RNA-protein interaction prediction method, apparatus, medium, and electronic device

Abstract

The present disclosure provides an RNA-protein interaction prediction method, device, medium and electronic equipment; it relates to artificial intelligence technical field.The method comprises: obtaining an RNA sequence and a protein sequence to be predicted; encoding the RNA sequence to be predicted to obtain an RNA vector sequence; encoding the protein sequence to be predicted to obtain a protein vector sequence; constructing a matching feature matrix according to the RNA vector sequence and the protein vector sequence; extracting features from the matching feature matrix, and determining the interaction between the RNA sequence to be predicted and the protein sequence to be predicted according to the extracted matching features.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and more specifically, to an RNA-protein interaction prediction method, an RNA-protein interaction prediction device, a computer-readable storage medium, and an electronic device. Background Technology

[0002] Noncoding RNAs (ncRNAs) participate in many complex cellular processes, playing crucial roles in life processes such as alternative splicing, chromatin modification, and epigenetics, and are closely linked to many diseases. Studies have shown that most ncRNAs exert their regulatory functions through interactions with proteins. Therefore, researching the interactions between ncRNAs and proteins is of great significance for elucidating the molecular mechanisms of ncRNA action in human diseases and life activities, and has become one of the important approaches to analyzing the functions of ncRNAs and proteins.

[0003] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] This disclosure provides a method for predicting RNA-protein interactions, an apparatus for predicting RNA-protein interactions, a computer-readable storage medium, and an electronic device.

[0005] This disclosure provides a method for predicting RNA-protein interactions, including:

[0006] Obtain the RNA and protein sequences to be predicted;

[0007] The RNA sequence to be predicted is encoded to obtain an RNA vector sequence;

[0008] The protein sequence to be predicted is encoded to obtain a protein vector sequence;

[0009] Construct a matching feature matrix based on the RNA vector sequence and protein vector sequence;

[0010] Feature extraction is performed on the matching feature matrix, and the interaction between the RNA sequence and protein sequence to be predicted is determined based on the extracted matching features.

[0011] In one exemplary embodiment of this disclosure, encoding the RNA sequence to be predicted to obtain an RNA vector sequence includes:

[0012] The RNA sequence to be predicted is converted into an N-base k-mer subsequence;

[0013] The RNA vector sequence is obtained by vectorizing each base k-mer subsequence.

[0014] In one exemplary embodiment of this disclosure, the vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes:

[0015] Each k-mer subsequence is encoded to obtain a first vector of N k-mer subsequences, and the first vector of the N k-mer subsequences forms the RNA vector sequence.

[0016] In one exemplary embodiment of this disclosure, the vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes:

[0017] Encode each base k-mer subsequence to obtain a first vector of N base k-mer subsequences;

[0018] The first vector of the N base k-mer subsequences is sequentially input into a pre-trained recurrent neural network, which outputs N base k-mer vectors, and the N base k-mer vectors form the RNA vector sequence.

[0019] In one exemplary embodiment of this disclosure, the vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes:

[0020] Encode each base k-mer subsequence to obtain a first vector of N base k-mer subsequences;

[0021] The first vector of the N base k-mer subsequences is processed using the first mapping matrix to obtain the second vector of the N base k-mer subsequences, and the second vector of the N base k-mer subsequences is used to form the RNA vector sequence.

[0022] In one exemplary embodiment of this disclosure, the vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes:

[0023] Encode each base k-mer subsequence to obtain a first vector of N base k-mer subsequences;

[0024] The first vector of the N base k-mer subsequences is obtained by performing operations on the first vector of the N base k-mer subsequences using the first mapping matrix;

[0025] The second vectors of the N base k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs N base k-mer vectors, and the N base k-mer vectors form the RNA vector sequence.

[0026] In one exemplary embodiment of this disclosure, encoding the protein sequence to be predicted to obtain a protein vector sequence includes:

[0027] The protein sequence to be predicted is converted into an M-amino acid k-mer subsequence;

[0028] The protein vector sequence is obtained by vectorizing each amino acid k-mer subsequence.

[0029] In one exemplary embodiment of this disclosure, the vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes:

[0030] Each amino acid k-mer subsequence is encoded to obtain a first vector of M amino acid k-mer subsequences, and the first vector of the M amino acid k-mer subsequences is used to form the protein vector sequence.

[0031] In one exemplary embodiment of this disclosure, the vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes:

[0032] Encode each amino acid k-mer subsequence to obtain a first vector of M amino acid k-mer subsequences;

[0033] The first vectors of the M amino acid k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs M amino acid k-mer vectors, and the protein vector sequence is composed of the M amino acid k-mer vectors.

[0034] In one exemplary embodiment of this disclosure, the vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes:

[0035] Encode each amino acid k-mer subsequence to obtain a first vector of M amino acid k-mer subsequences;

[0036] The first vector of the M amino acid k-mer subsequences is processed using the second mapping matrix to obtain the second vector of the M amino acid k-mer subsequences, and the second vector of the M amino acid k-mer subsequences is used to form the protein vector sequence.

[0037] In one exemplary embodiment of this disclosure, the vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes:

[0038] Encode each amino acid k-mer subsequence to obtain a first vector of M amino acid k-mer subsequences;

[0039] The second vector of the M amino acid k-mer subsequences is obtained by operating on the first vector of the M amino acid k-mer subsequences using the second mapping matrix;

[0040] The second vectors of the M amino acid k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs M amino acid k-mer vectors, and the protein vector sequence is composed of the M amino acid k-mer vectors.

[0041] In one exemplary embodiment of this disclosure, constructing a matching feature matrix based on the RNA vector sequence and the protein vector sequence includes:

[0042] Calculate the matching degree between the base k-mer vector in the RNA vector sequence and the k-mer vector of each amino acid in the protein vector sequence;

[0043] The calculated matching score is used as an element of the matching feature matrix to construct the matching feature matrix.

[0044] In one exemplary embodiment of this disclosure, calculating the matching degree between the base k-mer vector in the RNA vector sequence and the amino acid k-mer vector in the protein vector sequence includes:

[0045] according to:

[0046]

[0047] Calculate the k-mer vector of the i-th base in the RNA vector sequence. The k-mer vector of the j-th amino acid in the protein vector sequence Matching degree between in, express Length, express The length.

[0048] In one exemplary embodiment of this disclosure, constructing a matching feature matrix based on the RNA vector sequence and the protein vector sequence includes:

[0049] The matching feature matrix is ​​obtained by performing a dot product operation between the RNA vector sequence and the protein vector sequence.

[0050] In one exemplary embodiment of this disclosure, the feature extraction of the matching feature matrix includes:

[0051] The matching features are obtained by using a feature extraction network to extract features from the matching feature matrix.

[0052] In one exemplary embodiment of this disclosure, the step of using a feature extraction network to extract features from the matching feature matrix to obtain the matching features includes:

[0053] The matching feature matrix is ​​used to extract features to obtain the original features;

[0054] The original features are processed using the third mapping matrix to obtain the matching features.

[0055] In one exemplary embodiment of this disclosure, determining the interaction between the RNA sequence and the protein sequence to be predicted based on the extracted matching features includes:

[0056] The predicted interaction values ​​between the RNA sequence and the protein sequence to be predicted are obtained based on the matching features.

[0057] The interaction between the RNA sequence and the protein sequence to be predicted is determined based on the interaction prediction value.

[0058] In one exemplary embodiment of this disclosure, obtaining the predicted interaction value between the RNA sequence and the protein sequence to be predicted based on the matching features includes:

[0059] The matching features are input into the classifier, which outputs the probability that there is an interaction between the RNA sequence and the protein sequence to be predicted.

[0060] In one exemplary embodiment of this disclosure, the probability that there is an interaction between the RNA sequence and the protein sequence to be predicted is:

[0061]

[0062] Where r represents the RNA sequence to be predicted, p represents the protein sequence to be predicted, c0 represents the first feature value in the matching features, and c1 represents the second feature value in the matching features.

[0063] In one exemplary embodiment of this disclosure, determining the interaction between the RNA sequence and the protein sequence to be predicted based on the interaction prediction value includes:

[0064] If the predicted interaction value meets a preset threshold condition, it is determined that there is an interaction between the RNA sequence and the protein sequence to be predicted.

[0065] In one exemplary embodiment of this disclosure, the method further includes:

[0066] The recurrent neural network and the feature extraction network are trained.

[0067] In one exemplary embodiment of this disclosure, training the recurrent neural network and the feature extraction network includes:

[0068] Obtain a training dataset, which includes positive RNA-protein pairs and negative RNA-protein pairs;

[0069] The recurrent neural network and feature extraction network are used to determine the predicted interaction values ​​for each RNA-protein pair in the training dataset;

[0070] The loss function is used to calculate the predicted interaction value and label value of each RNA-protein pair in the training dataset to obtain the corresponding loss value;

[0071] The model parameters of the recurrent neural network and the feature extraction network are adjusted based on the loss value.

[0072] In one exemplary embodiment of this disclosure, the loss function is:

[0073]

[0074] Where, r i p represents the i-th RNA sequence in the training dataset. i Let y represent the i-th protein sequence in the training dataset. i p(1|r) represents the label value of the i-th RNA-protein pair in the training dataset. i ,p i ) represents the predicted value of the i-th RNA-protein pair in the training dataset that there is an interaction, p(0|r i ,p i ) represents the predicted value that the i-th RNA-protein pair in the training dataset does not interact, and K is the total number of RNA-protein pairs in the training dataset.

[0075] In one exemplary embodiment of this disclosure, adjusting the model parameters of the recurrent neural network and the feature extraction network based on the loss value includes:

[0076] Based on the loss value, the model parameters of the recurrent neural network and the feature extraction network are iteratively updated using the stochastic gradient descent algorithm. When the iteration termination condition is met, the training of all model parameters is completed.

