction model includes at least a weight adjustment layer;

[0022] The step of adjusting the weight parameters of the preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain the target protein prediction model comprises:

[0023] Inputting the enzyme characteristics and the substrate characteristics into a preset weighted supernetwork, and outputting target weight information from the weighted supernetwork;

[0024] The weight information of the weight adjustment layer in the preset protein prediction model is adjusted according to the target weight information to obtain a target protein prediction model.

[0025] In one embodiment, the preset protein prediction model further includes a projection layer;

[0026] Before the step of adjusting the weight parameters of the preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain the target protein prediction model, the method further includes:

[0027] Obtaining enzyme feature training data and substrate feature training data;

[0028] Inputting the enzyme feature training data and the substrate feature training data into an initial weighted supernetwork and a protein pre-training model, respectively, training the projection layer in the protein pre-training model and the initial weighted supernetwork to obtain a weighted supernetwork and a preset protein prediction model;

[0029] The projection layer is used to process the enzyme features and substrate features received by the preset protein prediction model.

[0030] In addition, to achieve the above-mentioned purpose, the present application also provides an enzyme-substrate interaction prediction device, which comprises:

[0031] A feature extraction module is used to extract the enzyme features of the target enzyme and the substrate features of the target substrate respectively through the pre-trained encoder;

[0032] A weight adjustment module, configured to adjust the weight parameters of a preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model;

[0033] A feature prediction module, configured to input the enzyme feature and the substrate feature into the target protein prediction model, and output the interaction feature of the target enzyme and the target substrate from the target protein prediction model;

[0034] The feature decoding module is used to input the interaction feature into a preset decoder, and the decoder decodes the interaction feature and outputs the interaction prediction result between the target enzyme and the target substrate.

[0035] In addition, to achieve the above-mentioned purpose, the present application also provides an enzyme-substrate interaction prediction device, which comprises at least: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the enzyme-substrate interaction prediction method as described above.

[0036] In addition, to achieve the above objectives, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the enzyme-substrate interaction prediction method described above are implemented.

[0037] In addition, to achieve the above objectives, the present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned enzyme-substrate interaction prediction method when executed by a processor.

[0038] The present application provides an enzyme-substrate interaction prediction method, which comprises the following steps: first, extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate respectively through a pre-trained encoder; then, adjusting the weight parameters of a preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model, so that the target protein prediction model can quickly and specifically adapt to the interaction prediction of the target enzyme and the target substrate; inputting the enzyme characteristics and the substrate characteristics into the target protein prediction model; and outputting the interaction characteristics of the target enzyme and the target substrate from the target protein prediction model; finally, inputting the interaction characteristics into a preset decoder, and decoding the interaction characteristics by the decoder to output the interaction prediction result of the target enzyme and the target substrate. In the technical solution of the present application, the target protein prediction model performs adaptive weight adjustment according to the characteristics of the enzyme and the substrate, so that the model can quickly learn the characteristics of the target enzyme and the target substrate. Compared with the traditional enzyme-substrate pair interaction prediction scheme, it maintains a higher prediction accuracy for out-of-distribution data, and enables the model to achieve better out-of-distribution generalization capabilities on rare data with unbalanced data volume, overcomes the shortcomings of existing schemes for long-tail distribution data and rare samples and small sample scenarios, and improves the generalization ability of the protein prediction model and the interaction prediction effect for rare enzyme-substrate pairs. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0041] Figure 1 Schematic diagram of the principle of traditional enzyme-substrate interaction prediction scheme for out-of-distribution data cases;

[0042] Figure 2 This is a schematic diagram of a process for predicting enzyme-substrate interactions in an embodiment of the present application;

[0043] Figure 3 Schematic diagram of the principle of the enzyme-substrate interaction prediction method in the embodiments of the present application for predicting out-of-distribution data examples;

[0044] Figure 4 Schematic diagram of data flow for a feasible enzyme-substrate interaction prediction scheme in the examples of this application;

[0045] Figure 5 Schematic diagram of the structure of the enzyme-substrate interaction prediction device in the embodiment of this application;

[0046] Figure 6 Schematic diagram of the device structure of the hardware operating environment involved in the enzyme-substrate interaction prediction method in the embodiment of the present application.

