A method and model for predicting ubiquitin ligase association with substrates

Abstract

The application discloses a method and a model for predicting ubiquitination enzyme and substrate association, the method comprises the following steps: obtaining ubiquitination enzyme protein data and substrate protein data; determining ubiquitination enzyme feature vectors and substrate feature vectors according to the ubiquitination enzyme protein data and the substrate protein data; obtaining initial mapping relationship corresponding to the ubiquitination enzyme and the substrate, and determining target mapping relationship corresponding to the ubiquitination enzyme and the substrate according to the initial mapping relationship, the ubiquitination enzyme feature vectors and the substrate feature vectors. Since the application obtains ubiquitination enzyme protein data and substrate protein data, and predicts the target mapping relationship between the ubiquitination enzyme and the substrate according to the ubiquitination enzyme protein data and the substrate protein data, the biological data types and biological knowledge involved are less, and thus the problem that the prior art is strongly dependent on field knowledge, needs multiple types of biological data, and makes the model difficult to be popularized and applied can be effectively solved.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics, and more particularly to a method and model for predicting the association between ubiquitinizing enzymes and substrates. Background Technology

[0002] Protein ubiquitination is a highly regulated process involving the covalent binding of ubiquitin molecules to specific lysine residues on target proteins via an enzyme cascade, including ubiquitin activators (E1), ubiquitin conjugators (E2), and ubiquitin ligases (E3), leading to substrate degradation. This post-translational modification plays a crucial regulatory role in various biological processes, such as protein degradation, DNA damage repair, signal transduction, growth, and metabolism. The dynamic balance between E3-mediated ubiquitination and DUB (deubiquitinating enzyme)-mediated deubiquitination determines the level of protein ubiquitination, ultimately determining protein stability and interactions. Dysregulation or disruption of dynamic ubiquitination and deubiquitination processes is associated with a wide range of human diseases, including neurodegenerative diseases, inflammation, and cancer, highlighting the critical role of these processes in maintaining cellular homeostasis and health. Given the central role of DUB in tumorigenesis and the specificity of E3, these enzymes have emerged as potential therapeutic targets for cancer treatment. E3 is recognized for its diversity and specificity, while DUB has been identified as a potential target involved in malignant progression. Therefore, elucidating the interaction patterns between E3 / DUB and its substrates provides a new perspective for developing disease treatment interventions.

[0003] Currently, experimental methods have been developed for identifying interactions between E3 / DUB and its substrates, including protein microarrays, GPS analysis, mass spectrometry, live phage display libraries, and biochemical experiments such as two-hybrid screening and immunoprecipitation. However, these methods are often laborious, expensive, and inefficient due to the low expression levels of E3 / DUB substrates and their weak intrinsic interactions. Recently, some high-throughput methods have been used to identify ESI, but they still produce false findings and are far from achieving ESI identification across the entire proteome. Therefore, more efficient and accurate methods are needed to identify E3 / DUB-substrate interactions, which could provide new perspectives for therapeutic interventions in diseases.

[0004] Identifying E3 / DUB substrate interactions is typically time-consuming and laborious, leading to the development of several computational methods to assist this process. Li et al. developed a Naive Bayes-based classifier model that incorporates various types of biological data, including gene ontology term pairs, enriched domain pairs, E3 / DUB recognition consensus motifs, and protein-protein interaction network loops, to predict potential interactions between E3 and DUB. Similarly, Chen et al. constructed a prediction model based on a random forest classifier that combines omics with network and pathway data to predict potential E3-substrate interactions. These computational methods provide an efficient approach to identifying ESI / DSI, but their strong reliance on domain knowledge and the need for diverse types of biological data hinder their generalization.

[0005] Therefore, existing technologies still need improvement and development. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and model for predicting the association between ubiquitination enzymes and substrates, in order to address the above-mentioned deficiencies of the prior art. This aims to solve the problem that the prior art relies heavily on domain knowledge and requires various types of biological data, making it difficult for the model to be widely applied.

[0007] The technical solution adopted by this invention to solve the problem is as follows:

[0008] In a first aspect, embodiments of the present invention provide a method for predicting the association between ubiquitinizing enzymes and substrates, wherein the method includes:

[0009] Acquire ubiquitination enzyme protein data and substrate protein data;

[0010] The ubiquitination enzyme feature vector and the substrate feature vector are determined based on the ubiquitination enzyme protein data and the substrate protein data, respectively.