[0077] In one exemplary embodiment of this disclosure, the method further includes:

[0078] Output the predicted results of the interaction between the RNA sequence and the protein sequence to be predicted.

[0079] This disclosure provides an RNA-protein interaction prediction device, comprising:

[0080] The data acquisition module is used to acquire the RNA and protein sequences to be predicted.

[0081] The first data encoding module is used to encode the RNA sequence to be predicted to obtain an RNA vector sequence;

[0082] The second data encoding module is used to encode the protein sequence to be predicted to obtain a protein vector sequence.

[0083] A feature matrix construction module is used to construct a matching feature matrix based on the RNA vector sequence and the protein vector sequence;

[0084] An interaction prediction module is used to extract features from the matching feature matrix and determine the interaction between the RNA sequence and protein sequence to be predicted based on the extracted matching features.

[0085] This disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any one of the above descriptions.

[0086] This disclosure provides an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method described in any of the preceding methods by executing the executable instructions.

[0087] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0088] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0089] Figure 1 A schematic diagram of an exemplary system architecture for predicting RNA-protein interactions, to which embodiments of the present disclosure can be applied, is shown;

[0090] Figure 2 A flowchart illustrating an RNA-protein interaction prediction method according to an embodiment of the present disclosure is shown schematically.

[0091] Figure 3 A flowchart illustrating the process of obtaining an RNA vector sequence according to an embodiment of the present disclosure is shown schematically.

[0092] Figure 4 A flowchart illustrating the process of obtaining a protein vector sequence according to an embodiment of the present disclosure is shown schematically.

[0093] Figure 5 A flowchart illustrating the construction of a matching feature matrix according to an embodiment of the present disclosure is shown schematically;

[0094] Figure 6 A flowchart illustrating model training according to an embodiment of the present disclosure is shown schematically;

[0095] Figure 7 A flowchart illustrating an RNA-protein interaction prediction method according to another embodiment of the present disclosure is shown schematically;

[0096] Figure 8 A block diagram of an RNA-protein interaction prediction device according to an embodiment of the present disclosure is shown schematically;

[0097] Figure 9 A schematic diagram of the structure of a computer system suitable for implementing the embodiments of the present disclosure is shown. Detailed Implementation

[0098] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0099] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0100] Figure 1 A schematic diagram of a system architecture for an exemplary application environment in which an RNA-protein interaction prediction method and apparatus according to embodiments of the present disclosure can be applied is shown.

[0101] like Figure 1 As shown, the system architecture 100 of the interaction prediction system may include one or more of terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables. Terminal devices 101, 102, and 103 may be various electronic devices, including but not limited to desktop computers, laptops, smartphones, and tablets. It should be understood that... Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be used. For example, server 105 can be a single server, a server cluster consisting of multiple servers, a cloud computing platform, or a virtualization center. Specifically, server 105 can be used to perform the following: acquiring the RNA sequence and protein sequence to be predicted; encoding the RNA sequence to be predicted to obtain an RNA vector sequence; encoding the protein sequence to be predicted to obtain a protein vector sequence; constructing a matching feature matrix based on the RNA vector sequence and the protein vector sequence; extracting features from the matching feature matrix; and determining the interaction between the RNA sequence and the protein sequence to be predicted based on the extracted matching features.

[0102] The RNA-protein interaction prediction method provided in this embodiment is generally executed by server 105. Correspondingly, the RNA-protein interaction prediction device is generally set in server 105. The server can send the prediction results of the interaction between the RNA sequence and the protein sequence to be predicted to the terminal device, and the terminal device can display them to the user. However, it is readily understood by those skilled in the art that the RNA-protein interaction prediction method provided in this embodiment can also be executed by one or more of terminal devices 101, 102, and 103. Correspondingly, the RNA-protein interaction prediction device can also be set in terminal devices 101, 102, and 103. For example, after execution by the terminal device, the prediction results can be directly displayed on the terminal device's screen, or the prediction results can be provided to the user through voice broadcast. This exemplary embodiment does not impose any special limitations on this.

[0103] The technical solutions of the embodiments of this disclosure are described in detail below:

[0104] This exemplary embodiment provides a method for predicting RNA-protein interactions. This method can be applied to the server 105 described above, or to one or more of the terminal devices 101, 102, and 103 described above; no special limitation is made in this exemplary embodiment. (See reference...) Figure 2 As shown, the RNA-protein interaction prediction method may include the following steps S210 to S250:

[0105] Step S210. Obtain the RNA sequence and protein sequence to be predicted;

[0106] Step S220. Encode the RNA sequence to be predicted to obtain an RNA vector sequence;

[0107] Step S230. Encode the protein sequence to be predicted to obtain a protein vector sequence;

[0108] Step S240. Construct a matching feature matrix based on the RNA vector sequence and the protein vector sequence;

[0109] Step S250. Extract features from the matching feature matrix and determine the interaction between the RNA sequence and protein sequence to be predicted based on the extracted matching features.

[0110] In the RNA-protein interaction prediction method provided in the exemplary embodiments of this disclosure, the following steps are taken: First, obtain the RNA sequence and protein sequence to be predicted. Then, encode the RNA sequence to be predicted to obtain an RNA vector sequence. Next, encode the protein sequence to be predicted to obtain a protein vector sequence. Finally, construct a matching feature matrix based on the RNA and protein vector sequences. Extract features from the matching feature matrix and determine the interaction between the RNA and protein sequences based on the extracted matching features. This disclosure improves the accuracy of RNA-protein interaction prediction by constructing a matching feature matrix using the matching relationship between RNA and protein sequences and predicting the interaction between them based on the matching feature matrix.

[0111] The steps described above in this example implementation will now be explained in more detail.

[0112] In step S210, the RNA sequence and protein sequence to be predicted are obtained.

[0113] In this example embodiment, at least one RNA-protein pair to be predicted, consisting of an RNA sequence and a protein sequence, can be obtained, and the interaction between the RNA and protein sequences in each RNA-protein pair is unknown. For example, a user can input the RNA-protein pair to be predicted via a terminal device. For instance, the user can input the RNA-protein pair manually or via voice; this example does not specifically limit this. For instance, an RNA sequence can be input, followed by a protein sequence; the input order is not limited. For instance, the RNA and protein sequences can be input into different text boxes or into the same text box. For instance, after inputting the information, clicking the "Start Prediction" button will begin executing the prediction steps provided in some embodiments of this application.

[0114] The interaction between RNA and protein refers to the function of a protein manifested through its interactions with other proteins and RNA. For example, the interaction between protein and RNA plays a crucial role in protein synthesis. Simultaneously, the performance of many RNA functions depends on its interaction with proteins. Interactions can be regulatory, guiding, etc., and are not limited here. For instance, in the presence of an interaction, RNA can guide protein synthesis, or RNA can regulate protein function. The interaction between RNA and protein can also refer to their ability to regulate each other's life cycle and function through physical interactions. For example, RNA-coding sequences can guide protein synthesis, and correspondingly, proteins can regulate RNA expression and function.

[0115] After obtaining the RNA-protein pairs to be predicted, an interaction prediction system can be used to predict the interactions of each input RNA-protein pair and determine whether each pair interacts based on the prediction results. Simultaneously, the prediction results of the RNA-protein pair interactions can be output to a terminal device for user viewing. For example, the prediction results can be displayed directly on the terminal device's screen, or they can be provided to the user via voice announcement; this example does not specify any particular limitation.

[0116] In other examples, at least one RNA sequence to be predicted can be obtained, and protein sequences that interact with each input RNA sequence can be searched in a database. For instance, after a user inputs the RNA sequence to be predicted via a terminal device, they can select at least one protein sequence from the database. Multiple RNA-protein pairs are formed by the RNA sequence to be predicted and the various protein sequences. The interaction prediction system can then predict the interaction of each RNA-protein pair and output the protein sequences that can interact with the RNA sequence to be predicted based on the prediction results. Preferably, several types of protein sequences can be pre-stored in the database for easy retrieval when predicting RNA-protein pair interactions. For example, protein sequences can be stored in a Redis database or a MySQL database, allowing for real-time querying and selection of the protein sequence to be predicted. Redis is a key-value storage system; when stored in a Redis database, it can include key-value pairs formed by sequence identifiers (such as sequence numbers) and corresponding protein sequences, where the key is the sequence identifier and the value is the corresponding protein sequence. Redis, as a high-efficiency caching technology, supports read / write frequencies exceeding 100KB / s, offering advantages in data reading and storage speed. MySQL, on the other hand, is a relational database management system. Relational databases store data across different tables, rather than storing all data uniformly, increasing storage speed and flexibility. They also offer stability in data storage and can prevent data loss.

[0117] It is understandable that several types of RNA sequences can be pre-stored in a database for retrieval when predicting RNA-protein pair interactions. Therefore, at least one protein sequence to be predicted can be obtained, and RNA sequences that interact with each input protein sequence can be searched in the database. Similarly, after a user inputs a protein sequence via a terminal device, they can select at least one RNA sequence from the database, forming multiple RNA-protein pairs with the protein sequence to be predicted and various RNA sequences. The interaction prediction system can then predict the interaction of each RNA-protein pair and output the RNA sequences that can interact with the protein sequence to be predicted based on the prediction results. This disclosure does not specifically limit this approach.

[0118] In step S220, the RNA sequence to be predicted is encoded to obtain an RNA vector sequence.

[0119] After obtaining the RNA sequence to be predicted, it can be encoded to obtain an RNA vector sequence. A matching feature matrix can then be constructed based on the RNA vector sequence and the protein vector sequence, thereby predicting the interaction between the RNA vector sequence and the protein vector sequence based on the matching feature matrix.