[0047] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0048] To make the above-mentioned purposes, features, and advantages of the present application more clearly understood, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0049] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0050] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0051] The traditional schemes used for predicting enzyme-substrate interactions have many limitations. Wet experimental analysis is not only time-consuming and costly, but also limited by experimental conditions and techniques, making it difficult to establish a complete enzyme-substrate mapping relationship. Although high-throughput technology has accelerated, the diversity of enzymes and substrates and the complexity of interactions still make data acquisition challenging. The existing enzyme-substrate related data have serious long-tail distribution problems. The imbalance of data limits researchers' comprehensive understanding of enzyme-substrate interactions and also poses a huge challenge to the development of deep learning computational models. Although deep learning technology has been applied to this field to a certain extent, its performance and generalization capabilities are significantly insufficient when dealing with imbalanced data distribution and few-sample and zero-sample learning scenarios. For example Figure 1 As shown in the figure, when faced with out-of-distribution data (hidden data), the representation distance between substrate and enzyme is large, and the model cannot accurately identify these rare enzyme-substrate pairs, resulting in poor prediction results. Because these traditional deep learning models often rely on fitting in-distribution data, their prediction ability for out-of-distribution data is poor. Especially in long-tail data distributions, the model tends to overfit the head data and ignore the tail data, resulting in poor prediction results for rare enzyme-substrate pairs.

[0052] The present invention provides a method for predicting enzyme-substrate interactions. Figure 2 , Figure 2 This is a flow chart of an embodiment of the enzyme-substrate interaction prediction method of the present application, which includes:

[0053] Step S10, extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate respectively through the pre-trained encoder;

[0054] Among them, the target enzyme and the target substrate refer to the enzyme and substrate that need to be predicted for interaction. Specifically, when extracting the corresponding enzyme features and substrate features, the enzyme sequence of the target enzyme and the molecular graph of the substrate can be used as the data basis. In addition, the pre-trained encoder refers to a pre-trained encoder that can accurately extract the corresponding features from the enzyme sequence of the target enzyme and the molecular graph of the substrate. It can be a deep learning model or a neural network. The goal of pre-training is to ensure that the extracted enzyme features of the target enzyme and the substrate features of the target substrate can accurately reflect the interaction strength between the target enzyme and the target substrate. For example, the higher the interaction strength between the two, the closer the distance between the enzyme features and the substrate features, thereby achieving the quantification of the interaction strength of the enzyme-substrate pair through a simple distance metric. Compared with the traditional manual feature extraction and direct splicing fusion strategy, it can significantly enhance its interpretability and final prediction accuracy, and lay the foundation for few-sample learning in data-scarce scenarios.

[0055] Step S20, adjusting the weight parameters of the preset protein prediction model according to the enzyme characteristics and substrate characteristics to obtain a target protein prediction model;

[0056] Emphasis in the present application's embodiment is that the weight parameters of the preset protein prediction model are carried out adaptive adjustment based on the enzyme characteristic and substrate characteristics extracted.It is understandable that, due to the long-tail characteristic of the enzyme-substrate interaction data set, its data distribution presents significant category imbalance phenomenon.So, in order to avoid the preset protein prediction model when the enzyme characteristic and substrate characteristics of input are predicted because the model overfitting causes the prediction result of this target enzyme and target substrate to be inaccurate, so in the present application's embodiment, a dynamic weight generation mechanism is introduced.

[0057] Among them, the preset protein prediction model is a pre-trained model for predicting the interaction between enzymes and substrates. It has a certain predictive ability, but it does not necessarily have a strong predictive ability for every enzyme-substrate pair. The preset protein prediction model is based on the input enzyme characteristics and substrate characteristics, and dynamically adaptively adjusts the parameters to ultimately obtain a target protein prediction model with a stronger predictive ability for the currently input enzyme characteristics and substrate characteristics. It should be noted that each time a new enzyme characteristic and substrate characteristic are input, the weight parameters are adaptively generated and updated based on the preset protein prediction model, thereby obtaining a target protein prediction model that is exclusive to the enzyme-substrate pair, so as to obtain a more accurate enzyme-substrate interaction prediction result.

[0058] Step S30, inputting the enzyme characteristics and substrate characteristics into a target protein prediction model, and the target protein prediction model outputs the interaction characteristics of the target enzyme and the target substrate;

[0059] After adjusting the weights of the preset protein prediction model and obtaining the target protein prediction model, the target protein prediction model can process and analyze the enzyme characteristics and substrate characteristics more accurately and deeply, thereby obtaining interaction characteristics that reflect the target enzyme and the target substrate. The interaction characteristics can directly reflect whether the target enzyme and the target substrate can interact and react with each other. If so, the interaction characteristics can also reflect the reaction rate between the two.

[0060] Step S40: Input the interaction characteristics into a preset decoder, which decodes the interaction characteristics and outputs a prediction result of the interaction between the target enzyme and the target substrate.