[0011] Obtain the initial mapping relationship between ubiquitinase and substrate, and determine the target mapping relationship between ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector.

[0012] In one implementation method, determining the ubiquitination enzyme feature vector and the substrate feature vector based on the ubiquitination enzyme protein data and the substrate protein data, respectively, includes:

[0013] Based on the ubiquitination enzyme protein data and the substrate protein data, the initial feature vector of the ubiquitination enzyme and the initial feature vector of the substrate are determined respectively.

[0014] The initial feature vector of the ubiquitinase and the initial feature vector of the substrate are respectively input into a one-dimensional convolutional neural network to obtain the feature vector of the ubiquitinase and the feature vector of the substrate.

[0015] In one implementation method, determining the initial feature vector of the ubiquitination enzyme and the initial feature vector of the substrate based on the ubiquitination enzyme protein data and the substrate protein data, respectively, includes:

[0016] Based on the ubiquitination enzyme protein data and the substrate protein data, the ubiquitination enzyme protein sequence and the substrate protein sequence were determined, respectively.

[0017] The ubiquitination enzyme protein sequence and the substrate protein sequence are respectively input into a bidirectional language model to obtain ubiquitination enzyme protein representation vector and substrate protein representation vector.

[0018] The initial feature vectors of the ubiquitination enzyme and the substrate are obtained by averaging the amino acid features in each ubiquitination enzyme protein representation vector and substrate protein representation vector, respectively.

[0019] In one implementation, determining the ubiquitinated protein sequence and the substrate protein sequence based on the ubiquitination enzyme protein data and the substrate protein data, respectively, includes:

[0020] For each different type of amino acid encoding, determine the amino acid encoding data;

[0021] The amino acids in the ubiquitinated enzyme protein data and the substrate protein data are encoded according to the amino acid encoding data to determine the ubiquitinated protein sequence and the substrate protein sequence.

[0022] In one implementation method, determining the target mapping relationship between the ubiquitinase and the substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector includes:

[0023] The ubiquitination enzyme feature vector and the substrate feature vector are concatenated to obtain an initial concatenation feature vector;

[0024] The initial mapping relationship and the initial connection feature vector are input into a Bayesian neural network to obtain the target mapping relationship.

[0025] In one implementation, the Bayesian neural network includes:

[0026] The posterior distribution of the weight parameters in the Bayesian neural network is determined based on the initial mapping relationship and the initial connection feature vector.

[0027] Obtain the approximate weight posterior distribution controlled by parameter θ, minimize the KL divergence between the posterior distribution and the approximate weight posterior distribution, and determine the optimization function of parameter θ;

[0028] Implement a Bayesian neural network based on the optimization function.

[0029] In one implementation, the optimization function for the parameter θ is:

[0030] Where, q θ (w) is the approximate weight posterior distribution controlled by parameter θ, Y is the initial mapping relationship, X is the initial connection feature vector, w is the weight parameter, p(w) is the prior distribution of the weight parameter, and p(Y|X,w) is the conditional distribution model.

[0031] Secondly, embodiments of the present invention also provide a model for predicting the association between ubiquitinase and substrate, wherein the apparatus for predicting the association between ubiquitinase and substrate includes:

[0032] The feature vector generation module includes a bidirectional language model and a one-dimensional convolutional neural network, used to acquire ubiquitination enzyme protein data and substrate protein data; and to determine ubiquitination enzyme feature vector and substrate feature vector based on the ubiquitination enzyme protein data and the substrate protein data, respectively.

[0033] The mapping relationship prediction module includes a Bayesian neural network, which is used to obtain the initial mapping relationship between ubiquitinase and substrate, and to determine the target mapping relationship between ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector.

[0034] Thirdly, embodiments of the present invention also provide a terminal, characterized in that the terminal includes a memory and one or more processors; the memory stores one or more programs; the programs include instructions for performing the method for predicting the association between ubiquitinase and substrate as described above; and the processor is used to execute the programs.

[0035] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having stored thereon a plurality of instructions, wherein the instructions are adapted to be loaded and executed by a processor to implement any of the methods described above for predicting the association between ubiquitinases and substrates.