[0120] In the exemplary embodiments of this disclosure, RNA sequences can be represented by base sequences. For example, an RNA sequence can be represented as AGCAAGCACCU… An RNA sequence can include four bases: adenine (A), uracil (U), guanine (G), and cytosine (C). Correspondingly, RNA sequences can also be represented using base k-mer subsequences. A k-mer subsequence refers to a k-connector consisting of k bases or k amino acids as a group. Specifically, all base k-mer subsequences can be obtained by arranging and combining the four bases; for a given k value, four k-mer subsequences can be obtained. k There are 4 base k-mer subsequences. For example, when k is 3, there are 4 3 =64 base 3-mer subsequences, when k is 4, there are a total of 4 4= 256 possible 4-mer subsequences. For example, AGC, AUA, GCA, and CCU are four different 3-mer subsequences, and AGCA, UAGC, and ACCU are three different 4-mer subsequences. Therefore, the RNA sequence AGCAUAGCACCU… can also be represented as {AGC, AUA, GCA, CCU, …}, or as {AGCA, UAGC, ACCU, …}. In other examples, the corresponding 3-mer or 4-mer subsequences can be obtained by reading the RNA sequence in an overlapping manner. Correspondingly, the 3-mer subsequences of the RNA sequence can also include AGC, GCA, CAU, AUA, etc., and the 4-mer subsequences of the RNA sequence can also include AGCA, GCAU, CAUA, etc., which are not specifically limited in this disclosure. In the example embodiments of this disclosure, k is a positive integer, such as 1, 2, 3…, and k can take one or more values. The specific value of k can be adjusted according to the actual situation and is not limited here.

[0121] When encoding the RNA sequence to be predicted, it can be converted into N base k-mer subsequences. For example, depending on the value of k, starting from the first base of the RNA sequence to be predicted, k consecutive bases can be taken to form a single k-mer subsequence, continuing until the last k bases are taken, resulting in all the k-mer subsequences of the RNA sequence to be predicted. Then, each k-mer subsequence can be vectorized to obtain N k-mer vectors, which together form an RNA vector sequence. For example, the RNA sequence to be predicted can be divided into N non-overlapping k-mer subsequences. For instance, if the RNA sequence to be predicted is AUCUGAAAU, it can be divided into three k-mer subsequences: AUC, UGA, and AAU. It is understood that dividing the RNA sequence into multiple non-overlapping k-mer subsequences is to vectorize the bases in the RNA sequence in a k-connected manner. Similarly, in other examples, each base in the RNA sequence to be predicted can be vectorized to obtain multiple base vectors, which can then form an RNA vector sequence. Alternatively, the overlapping parts of the RNA sequence to be predicted can be divided into P k-mer subsequences, each k-mer subsequence can be vectorized to obtain P k-mer vectors, which can then form an RNA vector sequence. This disclosure does not specifically limit the scope of this method.

[0122] In one example implementation, after converting the RNA sequence to be predicted into N base k-mer subsequences, each base k-mer subsequence in the RNA sequence to be predicted can be encoded to obtain a first vector of N base k-mer subsequences. This first vector of N base k-mer subsequences then forms the RNA vector sequence. For example, when k=3, there can be 64 possible 3-mer subsequences. Each 3-mer subsequence can be One-Hot encoded to obtain the first vector of the base k-mer subsequences. One-Hot encoding, also known as one-bit valid encoding, uses an N-bit state register to encode N states. Each state has an independent register bit, and at any given time, only one bit in the register is valid.

[0123] For example, for the i-th 3-mer subsequence (i.e., the 3-mer subsequence with index i), a 64-dimensional One-Hot vector can be obtained through encoding. The i-th element of this vector is set to 1, and all other elements are set to 0, in the form [0, 1, 0, 0, ..., 0]. Similarly, each 3-mer subsequence can correspond to a 3-mer One-Hot vector. Furthermore, when k = 1, each base is a 1-mer subsequence, meaning each base in the RNA sequence to be predicted can be encoded to obtain a representation vector for each base. For instance, if the RNA sequence to be predicted contains L bases, for the j-th base (i.e., the base with index j), an L-dimensional One-Hot vector can be obtained through encoding. The j-th element of this vector is set to 1, and all other elements are set to 0, thus obtaining the One-Hot vector for the j-th base. In other examples, each base in the RNA sequence to be predicted can be encoded as a 4-dimensional One-Hot vector based on the base type. For example, base A can be represented by the One-Hot vector [1, 0, 0, 0], U by [0, 0, 0, 1], G by [0, 1, 0, 0], and C by [0, 0, 1, 0]. Correspondingly, a One-Hot vector for each base in the RNA sequence to be predicted can be obtained.

[0124] For example, the RNA sequence AUCUGAAAU to be predicted may include three 3-mer subsequences: AUC, UGA, and AAU. The corresponding three 3-mer One-Hot vectors are as follows: and An RNA vector sequence can be composed of three 3-mer One-Hot bases. In the exemplary embodiments of this disclosure, by performing One-Hot encoding on the base k-mer subsequences, each base k-mer subsequence can be transformed into a binary feature, thereby compensating for the deficiencies of classifiers in processing attribute data, so that the classifier can more accurately predict the interaction between RNA sequences and protein sequences.

[0125] In some embodiments of this disclosure, dense vectors can be used to represent each 3-mer subsequence. This involves sequentially embedding (vector mapping) encoding each 3-mer subsequence, representing each 3-mer subsequence with a low-dimensional vector, resulting in multiple corresponding 3-mer embedding vectors. These multiple 3-mer embedding vectors then form an RNA vector sequence. For example, the Word2vec algorithm can be used to map each 3-mer subsequence in the RNA sequence to a vector space, where each 3-mer subsequence can be represented by a vector. Algorithms such as Doc2vec and Glove can also be used to convert the 3-mer subsequence into embedding vectors. Alternatively, a BERT (Bidirectional Encoding Representation from Transformer) pre-trained model can be used to encode each 3-mer subsequence, obtaining multiple corresponding 3-mer embedding vectors. This disclosure does not specifically limit the specific implementation of these methods. In the exemplary embodiments of this disclosure, by embedding encoding the base k-mer subsequences, discrete base k-mer subsequences can be converted into a low-dimensional continuous vector. This continuous vector can better represent each base k-mer subsequence. Moreover, the embedding encoding process is learnable; through continuous training, similar base k-mer subsequences can become closer in the vector space, achieving class distinction while encoding the base k-mer subsequences. This facilitates more accurate prediction of interactions between RNA and protein sequences. Furthermore, it also improves the efficiency of interaction prediction to some extent.

[0126] In one example implementation, after converting the RNA sequence to be predicted into N base k-mer subsequences, each base k-mer subsequence can be encoded to obtain a first vector of N base k-mer subsequences. The first vectors of the N base k-mer subsequences are then sequentially input into a pre-trained recurrent neural network, which outputs N base k-mer vectors, forming an RNA vector sequence. The recurrent neural network can include LSTM (Long Short-Term Memory) networks, bidirectional recurrent neural networks, and GRU (Gated Recurrent Unit) networks, etc.

[0127] For example, the first vector can be a One-Hot vector. It's understandable that there are relationships between the bases in an RNA sequence. In this example, all the 3-mer One-Hot vectors in the RNA sequence to be predicted can be considered as a temporal sequence, and then a recurrent neural network can be used to operate on each 3-mer One-Hot vector. For example, to obtain all the 3-mer One-Hot vectors in the RNA sequence AUCUGAAAU to be predicted (… and After that, the three 3-mer One-Hot vectors can be input into the trained LSTM network, which outputs the corresponding 3-mer vectors for each base, respectively. and The RNA vector sequence consists of three 3-mer bases. Among them, the LSTM network is a time recurrent neural network, which is suitable for processing and predicting important events with relatively long intervals and delays in time series.

[0128] In one example implementation, after converting the RNA sequence to be predicted into N base k-mer subsequences, each base k-mer subsequence can be encoded to obtain a first vector of N base k-mer subsequences. A first mapping matrix is ​​then used to operate on the first vector of the N base k-mer subsequences to obtain a second vector of N base k-mer subsequences, and the second vectors of the N base k-mer subsequences form an RNA vector sequence.

[0129] For example, the first vector can be a One-Hot vector, and the second vector can be an Embedding vector. For the RNA sequence AUCUGAAAU to be predicted, it can include three 3-mer subsequences: AUC, UGA, and AAU. Each 3-mer subsequence can be One-Hot encoded to obtain a 3-mer One-Hot vector, which is respectively... and Since the 3-mer One-Hot vector is a 64-dimensional sparse vector, it can be mapped to a dense embedding vector using the first mapping matrix W1, i.e., according to:

[0130]

[0131] Obtain the 3-mer embedding vector of the i-th base in the RNA sequence to be predicted. Among them, V i R This represents the 3-mer One-Hot vector of the i-th base in the RNA sequence to be predicted. The first mapping matrix W1 is an A*64 parameter matrix; for example, A can be 128 or 256, and this disclosure does not specifically limit it. Based on this, the 3-mer Embedding vectors corresponding to the three 3-mer subsequences can be obtained sequentially, respectively. and Furthermore, an RNA vector sequence can be formed from three-base 3-mer embedding vectors.

[0132] In one example implementation, after converting the RNA sequence to be predicted into an N-base k-mer subsequence, reference is made... Figure 3 As shown, each base k-mer subsequence can be encoded according to steps S310 to S330 to obtain an RNA vector sequence.

[0133] Step S310. Encode each base k-mer subsequence to obtain the first vector of N base k-mer subsequences.