[0061] The preset decoder is a pre-trained model or network that can decode the interaction features output by the target protein prediction model. The decoder can decode and output a more intuitive reflection of the predicted interaction results between the target enzyme and the target substrate. The interaction prediction results may include whether the interaction can occur or not. If the interaction prediction results include whether the interaction can occur, it can also include the interaction reaction rate between the two.

[0062] The interaction prediction results can be applied to downstream tasks (such as enzyme function annotation and substrate specificity prediction), providing an interpretable feature space for downstream tasks.

[0063] An embodiment of the present application provides an enzyme-substrate interaction prediction method, which includes first extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate through a pre-trained encoder, then adjusting the weight parameters of a preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model, so that the target protein prediction model can quickly adapt to the interaction prediction of the target enzyme and the target substrate in a targeted manner, inputting the enzyme characteristics and the substrate characteristics into the target protein prediction model, and outputting the interaction characteristics of the target enzyme and the target substrate from the target protein prediction model, and finally inputting the interaction characteristics into a preset decoder, which decodes the interaction characteristics and outputs the interaction prediction result of the target enzyme and the target substrate. In the technical solution of the embodiment of the present application, the target protein prediction model performs adaptive weight adjustment according to the characteristics of the enzyme and the substrate, so that the model can quickly learn the characteristics of the target enzyme and the target substrate. Compared with the traditional enzyme-substrate pair interaction prediction scheme, it maintains a higher prediction accuracy for out-of-distribution data, and enables the model to achieve better out-of-distribution generalization capabilities on rare data with unbalanced data volume, overcomes the shortcomings of existing schemes for long-tail distribution data and rare samples and small sample scenarios, and improves the generalization ability of the protein prediction model and the interaction prediction effect for rare enzyme-substrate pairs.

[0064] Furthermore, in a feasible embodiment, the encoder includes at least a protein language model and a graph neural network;

[0065] The steps of respectively extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate by the pre-trained encoder include:

[0066] Step S11, obtaining enzyme sequence information of the target enzyme and substrate molecule information of the target substrate;

[0067] Step S12, inputting the enzyme sequence information into a preset protein language model to obtain enzyme features of the target enzyme, wherein the enzyme features include hierarchical features of the enzyme;

[0068] In step S13, the substrate molecular information is input into a preset graph neural network to obtain substrate characteristics of the target substrate, wherein the substrate characteristics include topological structure and chemical properties.

[0069] The embodiments of the present application disclose extracting hierarchical features from enzyme sequence information through a protein language model to obtain enzyme features; and extracting topological structures and chemical properties from substrate molecule information through a graph neural network to obtain substrate features.

[0070] The pre-set protein language model and graph neural network are both pre-trained. The goal of pre-training is to ensure that enzyme-substrate pairs participating in the same biochemical reaction are consistently represented in the latent space, while unrelated combinations produce significant feature differences. Enzyme proteins and substrate molecules are considered as two complementary modalities of biochemical reactions, and cross-modal semantic associations are constructed through comparative learning.

[0071] Furthermore, after the step of inputting the substrate molecule information into a preset graph neural network to obtain substrate characteristics of the target substrate, the method further includes:

[0072] Step S13 , mapping the enzyme features and substrate features to a unified latent space, wherein the distance between the enzyme features and the substrate features in the latent space is used to characterize the interaction strength between the target enzyme and the target substrate.

[0073] After extracting the enzyme features of the target enzyme and the substrate features of the target substrate, the enzyme features and substrate features are mapped to a unified latent space. The purpose of the above steps is to reflect the specific rules of the reaction between different enzymes and different substrates through the distance between the enzyme features and substrate features in the latent space. That is, enzymes with similar catalytic functions will be closer to their substrate molecules in the latent space, while enzyme-substrate pairs of different functional categories maintain a significant distance. And through the geometric characteristics of this spatial distance, an interpretable feature space is provided for downstream tasks (such as enzyme function annotation and substrate specificity prediction).

[0074] It can be understood that when the enzyme-substrate interaction prediction method of the embodiment of the present application is applied to the out-of-distribution data (hidden data), such as Figure 3 As shown, the representation distance between the substrate and the enzyme is closer. Before inputting the enzyme features and substrate features into the model, the enzyme features and substrate features are combined with the super model to adjust the model weights, so that the model can eventually adapt to the currently input enzyme features and substrate features and output more accurate prediction results.