[0036] The beneficial effects of this invention are as follows: In this embodiment, the invention acquires ubiquitinase protein data and substrate protein data; determines ubiquitinase feature vectors and substrate feature vectors based on the ubiquitinase protein data and substrate protein data, respectively; obtains the initial mapping relationship between the ubiquitinase and substrate; and determines the target mapping relationship between the ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vectors, and the substrate feature vectors. Because this invention acquires ubiquitinase protein data and substrate protein data, and predicts the target mapping relationship between the ubiquitinase and substrate based on the ubiquitinase protein data and substrate protein data, it involves fewer types of biological data and requires less biological knowledge. Therefore, it can effectively solve the problem that existing technologies rely heavily on domain knowledge and require multiple types of biological data, making it difficult to widely apply the model. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating a method for predicting the association between ubiquitinase and substrate, provided in an embodiment of the present invention.

[0039] Figure 2 This is a schematic diagram of data collection for a method to predict the association between ubiquitination enzymes and substrates provided in an embodiment of the present invention.

[0040] Figure 3 This is a schematic diagram of a pre-trained model for a method to predict the association between ubiquitinases and substrates provided in an embodiment of the present invention.

[0041] Figure 4 This is a schematic diagram of a Bayesian neural network for predicting the association between ubiquitination enzymes and substrates, provided in an embodiment of the present invention.

[0042] Figure 5 This is a schematic diagram of the internal modules of a model for predicting the association between ubiquitinases and substrates, provided in an embodiment of the present invention.

[0043] Figure 6 This is a schematic diagram of the terminal provided in the embodiment of the present invention. Detailed Implementation

[0044] This invention discloses a method and model for predicting the association between ubiquitinizing enzymes and substrates. To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0045] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0046] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0047] The interaction patterns between ubiquitinases and their substrates offer new perspectives for developing therapeutic interventions for diseases. Currently, experimental methods have been developed to identify interactions between E3 / DUB and their substrates, including protein microarrays, GPS analysis, mass spectrometry, live phage display libraries, and biochemical experiments such as two-hybrid screening and immunoprecipitation. However, these methods are often laborious, expensive, and inefficient due to the low expression levels and weak intrinsic interactions of E3 / DUB substrates. Recently, some high-throughput methods have been used to identify ESIs, but they still produce false findings and are far from achieving whole-proteome ESI identification. Therefore, more efficient and accurate methods are needed to identify E3 / DUB-substrate interactions, which could provide new perspectives for therapeutic interventions for diseases.

[0048] Identifying E3 / DUB substrate interactions is typically time-consuming and laborious, leading to the development of several computational methods to assist this process. Li et al. developed a Naive Bayes-based classifier model that incorporates various types of biological data, including gene ontology term pairs, enriched domain pairs, E3 / DUB recognition consensus motifs, and protein-protein interaction network loops, to predict potential interactions between E3 and DUB. Similarly, Chen et al. constructed a prediction model based on a random forest classifier that combines omics with network and pathway data to predict potential E3-substrate interactions. These computational methods provide an efficient approach to identifying ESI / DSI, but their strong reliance on domain knowledge and the need for diverse types of biological data hinder their generalization.

[0049] To address the aforementioned shortcomings of existing technologies, this invention provides a method for predicting the association between ubiquitinases and substrates. The method involves acquiring ubiquitinase protein data and substrate protein data; determining ubiquitinase feature vectors and substrate feature vectors based on the ubiquitinase protein data and substrate protein data, respectively; obtaining an initial mapping relationship between the ubiquitinase and substrate; and determining a target mapping relationship between the ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vectors, and the substrate feature vectors. Because this invention acquires ubiquitinase protein data and substrate protein data, and predicts the target mapping relationship between ubiquitinase and substrate based on these data, it involves fewer types of biological data and requires less biological knowledge. Therefore, it effectively solves the problem that existing technologies rely heavily on domain knowledge and require various types of biological data, making it difficult to widely apply the model.

[0050] Exemplary methods

[0051] like Figure 1 As shown, the method includes:

[0052] Step S100: Obtain ubiquitinase protein data and substrate protein data.

[0053] Specifically, the ubiquitination enzyme protein data includes protein data corresponding to ubiquitin ligases and protein data corresponding to deubiquitination enzymes. To identify the interaction between ubiquitination ligases and substrates, or the interaction between deubiquitination enzymes and substrates, this embodiment extracts the protein data corresponding to ubiquitin ligases and substrates from a ubiquitination database, such as... Figure 2 As shown in the figure. This embodiment uses ubiquitination enzyme protein data and substrate protein data as data to identify the relationship between ubiquitination enzymes and substrates. The amount of biological data used is relatively small, which makes the method in this embodiment easier to promote and apply.