[0134] For example, the first vector can be a One-Hot vector. For the RNA sequence AUCUGAAAU to be predicted, it can include three 3-mer subsequences: AUC, UGA, and AAU. Each 3-mer subsequence can be One-Hot encoded to obtain a 3-mer One-Hot vector, which is respectively... and

[0135] Step S320. Calculate the first vector of the N base k-mer subsequences using the first mapping matrix to obtain the second vector of the N base k-mer subsequences.

[0136] The second vector can be an embedding vector. Since the 3-mer One-Hot vector is a 64-dimensional sparse vector, it can be mapped to a dense embedding vector using the first mapping matrix W1, resulting in three 3-mer embedding vectors, namely... and

[0137] Step S330. The second vectors of the N base k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs N base k-mer vectors, and the N base k-mer vectors form the RNA vector sequence.

[0138] It is understandable that there are relationships between the bases in an RNA sequence. In this example, the 3-mer embedding vectors of all bases in the RNA sequence to be predicted can be regarded as a temporal sequence, and then a recurrent neural network can be used to operate on each 3-mer embedding vector. For example, the 3-mer embedding vectors of all bases in the RNA sequence AUCUGAAAU to be predicted can be obtained ( and After that, the three base 3-mer embedding vectors can be sequentially input into the trained LSTM network, which outputs the corresponding 3-mer vector for each base, as follows: and The RNA vector sequence consists of three 3-mer bases.

[0139] Specifically, we can first embed the vector corresponding to "AUC". Input into an LSTM network allows for the processing of data through the LSTM network. Extract the latent features and output the latent vector at time t. Then, the hidden vector at time t can be... The Embedding vector corresponding to "UGA" at time t+1 The concatenated vectors are then fed into an LSTM network, where their latent features are extracted, and the latent vector at time t+1 is output. Similarly, the current embedding vector can be concatenated with the hidden vector passed down from the previous time step, and features can be extracted from the concatenated vector using an LSTM network. Finally, the embedding vector corresponding to "AAU" can be... Input the hidden vector at time t+1 into the LSTM network. With Embedding vector The concatenated vectors are then processed, and their latent features are extracted using an LSTM network to output the latent vector at the final time step. In other examples, a GRU network can be used to compute the 3-mer embedding vector for each base. The GRU network has a relatively simple structure and achieves the same results as the LSTM network. Alternatively, each 3-mer One-Hot vector from the RNA sequence to be predicted can be directly input into the GRU network to obtain the corresponding 3-mer vector. This disclosure does not impose any specific limitations on this approach.

[0140] In this implementation, when using an LSTM network to process multiple 3-mer embedding vectors in the RNA sequence to be predicted, the dependencies between the various 3-mer embedding vectors can be learned and memorized to obtain the final RNA vector sequence. When constructing a matching feature matrix using this RNA vector sequence, the matching relationship between the RNA and protein sequences can be more accurately represented, thereby improving the accuracy of RNA-protein interaction prediction.

[0141] In step S230, the protein sequence to be predicted is encoded to obtain a protein vector sequence.

[0142] After obtaining the protein sequence to be predicted, the protein sequence can be encoded to obtain a protein vector sequence. A matching feature matrix can then be constructed based on the protein vector sequence and the RNA vector sequence, thereby predicting the interaction between the RNA vector sequence and the protein vector sequence based on the matching feature matrix.

[0143] In one example implementation, a protein sequence can be represented by an amino acid sequence. This can include 20 amino acids, which are sequentially encoded as A, G, V, I, L, F, P, Y, M, T, S, H, N, Q, W, R, K, D, E, C. For example, a protein sequence can be represented as MTAQDDSYS… Correspondingly, a protein sequence can also be represented using amino acid k-mer subsequences. Specifically, all amino acid k-mer subsequences can be obtained by permuting and combining the 20 amino acids; for a given k value, 20 k-mer subsequences can be obtained. k A set of 20 amino acid k-mer subsequences. For example, when k is 3, there are 20 k-mer subsequences. 3=8000 amino acid 3-mer subsequences. For example, MTA, QDD, and SYS are three different amino acid 3-mer subsequences. Therefore, the protein sequence MTAQDDSYS… can also be represented as {MTA, QDD, SYS, …}. In other examples, the corresponding amino acid 3-mer subsequences can also be obtained by reading the protein sequence in an overlapping manner. Correspondingly, the amino acid 3-mer subsequences of this protein sequence can also include MTA, TAQ, AQD, etc. Furthermore, according to the physicochemical properties of amino acids, the 20 amino acids can be divided into 7 categories: {A, G, V}, {I, L, F, P}, {Y, M, T, S}, {H, N, Q, W}, {R, K}, {D, E}, and {C}, and each category of amino acids can be re-encoded, such as sequentially encoding them as 1, 2, 3, 4, 5, 6, and 7. For example, the protein sequence MTAQDDSYS… can be converted to 331466333…. Then, the seven classes of amino acids can be arranged and combined to obtain all the amino acid k-mer subsequences. For a certain k value, seven... k The k-mer subsequence of the 20 amino acids is not specifically limited in this disclosure. It is understood that dividing the 20 amino acids into 7 categories is merely illustrative; the 20 amino acids can also be classified according to their composition. Similarly, the four bases of the RNA sequence can also be classified according to actual needs.

[0144] When encoding a protein sequence to be predicted, the sequence can be converted into M amino acid k-mer subsequences. For example, based on the value of k, starting from the first amino acid, k consecutive amino acids can be taken to form an amino acid k-mer subsequence, continuing until the last k amino acids are taken, resulting in all amino acid k-mer subsequences of the protein sequence. Each amino acid k-mer subsequence can then be vectorized to obtain M amino acid k-mer vectors, which together form a protein vector sequence. For instance, the protein sequence to be predicted can be divided into M non-overlapping amino acid k-mer subsequences. For example, if the protein sequence to be predicted is MTAQDDSYS, it can be divided into three amino acid k-mer subsequences: MTA, QDD, and SYS. Similarly, in other examples, each amino acid in the protein sequence to be predicted can be vectorized to obtain multiple amino acid vectors, which together form a protein vector sequence. Alternatively, the overlapping protein sequences to be predicted can be divided into Q amino acid k-mer subsequences, each amino acid k-mer subsequence can be vectorized to obtain Q amino acid k-mer vectors, and the protein vector sequence can be composed of the Q amino acid k-mer vectors. This disclosure does not make any specific limitations on this.

[0145] In one example implementation, after converting the protein sequence to be predicted into M amino acid k-mer subsequences, each amino acid k-mer subsequence in the protein sequence to be predicted can be encoded to obtain a first vector of M amino acid k-mer subsequences, and the first vectors of the M amino acid k-mer subsequences can be used to form a protein vector sequence. For example, when k=3, there can be 8000 possible amino acid 3-mer subsequences, and each amino acid 3-mer subsequence can be One-Hot encoded to obtain the first vector of amino acid k-mer subsequences.

[0146] For example, for the j-th amino acid 3-mer subsequence, i.e., the amino acid 3-mer subsequence with an integer index j, an 8000-dimensional One-Hot vector can be obtained through encoding. The j-th element of this vector is set to 1, and all other elements are set to 0, in the form [1, 0, 0, ..., 0]. Similarly, each amino acid 3-mer subsequence can correspond to an amino acid 3-mer One-Hot vector. For another example, when k = 1, each amino acid is a 1-mer subsequence, meaning that each amino acid in the protein sequence to be predicted can be encoded to obtain a representation vector corresponding to each amino acid. For instance, if the protein sequence to be predicted contains S amino acids, for the j-th amino acid, i.e., the amino acid with an integer index j, an S-dimensional One-Hot vector can be obtained through encoding. The j-th element of this vector is set to 1, and all other elements are set to 0, thus obtaining the One-Hot vector for the j-th amino acid. In other examples, each amino acid in the protein sequence to be predicted can be encoded into a 20-dimensional One-Hot vector based on its amino acid type, thus obtaining a One-Hot vector for each amino acid in the protein sequence to be predicted. Alternatively, the 20 amino acids can be classified, and each amino acid in the protein sequence to be predicted can be encoded into a One-Hot vector with the same vector dimension as the number of classification categories. For example, when the 20 amino acids are divided into 7 categories, each amino acid in the protein sequence to be predicted can be encoded into a 7-dimensional One-Hot vector; this disclosure does not impose specific limitations on this approach.

[0147] For example, the protein sequence MTAQDDSYS to be predicted may include three amino acid 3-mer subsequences: MTA, QDD, and SYS. The corresponding three amino acid 3-mer One-Hot vectors are as follows: and An RNA vector sequence can be composed of three amino acid 3-mer One-Hot vectors. In the exemplary embodiments of this disclosure, by performing One-Hot encoding on the amino acid k-mer subsequences, each amino acid k-mer subsequence can be transformed into a binary feature, thereby compensating for the deficiencies of classifiers in processing attribute data, so that the classifier can more accurately predict the interaction between RNA sequences and protein sequences.

[0148] In some embodiments of this disclosure, dense vectors can be used to represent each amino acid 3-mer subsequence. This involves sequentially embedding each amino acid 3-mer subsequence, representing each subsequence with a low-dimensional vector, resulting in multiple corresponding amino acid 3-mer embedding vectors. These multiple embedding vectors then form the protein vector sequence. For example, the Word2vec algorithm can be used to map each amino acid 3-mer subsequence in the protein sequence to a vector space, where each subsequence can be represented by a vector. Algorithms such as Doc2vec and GloVe can also be used to convert the amino acid 3-mer subsequence into embedding vectors. Alternatively, a BERT pre-trained model can be used to encode each subsequence, yielding multiple corresponding embedding vectors. This disclosure does not specifically limit the specific implementation of these methods. In the example embodiments of this disclosure, by embedding the amino acid k-mer subsequences, discrete k-mer subsequences can be converted into low-dimensional continuous vectors, which can better represent each subsequence. Furthermore, the embedding encoding process is learnable. Through continuous training, similar amino acid k-mer subsequences can become closer in the vector space, enabling class distinction while encoding amino acid k-mer subsequences. This allows for more accurate prediction of interactions between RNA and protein sequences. Additionally, it improves the efficiency of interaction prediction to some extent.