[0075] Furthermore, in a feasible embodiment, before the step of respectively extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate by the pre-trained encoder, the method further includes:

[0076] Step A10: Input the enzyme sequence information and substrate molecule information into the initial protein language model and the initial graph neural network, respectively. The initial protein language model and the initial graph neural network then output prediction results corresponding to the enzyme sequence information and substrate molecule information, respectively.

[0077] In step A20, based on the preset contrast loss function, the true labels corresponding to the enzyme sequence information and the substrate molecule information, and the prediction results, the initial protein language model and the initial graph neural network are pre-trained in the preset shared embedding space to obtain the pre-trained protein language model and graph neural network.

[0078] The present application embodiment provides a method for pre-training a protein speech model and a graph neural network. In the pre-training stage, the present application embodiment constructs a comparative learning framework of enzyme protein sequence-substrate molecule graph, such as Figure 4 As shown in the figure, a GNN (Graph Neural Network) and ESM-2 (Evolution Scale Modeling v2, a protein sequence pre-training model) process substrate graphs and enzyme sequences (e.g., GASP), respectively, extracting corresponding substrate and enzyme features. These features are then processed by the Adapter layer and combined before being input into the weight adjustment layer of the target protein prediction model. During training, the distance between paired enzyme-substrates is shortened, while the distance between unpaired enzyme-substrates is increased.

[0079] Furthermore, during training, a contrastive loss function is used to measure the encoder loss to guide the pre-training process. Specifically, this symmetric loss design enforces bidirectional modal alignment, ensuring that enzyme proteins can predict their corresponding substrates and substrates can also trace back to the source enzyme proteins, thereby enhancing the interoperability of the representation. This stage primarily achieves alignment of the enzyme and molecule modalities through contrastive learning. A dual encoder architecture is employed to extract features from the enzyme sequence and substrate molecule graphs, respectively. Training is performed on enzyme-substrate pairs in a shared embedding space, ensuring that matching enzyme-substrate pair features are close to each other, while non-matching pair features are mutually exclusive. It should be noted that the initial protein language model and initial graph neural network serve as the encoders to be trained. The enzyme sequence information and substrate molecule information correspond to true labels for the true enzyme features and true substrate features corresponding to the enzyme sequence information and substrate molecule information, respectively. The prediction results reflect the predicted enzyme features and predicted substrate features. The contrastive loss function optimizes the feature space by comparing the similarity of sample pairs. Its core goal is to bring similar samples closer together in the embedding space and dissimilar samples farther apart.

[0080] In the examples of this application, only enzyme sequences and substrate molecular graphs are used as input, avoiding complex feature engineering, thereby improving the versatility and scalability of the model. A contrastive learning strategy is used in the pre-training stage to make the enzyme closer to the substrate it catalyzes in the potential feature space, thereby emphasizing the specificity of enzyme catalysis. This contrastive learning method not only enhances the model's understanding of the interaction between enzymes and substrates, but also lays a solid foundation for the subsequent model weight adjustment stage.

[0081] In a feasible embodiment, the preset protein prediction model includes at least a weight adjustment layer;

[0082] The step of adjusting the weight parameters of the preset protein prediction model according to the enzyme characteristics and substrate characteristics to obtain the target protein prediction model includes:

[0083] Step S21, inputting enzyme characteristics and substrate characteristics into a preset weighted supernetwork, and the weighted supernetwork outputs target weight information;

[0084] Step S22 , adjusting the weight information of the weight adjustment layer in the preset protein prediction model according to the target weight information to obtain the target protein prediction model.

[0085] The embodiment of the present application introduces a dynamic weight generation mechanism, which is mainly based on the parameter adaptation framework of the hypernetwork. This framework can break through the limitations of the traditional static fully connected layer and realize instance-level adaptive modeling (such as Figure 4 Specifically, the main network of the protein prediction model can be defined as the main structure that performs forward inference. The weight parameters of its final weight adjustment layer are no longer fixed, but are dynamically generated by the supernetwork based on the current input enzyme-substrate features. The weight adjustment layer can be an MLP (Multilayer Perceptron).

[0086] In the embodiment of the present application, by adopting the enzyme feature and substrate feature of the aforementioned encoder output, it is realized that the connection represented by enzyme and substrate is used as meta-information, and corresponding weight is generated for each data instance, so that the initial protein prediction model is adaptively updated. This dynamic weight generation mechanism essentially constructs a smooth mapping from input space to parameter space, which makes when facing samples outside the distribution, so the model is not simply performed extrapolation prediction, but automatically adjusts the decision boundary according to the input similarity. Considering that the connection comprises a rough enzyme-substrate interaction pattern, the embodiment of the present application can carry out fast instance-level adaptation to the catalytic interaction of each enzyme-substrate pair. In addition, from the perspective of weight generation, the data outside the distribution outside the distribution outside the data may not be regarded as completely outside the distribution at the perspective of optimal weight optimization, so this helps to improve the predictive performance of the model.