[0054] Step S200: Determine the ubiquitination enzyme feature vector and the substrate feature vector based on the ubiquitination enzyme protein data and the substrate protein data, respectively.

[0055] In simple terms, this embodiment first determines the ubiquitination enzyme feature vector based on ubiquitination enzyme protein data, and then determines the substrate feature vector based on substrate protein data, thereby enabling the analysis of ubiquitination enzyme protein data and substrate protein data based on deeper protein characteristics.

[0056] In one implementation, step S200 specifically includes:

[0057] Step S201: Determine the initial feature vector of the ubiquitination enzyme and the initial feature vector of the substrate based on the ubiquitination enzyme protein data and the substrate protein data, respectively.

[0058] Step S202: Input the initial feature vector of the ubiquitination enzyme and the initial feature vector of the substrate into a one-dimensional convolutional neural network to obtain the feature vector of the ubiquitination enzyme and the feature vector of the substrate.

[0059] Specifically, this embodiment first determines the initial feature vector of the ubiquitinase based on the ubiquitinase protein data, and then determines the feature vector of the substrate protein based on the substrate protein data. To further plan protein-level features and reduce the dimensionality of the initial ubiquitinase feature vector and the substrate protein feature vector, the initial ubiquitinase feature vector and the initial substrate feature vector are input into a one-dimensional convolutional neural network (CNN). The one-dimensional CNN in this embodiment consists of two cascaded convolutional layers, each characterized by the size and number of its convolutional kernels. The hyperparameters in the one-dimensional CNN are optimized to achieve optimal network performance. The first layer of the one-dimensional CNN has a kernel size and number of 8 and 16, respectively, while the second layer has a kernel size and number of 8 and 32, respectively. To activate the output of each convolutional layer, the one-dimensional CNN employs a modified linear unit (ReLU) nonlinear function, defined as follows:

[0060]

[0061] In one implementation, determining the initial feature vector of the ubiquitinase and the initial feature vector of the substrate based on the ubiquitinase protein data and the substrate protein data, respectively, includes:

[0062] Step S2011: Determine the ubiquitination enzyme protein sequence and the substrate protein sequence based on the ubiquitination enzyme protein data and the substrate protein data, respectively.

[0063] Step S2012: Input the ubiquitination enzyme protein sequence and the substrate protein sequence into the bidirectional language model to obtain the ubiquitination enzyme protein representation vector and the substrate protein representation vector, respectively.

[0064] Step S2013: Average the amino acid features in each ubiquitination enzyme protein representation vector and substrate protein representation vector to obtain the initial feature vector of the ubiquitination enzyme and the initial feature vector of the substrate.

[0065] Specifically, the ubiquitinase protein data and substrate protein data are pre-trained to obtain initial feature vectors for the ubiquitinase and substrate. To enable the data to be input into the pre-trained model, this embodiment first determines the ubiquitinase protein sequence based on the ubiquitinase protein data and the substrate protein sequence based on the substrate protein data. This determination method involves: encoding different types of amino acids to determine amino acid encoding data; and then encoding the amino acids in the ubiquitinase protein data and the substrate protein data according to the amino acid encoding data to determine the ubiquitinated protein sequence and the substrate protein sequence. Considering that all proteins are composed of 20 natural amino acids, which are represented by 20 different letters in the alphabet, in this embodiment, the protein data corresponding to a ubiquitin ligase / deubiquitinase / substrate containing one amino acid can be represented as x = x1x2...x l , where x t (1≤t≤l)) Each represents an amino acid from the 20 standard amino acids. It is worth noting that this embodiment uses only protein data as raw data, and does not use other biological data that provide advanced information about proteins.

[0066] To encode ubiquitinase and substrate protein sequences into protein-level representation vectors, this embodiment uses a pre-trained bidirectional language model from the seqVec repository. For a protein sequence of length L, seqVec outputs an embedding matrix X. p ∈R L×D Where D = 1024, corresponding to a 1024-dimensional vector for each amino acid, thus obtaining the ubiquitinase protein representation vector and the substrate protein representation vector. Based on the characteristics of each amino acid in the amino acid sequence of the ubiquitinase protein representation vector and the substrate protein representation vector, i.e., X... p Protein-level feature x is obtained by averaging L amino acid features. p ∈R D ,like Figure 3 As shown.