[0149] In one example implementation, after converting the protein sequence to be predicted into M amino acid k-mer subsequences, each amino acid k-mer subsequence can be encoded to obtain a first vector of M amino acid k-mer subsequences. The first vectors of the M amino acid k-mer subsequences are then sequentially input into a pre-trained recurrent neural network, which outputs M amino acid k-mer vectors, and the M amino acid k-mer vectors form a protein vector sequence.

[0150] For example, the first vector can be a One-Hot vector. It's understandable that there are relationships between the amino acids in a protein sequence. In this example, all the 3-mer One-Hot vectors of the amino acids in the protein sequence to be predicted can be considered as a temporal sequence, and then a recurrent neural network can be used to operate on each amino acid's 3-mer One-Hot vector. For example, the 3-mer One-Hot vectors of all amino acids in the protein sequence MTAQDDSYS to be predicted can be obtained (…). and After that, the three amino acid 3-mer One-Hot vectors can be sequentially input into the trained LSTM network, which outputs the corresponding 3-mer vector for each amino acid, as follows: and The protein vector sequence is composed of three amino acid 3-mer vectors.

[0151] In one example implementation, after converting the protein sequence to be predicted into M amino acid k-mer subsequences, each amino acid k-mer subsequence can be encoded to obtain a first vector of M amino acid k-mer subsequences. A second mapping matrix is ​​then used to operate on the first vector of the M amino acid k-mer subsequences to obtain a second vector of the M amino acid k-mer subsequences, and the second vectors of the M amino acid k-mer subsequences constitute the protein vector sequence.

[0152] For example, the first vector can be a One-Hot vector, and the second vector can be an Embedding vector. For the protein sequence MTAQDDSYS to be predicted, it can include three amino acid 3-mer subsequences: MTA, QDD, and SYS. Each amino acid 3-mer subsequence can be One-Hot encoded to obtain an amino acid 3-mer One-Hot vector, which is respectively... and Since the amino acid 3-mer One-Hot vector is an 8000-dimensional sparse vector, it can be mapped to a dense embedding vector using the second mapping matrix W2, i.e., according to:

[0153]

[0154] Obtain the 3-mer embedding vector of the j-th amino acid in the protein sequence to be predicted. Among them, V j P Let W1 represent the one-hot vector of the j-th amino acid 3-mer in the protein sequence to be predicted. The second mapping matrix W2 is a B*8000 parameter matrix; for example, B can be 256 or 128, and this disclosure does not specifically limit this. Based on this, the 3-mer embedding vectors corresponding to the three amino acid 3-mer subsequences can be obtained sequentially, respectively... and Furthermore, a protein vector sequence can be composed of three amino acid 3-mer embedding vectors.

[0155] In one example implementation, after converting the protein sequence to be predicted into an M-amino acid k-mer subsequence, reference is made... Figure 4 As shown, each amino acid k-mer subsequence can be encoded according to steps S410 to S430 to obtain a protein vector sequence.

[0156] Step S410. Encode each amino acid k-mer subsequence to obtain the first vector of M amino acid k-mer subsequences.

[0157] For example, the first vector can be a One-Hot vector. For the protein sequence MTAQDDSYS to be predicted, it can include three amino acid 3-mer subsequences: MTA, QDD, and SYS. Each amino acid 3-mer subsequence can be One-Hot encoded to obtain an amino acid 3-mer One-Hot vector, which is respectively... and

[0158] Step S420. Calculate the first vector of the M amino acid k-mer subsequences using the second mapping matrix to obtain the second vector of the M amino acid k-mer subsequences.

[0159] The second vector can be an embedding vector. Since the amino acid 3-mer One-Hot vector is an 8000-dimensional sparse vector, it can be mapped to a dense embedding vector using the second mapping matrix W2, resulting in three amino acid 3-mer embedding vectors, respectively. and

[0160] Step S430. The second vectors of the M amino acid k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs M amino acid k-mer vectors, and the protein vector sequence is composed of the M amino acid k-mer vectors.

[0161] In this example, the 3-mer embedding vectors of all amino acids in the protein sequence to be predicted can be viewed as a temporal sequence, and then a recurrent neural network can be used to operate on each amino acid 3-mer embedding vector. For example, the 3-mer embedding vectors of all amino acids in the protein sequence MTAQDDSYS to be predicted can be obtained ( and After that, the three amino acid 3-mer embedding vectors can be sequentially input into the trained LSTM network, which outputs the corresponding 3-mer vector for each amino acid, as follows: and The protein vector sequence is composed of three amino acid 3-mer vectors.

[0162] Specifically, you can first embed the vector corresponding to "MTA". Input into an LSTM network allows for the processing of data through the LSTM network. Extract the latent features and output the latent vector at time t. Then, the hidden vector at time t can be... The Embedding vector corresponding to "QDD" at time t+1 The concatenated vectors are then fed into an LSTM network, where their latent features are extracted, and the latent vector at time t+1 is output. Finally, the embedding vector corresponding to "SYS" can be... Input the hidden vector at time t+1 into the LSTM network. With Embedding vector The concatenated vectors are then processed, and their latent features are extracted using an LSTM network to output the latent vector at the final time step. In other examples, the GRU network can also be used to calculate the 3-mer embedding vector for each amino acid. Alternatively, the 3-mer one-hot vector of each amino acid in the protein sequence to be predicted can be directly input into the GRU network to obtain the corresponding base 3-mer vector; this disclosure does not impose specific limitations on this method.

[0163] In this implementation, when using an LSTM network to process multiple amino acid 3-mer embedding vectors in the protein sequence to be predicted, the dependencies between the various amino acid 3-mer embedding vectors can be learned and memorized to obtain the final protein vector sequence. When constructing a matching feature matrix using this protein vector sequence, the matching relationship between the protein sequence and the RNA sequence can be more accurately reflected, thereby improving the accuracy of RNA-protein interaction prediction.

[0164] In step S240, a matching feature matrix is ​​constructed based on the RNA vector sequence and the protein vector sequence.

[0165] In the exemplary embodiments of this disclosure, a matching feature matrix can be constructed using RNA vector sequences and protein vector sequences, and the interaction between the input RNA sequence and protein sequence can be predicted using the matching feature matrix, thereby improving the accuracy of RNA-protein interaction prediction.

[0166] For example, refer to Figure 5 As shown, a matching feature matrix can be constructed using RNA vector sequences and protein vector sequences according to steps S510 and S520.

[0167] Step S510. Calculate the matching degree between the base k-mer vector in the RNA vector sequence and the amino acid k-mer vector in the protein vector sequence.

[0168] When the RNA sequence to be predicted is converted into an N-base k-mer subsequence, the corresponding RNA vector sequence can be: The RNA vector sequence includes an N-base k-mer vector. Similarly, when the protein sequence to be predicted is converted into an M-amino acid k-mer subsequence, the corresponding protein vector sequence can be... The protein vector sequence includes M amino acid k-mer vectors. It is understood that the distance between two vectors is negatively correlated with their matching degree. Correspondingly, in the example embodiment of this disclosure, if the base k-mer vector... With amino acid k-mer vector The greater the distance, the lower the matching degree between the i-th base k-mer subsequence and the j-th amino acid k-mer subsequence. If the base k-mer vector... With amino acid k-mer vector The smaller the distance, the higher the matching degree between the i-th base k-mer subsequence and the j-th amino acid k-mer subsequence.

[0169] In one example, it can be based on:

[0170]

[0171] Calculate the k-mer vector of the i-th base in the RNA vector sequence. The k-mer vector of the j-th amino acid in the protein vector sequence The distance magnitude is then obtained. and Matching degree between in, express Length, express The length, ‖·‖2 represents the L2 norm. Based on this, the matching degrees between N base k-mer vectors and M amino acid k-mer vectors can be calculated in sequence to obtain N*M matching scores. It is also possible to calculate the matching degrees between some base k-mer vectors and some amino acid k-mer vectors. For example, calculate the matching degrees between X (N / 2 < X ≤ N) base k-mer vectors and Y (M / 2 < Y ≤ M) amino acid k-mer vectors to obtain X*Y matching scores. The present disclosure does not make specific limitations on this. It can be understood that by calculating the matching degrees between most of the base k-mer vectors in the RNA vector sequence and most of the base k-mer vectors in the protein vector sequence, the matching relationship between the RNA vector sequence and the protein vector sequence can also be accurately reflected.

[0172] In some embodiments of the present disclosure, the Euclidean distance, Manhattan distance, Mahalanobis distance, etc. between the base k-mer vectors in the RNA vector sequence and the amino acid k-mer vectors in the protein vector sequence can also be calculated to determine the matching degrees between the base k-mer vectors in the RNA vector sequence and the amino acid k-mer vectors in the protein vector sequence, so as to construct a matching feature matrix. The present disclosure does not make specific limitations on this.

[0173] Step S520. Use the calculated matching scores as the elements of the matching feature matrix to construct the matching feature matrix.

[0174] Exemplarily, when calculating the matching degrees between the base k-mer vectors in the RNA vector sequence and the amino acid k-mer vectors in the protein vector sequence to obtain N*M matching scores, each matching score can be used as an element of the matching feature matrix to construct a matching feature matrix F with a size of N*M. Among them, each element F i,j (i = 1, 2, …, N; j = 1, 2, …, M) in the matching feature matrix F is the i-th element (base k-mer vector ) in the RNA vector sequence and the j-th element (amino acid k-mer vector ) in the protein vector sequence The matching score can represent the matching degree between the i-th base k-mer vector in the RNA vector sequence and the j-th amino acid k-mer vector in the protein vector sequence The matching degree.