[0087] It should be noted that the embodiment of the present application uses a two-stage alignment method, using enzyme-substrate features as the input of the hypernetwork to enable the model to quickly adjust weights. Contrastive learning enables enzyme-substrate features to be better extracted by the hypernetwork.

[0088] like Figure 4 As shown, enzyme and substrate features are input into a weighted hypernetwork (i.e., a weighted hypernetwork). The hypernetwork then outputs a weight stream (i.e., target weight information) to the weight adjustment layer of the initial protein prediction model for weight adjustment, resulting in a target protein prediction model. After obtaining the target protein prediction model, interaction features are predicted using the input enzyme-substrate features. These are then output to the decoder to obtain interaction prediction results, which are then applied to downstream tasks.

[0089] In the embodiment of the present application, by introducing a super network structure, it is used to generate the model weight adapted to the current task (i.e., the enzyme characteristics and substrate weights of input). The effect of the weight super network is to utilize the enzyme characteristics and substrate characteristics of the enzyme pre-trained and substrate encoder output in the first phase, so as to efficiently learn the most generalizable weight, rather than simply fitting the bad condition distribution in the training data. In this way, it is possible to when facing data outside the distribution, the target protein prediction model can still maintain higher prediction performance. Moreover, the introduction of the super network not only improves the generalization ability of the model, but also significantly reduces the number of training cycles, so that the model can converge under the number of cycles of less amount. Based on the above design, the technical scheme of the embodiment of the present application can realize better generalization ability outside the distribution on imbalanced data, overcome the shortcomings of existing methods under long-tail distribution data and few sample scenes, for biomedicine field provides more powerful tools, promotes the innovation and development of the related fields such as enzyme engineering and drug discovery.

[0090] In a feasible embodiment, the preset protein prediction model further includes a projection layer;

[0091] Before the step of adjusting the weight parameters of the preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain the target protein prediction model, the method further includes:

[0092] Step B10, obtaining enzyme feature training data and substrate feature training data;

[0093] Step B20: inputting the enzyme feature training data and the substrate feature training data into the initial weighted supernetwork and the protein pre-training model, respectively, training the projection layer and the initial weighted supernetwork in the protein pre-training model to obtain the weighted supernetwork and the preset protein prediction model;

[0094] Among them, the projection layer is used to process the enzyme features and substrate features received by the preset protein prediction model.

[0095] It is understandable that the initial protein prediction model and the preset protein prediction model all include projection layers. In the deep learning model, projection layer (Projection Layer) is a kind of important structural design, and its core function is to carry out feature mapping and dimension adjustment to input data by linear or nonlinear transformation. In the embodiment of the present application, projection layer is used to receive the enzyme characteristics and substrate characteristics output by the encoder, and feature mapping and dimensionality reduction are carried out to enzyme characteristics and substrate characteristics, so that the protein prediction model can be further analyzed and processed. Enzyme feature training data and substrate feature training data are prepared in advance with real labels for training weight super network and protein pre-training model data.

[0096] Taking into account the constraints of computing resources, the embodiment of the present application freezes the parameters of the protein pre-training model during the training phase, and only trains the projection layer therein, reducing the amount of calculation and improving training efficiency. Taking into account the complexity of the super network training itself, the embodiment of the present application can adopt a learning rate restart strategy during training. Specifically, the learning rate is reset to the initial value after a certain training step using a learning rate decay strategy. This training strategy can effectively improve the local optimal problems that the model may suffer during training, and improve the model convergence speed, so that the model has the ability to converge quickly. Exemplarily, the mean square error loss (MSE) can be used to optimize the model parameters during iterative training.

[0097] The purpose of the above training steps is to train a hypernetwork that can output accurate target weight information based on the input enzyme characteristics and substrate characteristics, as well as a preset protein prediction model that can output accurate interaction characteristics based on the input enzyme characteristics and substrate characteristics (this model also requires further weight adjustment).

[0098] Illustratively, the enzyme-substrate interaction prediction method of the embodiment of the present application can be implemented in two stages. The first stage is to use a contrastive learning strategy for pre-training. By using the enzyme-substrate sequence as input, the enzyme is brought closer to the substrate it catalyzes in the potential feature space. Thereby, an enzyme-substrate encoding model with stronger recognition ability is obtained, which enables it to better capture the complex relationship between the enzyme and the substrate. The second stage is to train the corresponding super network, and use its weight generation ability and generalization to perform targeted weight adjustment on the model, so that the model can perform targeted weight generation for different data, improve the generalization ability of the target protein prediction model, and be more adaptable to the prediction task of rare or a few enzyme-substrate pairs.