[0067] Step S300: Obtain the initial mapping relationship between ubiquitinase and substrate, and determine the target mapping relationship between ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector.

[0068] Specifically, the mapping relationship between ubiquitinase and substrate is extracted from the ubiquitination database, i.e., the connection between ubiquitin ligase and substrate or the connection between deubiquitinase and substrate. In this embodiment, the initial mapping relationship is the known connection relationship between ubiquitinase and substrate. This connection relationship may contain erroneous values, and there may be other unknown connection relationships between ubiquitinase and substrate. Therefore, in order to confirm the accuracy of the initial mapping relationship and to obtain unknown mapping relationships, this embodiment uses the initial mapping relationship between ubiquitinase and substrate as a reference, and learns and predicts the target mapping relationship between ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector.

[0069] In one implementation, step S300 specifically includes:

[0070] Step S301: Connect the ubiquitination enzyme feature vector and the substrate feature vector to obtain an initial connection feature vector;

[0071] Step S302: Input the initial mapping relationship and the initial connection feature vector into the Bayesian neural network to obtain the target mapping relationship.

[0072] Specifically, the ubiquitination enzyme feature vector and the substrate characteristic vector are concatenated to obtain an initial connection feature vector. This initial mapping relationship and the initial connection feature vector are then input into a Bayesian neural network for learning. The target mapping relationship is then predicted by the Bayesian neural network. By using a Bayesian neural network to predict the target mapping relationship, the reliance on domain knowledge is reduced, and because it is based on a deep learning model, it has high accuracy, which facilitates the widespread use of the model.

[0073] In one implementation, the Bayesian neural network includes:

[0074] The posterior distribution of the weight parameters in the Bayesian neural network is determined based on the initial mapping relationship and the initial connection feature vector.

[0075] Obtain the approximate weight posterior distribution controlled by parameter θ, minimize the KL divergence between the posterior distribution and the approximate weight posterior distribution, and determine the optimization function of parameter θ;

[0076] Implement a Bayesian neural network based on the optimization function.

[0077] Specifically, such as Figure 4As shown, in a given training set {X,Y}, where X represents protein features (i.e., the initial connectivity feature vector) and Y represents the presence or absence of interactions (i.e., the initial mapping relationship), the probability of the model is represented as p(Y|X,w), and p(w) is the prior distribution. The model weight parameter set w={W1,W2,...,W k It has a k-layer structure. In the Bayesian network framework, the problem of predicting the distribution of the model can be defined as the integral of the product of the likelihood distribution and the prior distribution with respect to the weight parameters w, as shown below:

[0078] p(y * |x * ,X,Y)=∫p(y * |x * ,w)p(w|X,Y)dw

[0079] Among them, y * Is input x * The output, x * In this embodiment, it is equivalent to X. The posterior calculation of p(w|X,Y) is as follows:

[0080]

[0081] However, directly calculating the predicted distribution of the above model in a neural network is computationally expensive, and the posterior distribution is difficult to compute analytically. To address these limitations, variational inference uses an approximate weighted posterior distribution q controlled by a parameter θ. θ (w). q θ The KL divergence between (w) and the true posterior distribution is minimized to obtain a tractable distribution, with the equation:

[0082]

[0083] By combining the above formulas for calculating the posterior distribution of the weight parameters and the formula for calculating the KL divergence, the negative likelihood lower bound is minimized, and the optimization function of parameter θ can be obtained. The Bayesian neural network can then be implemented based on the optimization function of parameter θ.

[0084] Obtain the approximate weighted posterior distribution controlled by parameter θ, minimize the KL divergence between the posterior distribution and the approximate weighted posterior distribution, and determine the optimization function for parameter θ.