[0175] In some embodiments of the present disclosure, the RNA vector sequence and the protein vector sequence can also be subjected to a dot product operation to obtain a matching feature matrix. Exemplarily, the RNA vector sequence With protein vector sequence Perform a dot product operation. For example, the RNA vector sequence can be transposed first, and then the dot product operation can be performed between the transposed RNA vector sequence and the protein vector sequence to obtain a matching feature matrix containing N*M elements, where each element corresponds to the matching score between a base k-mer vector and an amino acid k-mer vector. Alternatively, the protein vector sequence can be transposed and then the dot product operation performed between it and the RNA vector sequence. In some cases, if one of the protein vector sequence and the RNA vector sequence is a row vector and the other is a column vector, the transposition operation can be omitted.

[0176] In the exemplary embodiments of this disclosure, a matching feature matrix is ​​constructed using RNA vector sequences and protein vector sequences. This matching feature matrix can accurately reflect the matching relationship between RNA vector sequences and protein vector sequences. Therefore, when predicting the interaction between input RNA sequences and protein sequences using this matching feature matrix, the accuracy of RNA-protein interaction prediction can be improved.

[0177] In step S250, feature extraction is performed on the matching feature matrix, and the interaction between the RNA sequence and protein sequence to be predicted is determined based on the extracted matching features.

[0178] In this exemplary embodiment, it is necessary to predict the interaction between the RNA and protein sequences to be predicted. The prediction result may indicate that there is an interaction between the RNA and protein sequences, or it may indicate that there is no interaction between them, i.e., a binary classification prediction is performed. A feature extraction network can be used to extract features from the matching feature matrix, and the existence of an interaction between the RNA and protein sequences to be predicted can be determined based on the extracted matching features. The feature extraction network can be a convolutional neural network. For example, a convolutional neural network can be used to extract features from the matching feature matrix F to obtain the matching features.

[0179] In the exemplary embodiments of this disclosure, the network structure of the convolutional neural network may include at least one convolutional layer and at least one pooling layer. After inputting the matching feature matrix F into the convolutional neural network, it can output a feature matrix FM of size D*E. In some embodiments, to facilitate subsequent binary classification prediction, the feature matrix FM can be dimensionality reduced. For example, the feature matrix FM can be flattened into a D×E dimensional feature vector, and a classifier can be used to perform classification prediction on this feature vector to obtain the predicted interaction values ​​between the RNA sequence and the protein sequence to be predicted.

[0180] Since different RNA sequences may have different lengths (i.e., different numbers of bases), and different protein sequences may also have different lengths, to improve processing efficiency, the sequence length used for prediction can be set to 1000, i.e., N = M = 1000, and the corresponding size of the matching feature matrix F is 1000 x 1000. If the sequence length is less than 1000, zero padding can be performed. If the sequence length exceeds 1000, the first 1000 bytes of the sequence can be selected. In the example embodiment of this disclosure, the network structure of the convolutional neural network can be: six 5x5 convolutional layers, i.e., each convolutional layer has a kernel size of 5x5, a 3x3 average pooling layer, and a ReLU (Rectified Linear Unit) activation layer; six 5x5 convolutional layers, a 4x4 average pooling layer, and a ReLU activation layer; six 3x3 convolutional layers, a 4x4 average pooling layer, and a ReLU activation layer. When a matching feature matrix F of size 1000*1000 is input into this convolutional neural network, a feature matrix FM of size 20*20 can be output.

[0181] Furthermore, the 20*20 feature matrix FM can be flattened into a 400-dimensional feature vector v. For example, all elements of the feature matrix FM can be FM... 0,0 FM 0,1 ...FM 19,19 Arranged sequentially into a row or a column, when arranged in a row, the resulting 400-dimensional feature vector is:

[0182] v = [FM] 0,0 ,…,FM 0,19 FM 1,0 ,…,FM 1,19 ,…,FM 19,19 ]

[0183] In some implementations, a convolutional neural network can be used to extract features from the matching feature matrix F to obtain the original features, and then a third mapping matrix can be used to perform operations on the original features to obtain the matching features.

[0184] To perform binary classification prediction, the original features extracted by the convolutional neural network need to be reduced in dimensionality. For example, a fully connected layer can be used to reduce the dimensionality of the original features, resulting in matching features for interaction prediction. Correspondingly, the feature vector can be operated on using a third mapping matrix, i.e., according to:

[0185] c = W6 × v (4)

[0186] The matching feature c is obtained; where c is a 2-dimensional feature vector [c0, c1], the third mapping matrix W6 is a 2*C parameter matrix, v is the feature vector, and the value of C is consistent with the dimension of the feature vector.

[0187] In other examples, after inputting the 1000 x 1000 matching feature matrix F into a convolutional neural network, a specific convolutional neural network can be used to directly output a feature vector, thus obtaining the matching features used for interaction prediction, such as a 2-dimensional feature vector. In this case, the aforementioned dimensionality reduction step can be omitted, and this disclosure does not specifically limit this. In the exemplary embodiments of this disclosure, using a convolutional neural network to extract features from the matching feature matrix and using the extracted invariant matching features for interaction prediction can improve the accuracy of RNA-protein interaction prediction.

[0188] In one example implementation, after obtaining the matching features for interaction prediction, the predicted interaction values ​​between the RNA and protein sequences to be predicted can be obtained based on the matching features, and the interaction between the RNA and protein sequences to be predicted can be determined based on the predicted interaction values. For example, the matching features can be input into a classifier to classify the interaction between the RNA and protein sequences to be predicted based on the matching features. After classification, the predicted interaction value between the RNA and protein sequences to be predicted is output. For example, a Softmax classifier can be used to predict the interaction between the RNA and protein sequences to be predicted. Specifically, the matching features can be transformed using a Softmax classifier to obtain probability distributions for the interaction between the RNA and protein sequences to be predicted, indicating whether the interaction exists or not.

[0189] For example, the probability of an interaction between the RNA sequence and the protein sequence to be predicted can be obtained using the Softmax classifier:

[0190]

[0191] The probability that there is no interaction between the RNA sequence and the protein sequence to be predicted is:

[0192]

[0193] Where r represents the RNA sequence to be predicted, p represents the protein sequence to be predicted, c0 is the first feature value of the matching feature, and c1 is the second feature value of the matching feature. When the matching feature is a 2D vector, the vector is (c0, c1). In other examples, logistic regression classifiers and SVM (Support Vector Machine) classifiers can also be used for binary classification prediction to obtain the predicted interaction value between the RNA sequence and the protein sequence to be predicted based on the matching feature. This disclosure does not specifically limit this.

[0194] After obtaining the predicted interaction values ​​between the RNA and protein sequences to be predicted using a classifier, the interaction between the sequences can be determined based on these predicted values. For example, if the predicted interaction values ​​meet a preset threshold condition, it can be determined that there is an interaction between the RNA and protein sequences to be predicted.

[0195] For example, the Softmax classifier can be used to obtain the probability P(1|r, p) of an interaction between the RNA and protein sequences to be predicted. Here, P(1|r, p) can be any value between 0 and 1. For instance, a preset threshold of 0.5 can be used. When P(1|r, p) > 0.5, the prediction result can be marked as 1, indicating an interaction between the RNA and protein sequences. When P(1|r, p) ≤ 0.5, the prediction result can be marked as 0, indicating no interaction between the RNA and protein sequences. In other examples, it can also be set so that when P(1|r, p) ≥ 0.5, an interaction between the RNA and protein sequences is determined, and when P(1|r, p) < 0.5, no interaction is determined. Finally, the prediction result of the interaction between the RNA and protein sequences can be output to the terminal device for user viewing. It should be noted that the output may only show the probability that there is an interaction between the RNA sequence and the protein sequence to be predicted, or it may only show the probability that there is no interaction between the RNA sequence and the protein sequence to be predicted, or it may simultaneously show the probability that there is an interaction between the RNA sequence and the protein sequence to be predicted and the probability that there is no interaction. This disclosure does not make any specific limitations on this.

[0196] In the exemplary embodiments of this disclosure, reference is made to Figure 6 As shown, the recurrent neural network and feature extraction network can be pre-trained according to steps S610 to S650. At the same time, the parameters in each parameter matrix are trained to optimize each prediction model and all parameters in the parameter matrix. Then, the final model obtained from the training can be used to predict RNA sequences and protein sequences with unknown interactions.

[0197] Step S610. Obtain a training dataset, which includes positive RNA-protein pairs and negative RNA-protein pairs.

[0198] For example, various models can be trained based on the RPI1807 dataset. This dataset contains 3243 RNA-protein pairs, specifically 1807 positive pairs and 1436 negative pairs. Positive pairs indicate an interaction between the RNA and protein sequences in the RNA-protein pair, while negative pairs indicate no interaction. A training dataset of 1200 positive and 1000 negative pairs can be selected, or all RNA-protein pairs can be used. It's important to understand that the number of RNA-protein pairs in the training dataset is merely illustrative; any number of RNA-protein pairs can be used to train each model multiple times to improve its performance. Positive RNA-protein pairs can be labeled with a value of "1," indicating an interaction, while negative RNA-protein pairs can be labeled with a value of "0," indicating no interaction. It is understandable that other examples can also be based on the RPI2241 dataset, RPI369 dataset, etc., and this disclosure does not make any specific restrictions on this.

[0199] Step S620. Use the recurrent neural network and feature extraction network to determine the predicted interaction value for each RNA-protein pair in the training dataset.