[0099] The present application embodiment utilizes the emerging technology of super network, and is intended to solve the difficult problem in the prediction of enzyme-substrate interaction of few samples. As a kind of meta-network model that can generate network parameters, super network has excellent generalization ability, can make the model quickly adapt to new tasks by learning from multiple tasks. This characteristic makes it show huge potential in the few-sample learning scenario, and can effectively improve the learning efficiency and prediction accuracy of the model under a small amount of samples. The enzyme-substrate interaction prediction method that the present application embodiment provides can not only improve the accuracy and efficiency of molecular interaction prediction, but also provide more powerful tools for biomedicine field, help researchers make faster decision in new drug development, and can also realize effective prediction of enzyme-substrate interaction when data is scarce, so as to promote the full utilization of biological data.

[0100] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the enzyme-substrate interaction prediction method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0101] The present application also provides an enzyme-substrate interaction prediction device, such as Figure 5 As shown, the enzyme-substrate interaction prediction device includes:

[0102] A feature extraction module 10 is used to extract enzyme features of the target enzyme and substrate features of the target substrate respectively through a pre-trained encoder;

[0103] A weight adjustment module 20 is used to adjust the weight parameters of the preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model;

[0104] A feature prediction module 30 is used to input the enzyme feature and the substrate feature into the target protein prediction model, and the target protein prediction model outputs the interaction feature of the target enzyme and the target substrate;

[0105] The feature decoding module 40 is used to input the interaction feature into a preset decoder, and the decoder decodes the interaction feature and outputs a prediction result of the interaction between the target enzyme and the target substrate.

[0106] In one embodiment, the encoder includes at least a protein language model and a graph neural network; the feature extraction module 10 is further configured to:

[0107] Obtaining enzyme sequence information of the target enzyme and substrate molecule information of the target substrate;

[0108] Inputting the enzyme sequence information into a preset protein language model to obtain enzyme features of the target enzyme, wherein the enzyme features include hierarchical features of the enzyme;

[0109] The substrate molecule information is input into a preset graph neural network to obtain the substrate characteristics of the target substrate, wherein the substrate characteristics include topological structure and chemical properties.

[0110] In one embodiment, the feature extraction module 10 is further configured to:

[0111] The enzyme features and the substrate features are mapped to a unified latent space, wherein the distance between the enzyme features and the substrate features in the latent space is used to characterize the interaction strength between the target enzyme and the target substrate.

[0112] In one embodiment, the enzyme-substrate interaction prediction apparatus further includes a pre-training module, which is used to:

[0113] Inputting the enzyme sequence information and the substrate molecule information into an initial protein language model and an initial graph neural network, respectively, and then the initial protein language model and the initial graph neural network respectively output prediction results corresponding to the enzyme sequence information and the substrate molecule information;

[0114] Based on a preset contrast loss function, the true labels corresponding to the enzyme sequence information and the substrate molecule information, and the prediction results, the initial protein language model and the initial graph neural network are pre-trained in a preset shared embedding space to obtain a pre-trained protein language model and graph neural network.

[0115] In one embodiment, the preset protein prediction model includes at least a weight adjustment layer; the weight adjustment module 20 is further configured to:

[0116] Inputting the enzyme characteristics and the substrate characteristics into a preset weighted supernetwork, and outputting target weight information from the weighted supernetwork;

[0117] The weight information of the weight adjustment layer in the preset protein prediction model is adjusted according to the target weight information to obtain a target protein prediction model.

[0118] In one embodiment, the preset protein prediction model further includes a projection layer, and the enzyme-substrate interaction prediction apparatus further includes a model training module, which is further configured to:

[0119] Obtaining enzyme feature training data and substrate feature training data;

[0120] Inputting the enzyme feature training data and the substrate feature training data into an initial weighted supernetwork and a protein pre-training model, respectively, training the projection layer in the protein pre-training model and the initial weighted supernetwork to obtain a weighted supernetwork and a preset protein prediction model;

[0121] The projection layer is used to process the enzyme features and substrate features received by the preset protein prediction model.