[0085] The optimization function for parameter θ is:

[0086] Where, q θ(w) is the approximate weight posterior distribution controlled by parameter θ, Y is the initial mapping relation, X is the initial connection feature vector, w is the weight parameter, p(w) is the prior distribution of the weight parameter, and p(Y|X,w) is the conditional distribution model. Additionally, the first term in the formula, i.e., -∫q... θ (w)logp(Y|X,w)dw is the expected log-likelihood, used to adjust model parameters to better fit the observed data, thus giving more weight to the latent variables. The second term, KL(q) θ The KL divergence (p(w)) measures the approximate weight posterior distribution and the prior distribution, used to ensure that the approximate weight posterior distribution is similar to the prior distribution by adjusting the model parameters. The implementation of a tractable distribution (approximate weight posterior distribution) is crucial for implementing Bayesian neural networks. This embodiment employs the MC-dropout method, which uses the product of Bernoulli distributions to approximate the posterior distribution of network units. The MC-dropout method is feasible because a tractable posterior distribution can be directly obtained without additional parameters. Furthermore, when the dropout layer is activated, the expectation and variance of the output can be easily computed by repeatedly sampling the output of the input.

[0087] In this embodiment, the AUC value reached 95.93% in the task of predicting the interaction between ubiquitin ligase and substrate, and the AUC value reached 86.02% in the task of predicting the interaction between deubiquitinating enzyme and substrate.

[0088] Mathematical definition of AUC value:

[0089]

[0090] Where M represents the positive sample set, N represents the negative sample set, and rank i Let i represent the predicted score ranking of edge i.

[0091] This embodiment uses a Bayesian neural network to predict the interaction between potential ubiquitinases and substrates, i.e., the target mapping relationship, based on the initial mapping relationship and initial connection feature vectors. Specifically, users can query potential substrates by providing the ubiquitinase. Similarly, users can also query potential ubiquitinases of substrates.

[0092] Based on the above embodiments, the present invention also provides a model for predicting the association between ubiquitinizing enzymes and substrates, such as... Figure 5 As shown, the model includes:

[0093] Feature vector generation module 01 includes a bidirectional language model and a one-dimensional convolutional neural network, used to acquire ubiquitination enzyme protein data and substrate protein data; and to determine ubiquitination enzyme feature vector and substrate feature vector based on the ubiquitination enzyme protein data and the substrate protein data, respectively.

[0094] The mapping relationship prediction module 02 includes a Bayesian neural network, which is used to obtain the initial mapping relationship between ubiquitinase and substrate, and to determine the target mapping relationship between ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector.

[0095] Based on the above embodiments, the present invention also provides a terminal, the principle block diagram of which can be as follows: Figure 6 As shown, the terminal includes a processor, memory, network interface, and display screen connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for predicting the association between ubiquitin-forming enzymes and substrates. The display screen can be a liquid crystal display (LCD) or an e-ink display.

[0096] Those skilled in the art will understand that Figure 6 The schematic diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the terminal to which the present invention is applied. A specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0097] In one implementation, the terminal's memory stores one or more programs, and these programs are configured to be executed by one or more processors, and the programs contain instructions for performing methods for predicting the association of ubiquitinases with substrates.

[0098] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0099] In summary, this invention discloses a method and model for predicting the association between ubiquitinases and substrates. The method involves acquiring ubiquitinase protein data and substrate protein data; determining ubiquitinase feature vectors and substrate feature vectors based on the ubiquitinase protein data and substrate protein data, respectively; obtaining an initial mapping relationship between the ubiquitinase and substrate; and determining a target mapping relationship between the ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vectors, and the substrate feature vectors. Because this invention acquires ubiquitinase protein data and substrate protein data, and predicts the target mapping relationship between ubiquitinase and substrate based on these data, it involves fewer types of biological data and requires less biological knowledge. Therefore, it effectively solves the problem that existing technologies rely heavily on domain knowledge and require various types of biological data, making it difficult to widely apply the model.

[0100] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for predicting the association between ubiquitinase and substrate, characterized in that, The method includes: Acquire ubiquitination enzyme protein data and substrate protein data; The ubiquitination enzyme feature vector and the substrate feature vector are determined based on the ubiquitination enzyme protein data and the substrate protein data, respectively. Obtain the initial mapping relationship between ubiquitinase and substrate; determine the target mapping relationship between ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector; wherein the initial mapping relationship is the known connection relationship between ubiquitinase and substrate. The step of determining the target mapping relationship between the ubiquitinase and the substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector includes: The ubiquitination enzyme feature vector and the substrate feature vector are concatenated to obtain an initial concatenation feature vector; The initial mapping relationship and the initial connection feature vector are input into a Bayesian neural network to obtain the target mapping relationship; The Bayesian neural network includes: The posterior distribution of the weight parameters in the Bayesian neural network is determined based on the initial mapping relationship and the initial connection feature vector. Get parameters The parameters are determined by minimizing the KL divergence between the controlled approximate weighted posterior distribution and the posterior distribution. The optimization function; Implement the Bayesian neural network based on the optimization function; The parameters The optimization function is: ,in, For parameters The approximate weight posterior distribution is controlled, Y is the initial mapping relationship, and X is the initial connection feature vector. The weight parameter is... The prior distribution of the weight parameters, This is a conditional distribution model.