[0200] Similarly, a recurrent neural network can be used to encode the RNA and protein sequences in each RNA-protein pair in the training dataset, obtaining corresponding RNA and protein vector sequences. A matching feature matrix is ​​then constructed for each RNA-protein pair based on the RNA and protein vector sequences, and this matrix is ​​input into a feature extraction network for feature extraction. Finally, a classifier can be used to classify and predict the extracted matching features, obtaining the predicted interaction value for each RNA-protein pair.

[0201] Step S630. Calculate the predicted interaction value and label value of each RNA-protein pair in the training dataset using the loss function to obtain the corresponding loss value.

[0202] Each RNA-protein pair in the training dataset has a label value, such as 1 for each positive pair and 0 for each negative pair. For example, the i-th RNA-protein pair is a positive example, and its label value is 1. The interaction value p(1|r) can be predicted based on this RNA-protein pair. i ,p i p(0|r) i ,p iThe loss function is calculated using the label value 1 and the target value 1, yielding the corresponding loss value. During model training, the goal is to minimize the interaction predictions to be infinitely close to the label values, i.e., to minimize the objective function. In one example, the cross-entropy loss function can be chosen as the objective function. When calculating the cross-entropy loss function, if the label value is 1, p(1|r) = 1 / 2 * ... i ,p i The closer p(1|r) is to 1, the smaller the calculated loss value. i ,p i The closer p(0|r) is to 0, the larger the calculated loss value. At the same time, p(0|r) i ,p i The closer p(0|r) is to 1, the larger the calculated loss value. i ,p i The closer the cross-entropy loss function is to 0, the smaller the calculated loss value. It's understandable that the cross-entropy loss function is a performance function in the prediction model, used to estimate the degree of discrepancy between the model's predicted values ​​and the labeled values. The smaller the calculated cross-entropy loss function value, the better the model's prediction performance.

[0203] Specifically, the cross-entropy loss function can be:

[0204]

[0205] Where, r i p represents the i-th RNA sequence in the training dataset. i Let y represent the i-th protein sequence in the training dataset. i p(1|r) represents the label value of the i-th RNA-protein pair in the training dataset. i ,p i ) represents the predicted value of the i-th RNA-protein pair in the training dataset that there is an interaction, p(0|r i ,p i ) represents the predicted value that the i-th RNA-protein pair in the training dataset does not interact, and K is the total number of RNA-protein pairs in the training dataset.

[0206] Step S640. Adjust the model parameters of the recurrent neural network and the feature extraction network according to the loss value.

[0207] The model parameters can be weight parameters, bias parameters, and parameter matrices, such as mapping matrices W1, W2, and W3. For example, the model parameters of each model can be iteratively updated based on the calculated loss value. When the iteration termination condition is met, the training of model parameters for multiple interaction prediction models is complete. For example, the stochastic gradient descent algorithm can be used to update the model parameters. According to the backpropagation principle, the objective function, such as the cross-entropy loss function, is continuously calculated, and the model parameters of each model are updated simultaneously based on the calculated loss value. When the objective function converges to its minimum value, the training of all model parameters is complete. Alternatively, the model parameters can be updated iteratively in reverse order. When the preset number of iterations is met, the training of all model parameters is complete. After iteration, the optimized model parameters can be obtained. In other examples, the objective function can be minimized alternately using least squares, Adam optimization algorithms, etc., and the model parameters can be updated sequentially from back to front to optimize the parameters.

[0208] During the training process described above, the parameters of the recurrent neural network (RNN) and the feature extraction network can be trained simultaneously. For example, using L as the objective function, the mapping matrix W6 in the fully connected layer can be adjusted first. Since a convolutional neural network is needed to extract features from the matching feature matrix before binary classification prediction, and a recurrent neural network is needed to encode the RNA and protein sequences to be predicted, backpropagation can be further performed on the convolutional neural network and the recurrent neural network to adjust the model parameters and mapping matrices W1 and W2 in the convolutional neural network and the recurrent neural network. Through multiple backpropagation layers, the parameters of each model can eventually converge, or training can terminate after a certain number of iterations. This training method allows for the simultaneous training of the recurrent neural network and the feature extraction network, ensuring higher accuracy and precision of each model while improving training efficiency. After training, the final models can be used to predict the interactions between the RNA and protein sequences to be predicted.

[0209] In one specific example implementation, reference is made to Figure 7 As shown, the interaction between the RNA and protein sequences to be predicted can be predicted using the trained LSTM network, convolutional neural network, and Softmax classifier according to steps S701 to S706.

[0210] Step S701. The RNA sequence to be predicted, AGCAA…GCA, is converted into N 3-mer subsequences such as AGC and AUA. Embedding encoding can be performed on each 3-mer subsequence to obtain an N 3-mer embedding vector. The protein sequence to be predicted, MTAQDD…SYS, is converted into M amino acid 3-mer subsequences such as MTA and QDD. Embedding encoding can be performed on each amino acid 3-mer subsequence to obtain an M amino acid 3-mer embedding vector.

[0211] Step S702. Input the obtained N base 3-mer embedding vectors and M amino acid 3-mer embedding vectors into the LSTM network respectively, and output the vector corresponding to each base 3-mer. and the vector corresponding to each amino acid 3-mer And a vector corresponding to N base 3-mers. Composition of RNA vector sequence Vectors corresponding to M amino acid 3-mers Composition of protein vector sequences

[0212] Step S703. Construct an N×M matching feature matrix. Based on the RNA vector sequence and protein vector sequences Construct a matching feature matrix of size N*M;

[0213] Step S704. Feature extraction using a convolutional neural network. The matching feature matrix is ​​input into the convolutional neural network for feature extraction to obtain the matching feature vector;

[0214] Step S705. Softmax classifier for prediction. The Softmax classifier is used to perform binary classification prediction on the matching feature vectors to obtain the predicted values ​​of the interaction between the RNA sequence and the protein sequence to be predicted;

[0215] Step S706. Output prediction results. Output the predicted values ​​of the interaction between the RNA sequence and protein sequence to be predicted to the terminal device for user viewing.

[0216] In this example embodiment, a matching feature matrix is ​​constructed using the matching relationship between RNA and protein sequences. This matching feature matrix accurately reflects the matching relationship between RNA and protein vector sequences. Compared to existing technologies that do not consider the association between individual k-mer subsequences during ncRPI prediction but treat each k-mer subsequence as an independent factor, this disclosure improves the accuracy of RNA-protein interaction prediction when predicting interactions based on the matching feature matrix. It should be noted that the interaction prediction method provided in this disclosure is applicable to, but not limited to, predicting interactions between RNA and protein sequences. For example, this interaction prediction method can be used to predict the interaction between a first substance and a second substance, where both the first and second substances can be represented by sequence segments; this disclosure does not limit this.

[0217] In this exemplary embodiment of the present disclosure, at least one RNA sequence can be obtained, and protein sequences that interact with each input RNA sequence can be searched in a database. For example, after a user inputs at least one RNA sequence, each input RNA sequence can be combined with all protein sequences in the database to form several RNA-protein pairs. Further, the interaction of each RNA-protein pair can be predicted according to steps S220 to S260. Specifically, a recurrent neural network can be used to encode the RNA and protein sequences in each RNA-protein pair to obtain corresponding RNA vector sequences and protein vector sequences. A matching feature matrix corresponding to each RNA-protein pair is constructed based on the RNA vector sequences and protein vector sequences, and the matching feature matrix is ​​input into a feature extraction network for feature extraction. Finally, a classifier can be used to classify and predict the extracted matching features to obtain the interaction prediction value for each RNA-protein pair. An interaction prediction value of 1 indicates that the RNA-protein pair interacts, and an interaction prediction value of 0 indicates that the RNA-protein pair does not interact. Then, all RNA-protein pairs with an interaction prediction value of 1 can be filtered out, and the protein sequences in each RNA-protein pair can be output to a terminal device for the user to view the protein sequences that interact with the input RNA sequence.

[0218] Similarly, in the exemplary embodiments of this disclosure, at least one protein sequence can be obtained, and RNA sequences that interact with each input protein sequence can be searched in a database. For example, after a user inputs at least one protein sequence, each input protein sequence can be combined with all RNA sequences in the database to form several RNA-protein pairs. Further, the interaction of each RNA-protein pair can be predicted according to steps S220 to S260. Specifically, a recurrent neural network can be used to encode the RNA and protein sequences in each RNA-protein pair to obtain corresponding RNA vector sequences and protein vector sequences. A matching feature matrix corresponding to each RNA-protein pair is constructed based on the RNA vector sequences and protein vector sequences, and the matching feature matrix is ​​input into a feature extraction network for feature extraction. Finally, a classifier can be used to classify and predict the extracted matching features to obtain the interaction prediction value for each RNA-protein pair. An interaction prediction value of 1 indicates that the RNA-protein pair interacts, and an interaction prediction value of 0 indicates that the RNA-protein pair does not interact. Then, all RNA-protein pairs with an interaction prediction value of 1 can be screened out, and the RNA sequence in each RNA-protein pair can be output to the terminal device so that the user can view the RNA sequences that interact with the input protein sequence.

[0219] In the RNA-protein interaction prediction method provided in the exemplary embodiments of this disclosure, the following steps are taken: First, obtain the RNA sequence and protein sequence to be predicted. Then, encode the RNA sequence to be predicted to obtain an RNA vector sequence. Next, encode the protein sequence to be predicted to obtain a protein vector sequence. Finally, construct a matching feature matrix based on the RNA and protein vector sequences. Extract features from the matching feature matrix and determine the interaction between the RNA and protein sequences based on the extracted matching features. This disclosure improves the accuracy of RNA-protein interaction prediction by constructing a matching feature matrix using the matching relationship between RNA and protein sequences and predicting the interaction between them based on the matching feature matrix.