[0122] The enzyme-substrate interaction prediction device provided in the embodiments of the present application, which employs the enzyme-substrate interaction prediction method of the above-described embodiments, can address the technical problem that conventional enzyme-substrate pair interaction prediction schemes have poor interaction prediction results for rare enzyme-substrate pairs. Compared to the prior art, the beneficial effects of the enzyme-substrate interaction prediction device provided in the embodiments of the present application are the same as those of the enzyme-substrate interaction prediction method provided in the above-described embodiments, and other technical features of the enzyme-substrate interaction prediction device are the same as those disclosed in the above-described embodiments, which are not further described herein.

[0123] An embodiment of the present application also provides an enzyme-substrate interaction prediction device, which includes at least: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the enzyme-substrate interaction prediction method in the above embodiment.

[0124] Reference below Figure 6 , which shows a structural schematic diagram of a control unit suitable for implementing an embodiment of the present application. Figure 6 The control unit shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0125] like Figure 6As shown, the control unit may include a processing device 101 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 102 or programs loaded from a storage device 103 into a random access memory (RAM) 104. RAM 104 also stores various programs and data required for the operation of the control unit. Processing device 101, ROM 102, and RAM 104 are interconnected via a bus 105. An input / output (I / O) interface 106 is also connected to the bus. Typically, the following systems may be connected to I / O interface 106: input devices 107 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 108 including, for example, a liquid crystal display (LCD), speakers, vibrator, etc.; storage device 103 including, for example, a magnetic tape, hard disk, etc.; and communication device 109. Communication device 109 may allow the control unit to communicate with other devices wirelessly or by wire to exchange data. Although the figures show a control unit with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented or have instead.

[0126] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 103, or installed from a ROM 102. When the computer program is executed by the processing device 101, the above-mentioned functions defined in the method of the embodiment of the present application are performed.

[0127] The enzyme-substrate interaction prediction device provided in the embodiments of the present application, which utilizes the enzyme-substrate interaction prediction method of the above-described embodiments, can address the technical problem that conventional enzyme-substrate pair interaction prediction schemes have poor results in predicting interactions for rare enzyme-substrate pairs. Compared to the prior art, the beneficial effects of the enzyme-substrate interaction prediction device provided in the embodiments of the present application are the same as those of the enzyme-substrate interaction prediction method provided in the above-described embodiments, and other technical features of the enzyme-substrate interaction prediction device are the same as those disclosed in the above-described embodiments, which are not further described herein.

[0128] It should be understood that the various parts of the embodiments of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any appropriate manner in any one or more embodiments or examples.

[0129] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the above claims.

[0130] An embodiment of the present application further provides a computer-readable storage medium storing a computer program executable on a processor, wherein the computer program is used to execute the enzyme-substrate interaction prediction method in the above embodiment.

[0131] The computer-readable storage medium provided in the embodiments of the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in combination with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0132] The computer-readable storage medium may be included in the enzyme-substrate interaction prediction device; or may exist independently without being assembled into the enzyme-substrate interaction prediction device.

[0133] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the enzyme-substrate interaction prediction device, the enzyme-substrate interaction prediction device executes: extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate respectively through a pre-trained encoder; adjusting the weight parameters of a preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model; inputting the enzyme characteristics and the substrate characteristics into the target protein prediction model, and the target protein prediction model outputs the interaction characteristics of the target enzyme and the target substrate; inputting the interaction characteristics into a preset decoder, and the decoder decodes the interaction characteristics and outputs the interaction prediction result of the target enzyme and the target substrate.

[0134] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0135] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0136] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0137] The computer-readable storage medium provided in the embodiment of the present application stores computer-readable program instructions for executing the above-mentioned enzyme-substrate interaction prediction method, which can solve the technical problem that traditional enzyme-substrate pair interaction prediction schemes have poor interaction prediction effects on rare enzyme-substrate pairs. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the embodiment of the present application are the same as the beneficial effects of the enzyme-substrate interaction prediction method provided in the above-mentioned embodiment, and are not described in detail here.

[0138] An embodiment of the present application also provides a computer program product, including a computer program, which implements the steps of the enzyme-substrate interaction prediction method as described above when the computer program is executed by a processor.

[0139] The computer program products provided in the embodiments of the present application can address the technical problem that conventional enzyme-substrate pair interaction prediction schemes have poor prediction results for rare enzyme-substrate pair interactions. Compared to the prior art, the beneficial effects of the computer program products provided in the embodiments of the present application are the same as those of the enzyme-substrate interaction prediction methods provided in the above-mentioned embodiments, and are not further elaborated here.

[0140] The above is only an exemplary solution of the present application and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the description and drawings of this application, or directly or indirectly applied in other related technical fields, are also included in the patent processing scope of the present application.