2. The method for predicting ubiquitination enzyme-substrate association according to claim 1, characterized in that, The step of determining the ubiquitination enzyme feature vector and the substrate feature vector based on the ubiquitination enzyme protein data and the substrate protein data, respectively, includes: Based on the ubiquitination enzyme protein data and the substrate protein data, the initial feature vector of the ubiquitination enzyme and the initial feature vector of the substrate are determined respectively. The initial feature vector of the ubiquitinase and the initial feature vector of the substrate are respectively input into a one-dimensional convolutional neural network to obtain the feature vector of the ubiquitinase and the feature vector of the substrate.

3. The method for predicting ubiquitination enzyme-substrate association according to claim 2, characterized in that, The step of determining the initial feature vector of the ubiquitination enzyme and the initial feature vector of the substrate based on the ubiquitination enzyme protein data and the substrate protein data, respectively, includes: Based on the ubiquitination enzyme protein data and the substrate protein data, the ubiquitination enzyme protein sequence and the substrate protein sequence were determined, respectively. The ubiquitination enzyme protein sequence and the substrate protein sequence are respectively input into a bidirectional language model to obtain ubiquitination enzyme protein representation vector and substrate protein representation vector. The initial feature vectors of the ubiquitination enzyme and the substrate are obtained by averaging the amino acid features in each ubiquitination enzyme protein representation vector and substrate protein representation vector, respectively.

4. The method for predicting ubiquitination enzyme-substrate association according to claim 3, characterized in that, The step of determining the ubiquitinated protein sequence and the substrate protein sequence based on the ubiquitination enzyme protein data and the substrate protein data, respectively, includes: For each different type of amino acid encoding, determine the amino acid encoding data; The amino acids in the ubiquitinated enzyme protein data and the substrate protein data are encoded according to the amino acid encoding data to determine the ubiquitinated protein sequence and the substrate protein sequence.

5. An apparatus for predicting the association between ubiquitinase and substrate, characterized in that, The device includes: The feature vector generation module includes a bidirectional language model and a one-dimensional convolutional neural network, used to acquire ubiquitination enzyme protein data and substrate protein data; and to determine ubiquitination enzyme feature vector and substrate feature vector based on the ubiquitination enzyme protein data and the substrate protein data, respectively. A mapping relationship prediction module, comprising a Bayesian neural network, is used to obtain an initial mapping relationship between ubiquitinase and substrate, and to determine a target mapping relationship between ubiquitinase and substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector. The initial mapping relationship is a known connection relationship between the ubiquitinase and substrate. The step of determining the target mapping relationship between the ubiquitinase and the substrate based on the initial mapping relationship, the ubiquitinase feature vector, and the substrate feature vector includes: The ubiquitination enzyme feature vector and the substrate feature vector are concatenated to obtain an initial concatenation feature vector; The initial mapping relationship and the initial connection feature vector are input into a Bayesian neural network to obtain the target mapping relationship; The Bayesian neural network includes: The posterior distribution of the weight parameters in the Bayesian neural network is determined based on the initial mapping relationship and the initial connection feature vector. Get parameters The parameters are determined by minimizing the KL divergence between the controlled approximate weighted posterior distribution and the posterior distribution. The optimization function; Implement the Bayesian neural network based on the optimization function; The parameters The optimization function is: ,in, For parameters The approximate weight posterior distribution is controlled, Y is the initial mapping relationship, and X is the initial connection feature vector. The weight parameter is... The prior distribution of the weight parameters, This is a conditional distribution model.

6. A terminal, characterized in that, The terminal includes a memory and one or more processors; the memory stores one or more programs; the programs contain instructions for performing the method for predicting the association between ubiquitinase and substrate as described in any one of claims 1-4; the processor is used to execute the programs.

7. A computer-readable storage medium storing a plurality of instructions thereon, characterized in that, The instructions are applicable to be loaded and executed by a processor to implement the steps of the method for predicting the association between ubiquitinase and substrate as described in any one of claims 1-4.

Images