Claims

1. A method for predicting enzyme-substrate interactions, characterized in that: The enzyme-substrate interaction prediction method comprises: The enzyme features of the target enzyme and the substrate features of the target substrate are extracted respectively through the pre-trained encoder; Adjusting the weight parameters of a preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model; Inputting the enzyme characteristics and the substrate characteristics into the target protein prediction model, and outputting the interaction characteristics of the target enzyme and the target substrate from the target protein prediction model; The interaction characteristics are input into a preset decoder, which decodes the interaction characteristics and outputs a predicted result of the interaction between the target enzyme and the target substrate.

2. The enzyme-substrate interaction prediction method according to claim 1, wherein The encoder includes at least a protein language model and a graph neural network; The steps of respectively extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate by the pre-trained encoder include: Obtaining enzyme sequence information of the target enzyme and substrate molecule information of the target substrate; Inputting the enzyme sequence information into a preset protein language model to obtain enzyme features of the target enzyme, wherein the enzyme features include hierarchical features of the enzyme; The substrate molecule information is input into a preset graph neural network to obtain the substrate characteristics of the target substrate, wherein the substrate characteristics include topological structure and chemical properties.

3. The enzyme-substrate interaction prediction method according to claim 2, wherein: After the step of inputting the substrate molecule information into a preset graph neural network to obtain substrate characteristics of the target substrate, the method further includes: The enzyme features and the substrate features are mapped to a unified latent space, wherein the distance between the enzyme features and the substrate features in the latent space is used to characterize the interaction strength between the target enzyme and the target substrate.

4. The enzyme-substrate interaction prediction method according to claim 2, wherein Before the step of respectively extracting the enzyme characteristics of the target enzyme and the substrate characteristics of the target substrate by the pre-trained encoder, the method further comprises: Inputting the enzyme sequence information and the substrate molecule information into an initial protein language model and an initial graph neural network, respectively, and then the initial protein language model and the initial graph neural network respectively output prediction results corresponding to the enzyme sequence information and the substrate molecule information; Based on a preset contrast loss function, the true labels corresponding to the enzyme sequence information and the substrate molecule information, and the prediction results, the initial protein language model and the initial graph neural network are pre-trained in a preset shared embedding space to obtain a pre-trained protein language model and graph neural network.

5. The enzyme-substrate interaction prediction method according to claim 1, wherein The preset protein prediction model at least includes a weight adjustment layer; The step of adjusting the weight parameters of the preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain the target protein prediction model comprises: Inputting the enzyme characteristics and the substrate characteristics into a preset weighted supernetwork, and outputting target weight information from the weighted supernetwork; The weight information of the weight adjustment layer in the preset protein prediction model is adjusted according to the target weight information to obtain a target protein prediction model.

6. The enzyme-substrate interaction prediction method according to claim 5, wherein The preset protein prediction model further includes a projection layer; Before the step of adjusting the weight parameters of the preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain the target protein prediction model, the method further includes: Obtaining enzyme feature training data and substrate feature training data; Inputting the enzyme feature training data and the substrate feature training data into an initial weighted supernetwork and a protein pre-training model, respectively, training the projection layer in the protein pre-training model and the initial weighted supernetwork to obtain a weighted supernetwork and a preset protein prediction model; The projection layer is used to process the enzyme features and substrate features received by the preset protein prediction model.

7. An enzyme-substrate interaction prediction device, characterized in that: The enzyme-substrate interaction prediction device comprises: A feature extraction module is used to extract the enzyme features of the target enzyme and the substrate features of the target substrate respectively through the pre-trained encoder; A weight adjustment module, configured to adjust the weight parameters of a preset protein prediction model according to the enzyme characteristics and the substrate characteristics to obtain a target protein prediction model; A feature prediction module, configured to input the enzyme feature and the substrate feature into the target protein prediction model, and output the interaction feature of the target enzyme and the target substrate from the target protein prediction model; The feature decoding module is used to input the interaction feature into a preset decoder, and the decoder decodes the interaction feature and outputs the interaction prediction result between the target enzyme and the target substrate.

8. An enzyme-substrate interaction prediction device, characterized in that: The enzyme-substrate interaction prediction device comprises at least: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the enzyme-substrate interaction prediction method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program for implementing the enzyme-substrate interaction prediction method, and the program for implementing the enzyme-substrate interaction prediction method is executed by a processor to implement the steps of the enzyme-substrate interaction prediction method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the enzyme-substrate interaction prediction method according to any one of claims 1 to 6 are implemented.

Images