[0220] It should be noted that although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0221] Furthermore, this example embodiment also provides an RNA-protein interaction prediction device. This device can be applied to a server or terminal device. (Reference) Figure 8 As shown, the RNA-protein interaction prediction device 800 may include a data acquisition module 810, a first data encoding module 820, a second data encoding module 830, a feature matrix construction module 840, and an interaction prediction module 850, wherein:

[0222] The data acquisition module 810 is used to acquire the RNA-protein pairs to be predicted;

[0223] The first data encoding module 820 is used to encode the RNA sequence to be predicted to obtain an RNA vector sequence;

[0224] The second data encoding module 830 is used to encode the protein sequence to be predicted to obtain a protein vector sequence.

[0225] The feature matrix construction module 840 is used to construct a matching feature matrix based on the RNA vector sequence and the protein vector sequence.

[0226] The interaction prediction module 850 is used to extract features from the matching feature matrix and determine the interaction between the RNA sequence and protein sequence to be predicted based on the extracted matching features.

[0227] In one optional implementation, the first data encoding module 820 includes:

[0228] The first sequence conversion module is used to convert the RNA sequence to be predicted into an N-base k-mer subsequence;

[0229] The first sequence encoding module is used to vectorize each base k-mer subsequence to obtain the RNA vector sequence.

[0230] In one optional implementation, the first sequence encoding module includes:

[0231] The first sequence encoding unit is used to encode each base k-mer subsequence to obtain a first vector of N base k-mer subsequences;

[0232] The first vector sequence determination unit is used to form the RNA vector sequence from the first vector of the N base k-mer subsequences.

[0233] In one optional implementation, the first sequence encoding module includes:

[0234] The second sequence encoding unit is used to encode each base k-mer subsequence to obtain a first vector of N base k-mer subsequences;

[0235] The second vector sequence determination unit is used to sequentially input the first vector of the N base k-mer subsequences into a pre-trained recurrent neural network, output N base k-mer vectors, and form the RNA vector sequence from the N base k-mer vectors.

[0236] In one optional implementation, the first sequence encoding module includes:

[0237] The third sequence encoding unit is used to encode each base k-mer subsequence to obtain a first vector of N base k-mer subsequences;

[0238] The third vector sequence determination unit is used to perform operations on the first vector of the N base k-mer subsequences using the first mapping matrix to obtain the second vector of the N base k-mer subsequences, and the second vector of the N base k-mer subsequences forms the RNA vector sequence.

[0239] In one optional implementation, the first sequence encoding module includes:

[0240] The fourth sequence encoding unit is used to encode each base k-mer subsequence to obtain a first vector of N base k-mer subsequences;

[0241] The first vector operation unit is used to perform operations on the first vector of the N base k-mer subsequences using the first mapping matrix to obtain the second vector of the N base k-mer subsequences.

[0242] The fourth vector sequence determination unit is used to sequentially input the second vector of the N base k-mer subsequences into a pre-trained recurrent neural network, output N base k-mer vectors, and form the RNA vector sequence from the N base k-mer vectors.

[0243] In one alternative implementation, the second data encoding module 830 includes:

[0244] The second sequence conversion module is used to convert the protein sequence to be predicted into an M-amino acid k-mer subsequence;

[0245] The second sequence encoding module is used to vectorize each amino acid k-mer subsequence to obtain the protein vector sequence.

[0246] In one optional implementation, the second sequence encoding module includes:

[0247] The fifth sequence coding unit is used to encode each amino acid k-mer subsequence to obtain a first vector of M amino acid k-mer subsequences;

[0248] The fifth vector sequence determination unit is used to form the protein vector sequence from the first vector of the M amino acid k-mer subsequences.

[0249] In one optional implementation, the second sequence encoding module includes:

[0250] The sixth sequence coding unit is used to encode each amino acid k-mer subsequence to obtain a first vector of M amino acid k-mer subsequences;

[0251] The sixth vector sequence determination unit is used to sequentially input the first vector of the M amino acid k-mer subsequences into a pre-trained recurrent neural network, output M amino acid k-mer vectors, and form the protein vector sequence from the M amino acid k-mer vectors.

[0252] In one optional implementation, the second sequence encoding module includes:

[0253] The seventh sequence coding unit is used to encode each amino acid k-mer subsequence to obtain a first vector of M amino acid k-mer subsequences;

[0254] The seventh vector sequence determination unit is used to perform operations on the first vector of the M amino acid k-mer subsequences using the second mapping matrix to obtain the second vector of the M amino acid k-mer subsequences, and the second vector of the M amino acid k-mer subsequences forms the protein vector sequence.

[0255] In one optional implementation, the second sequence encoding module includes:

[0256] The eighth sequence coding unit is used to encode each amino acid k-mer subsequence to obtain a first vector of M amino acid k-mer subsequences;

[0257] The second vector operation unit is used to perform operations on the first vector of the M amino acid k-mer subsequences using the second mapping matrix to obtain the second vector of the M amino acid k-mer subsequences.

[0258] The eighth vector sequence determination unit is used to sequentially input the second vectors of the M amino acid k-mer subsequences into a pre-trained recurrent neural network, output M amino acid k-mer vectors, and form the protein vector sequence from the M amino acid k-mer vectors.

[0259] In one alternative implementation, the feature matrix construction module 840 includes:

[0260] The matching degree calculation unit is used to calculate the matching degree between each base k-mer vector in the RNA vector sequence and each amino acid k-mer vector in the protein vector sequence;

[0261] The feature matrix construction unit is used to construct the matching feature matrix by using the calculated matching score as an element of the matching feature matrix.

[0262] In one optional implementation, the matching degree calculation unit includes:

[0263] The first matching degree calculation subunit is used to calculate based on:

[0264]

[0265] Calculate the k-mer vector of the i-th base in the RNA vector sequence. The k-mer vector of the j-th amino acid in the protein vector sequence Matching degree between in, express Length, express The length.

[0266] In an optional implementation, the feature matrix construction module 840 is configured to perform a dot product operation between the RNA vector sequence and the protein vector sequence to obtain the matching feature matrix.

[0267] In one alternative implementation, the interaction prediction module 850 includes:

[0268] The feature extraction module is used to extract features from the matching feature matrix using a feature extraction network to obtain the matching features.

[0269] In one optional implementation, the feature extraction module includes:

[0270] The feature extraction unit is used to extract features from the matching feature matrix using a feature extraction network to obtain the original features;

[0271] The third vector operation unit is used to perform operations on the original features using the third mapping matrix to obtain the matching features.

[0272] In an optional implementation, the interaction prediction module 850 further includes:

[0273] The prediction value acquisition module is used to obtain the predicted interaction value between the RNA sequence and the protein sequence to be predicted based on the matching features;

[0274] An interaction determination module is used to determine the interaction between the RNA sequence and the protein sequence to be predicted based on the interaction prediction value.

[0275] In an optional implementation, the interaction prediction module 850 further includes:

[0276] The interaction prediction module is used to input the matching features into the classifier and output the predicted interaction value between the RNA sequence and the protein sequence to be predicted.

[0277] In one optional implementation, the probability of an interaction between the RNA sequence and the protein sequence to be predicted is:

[0278]

[0279] Where r represents the RNA sequence to be predicted, p represents the protein sequence to be predicted, c0 represents the first feature value in the matching features, and c1 represents the second feature value in the matching features.

[0280] In one alternative implementation, the interaction determination module is configured to determine that there is an interaction between the RNA sequence and the protein sequence to be predicted if the interaction prediction value meets a preset threshold condition.

[0281] In an optional embodiment, the RNA-protein interaction prediction device 800 further includes:

[0282] The training module is used to train the recurrent neural network and the feature extraction network.

[0283] In one alternative implementation, the training module includes:

[0284] The training data acquisition module is used to acquire a training dataset, which includes positive RNA-protein pairs and negative RNA-protein pairs.

[0285] A prediction output module is used to determine the predicted interaction value of each RNA-protein pair in the training dataset using the recurrent neural network and the feature extraction network;

[0286] The loss calculation module is used to calculate the interaction prediction value and label value of each RNA-protein pair in the training dataset using a loss function to obtain the corresponding loss value;

[0287] The model parameter adjustment module is used to adjust the model parameters of the recurrent neural network and the feature extraction network according to the loss value.

[0288] In one optional implementation, the loss function is:

[0289]

[0290] Where, r i p represents the i-th RNA sequence in the training dataset. i Let y represent the i-th protein sequence in the training dataset. i p(1|r) represents the label value of the i-th RNA-protein pair in the training dataset. i ,p i ) represents the predicted value of the i-th RNA-protein pair in the training dataset that there is an interaction, p(0|r i ,p i ) represents the predicted value that the i-th RNA-protein pair in the training dataset does not interact, and K is the total number of RNA-protein pairs in the training dataset.

[0291] In one alternative implementation, the model parameter adjustment module is configured to iteratively update the model parameters of the recurrent neural network and the feature extraction network based on the loss value using a stochastic gradient descent algorithm, and complete the training of all model parameters when the iteration termination condition is met.

[0292] In an optional embodiment, the RNA-protein interaction prediction device 800 further includes:

[0293] The data output module is used to output the prediction results of the interaction between the RNA sequence and the protein sequence to be predicted.

[0294] The specific details of each module in the above-mentioned RNA-protein interaction prediction device have been described in detail in the corresponding RNA-protein interaction prediction methods, so they will not be repeated here.

[0295] The modules in the above-described device can be general-purpose processors, including central processing units (CPUs), network processors, etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. Each module can also be implemented using software, firmware, etc. The processors in the above-described device can be independent processors or integrated together.

[0296] Exemplary embodiments of this disclosure also provide a computer-readable storage medium having a program product stored thereon capable of implementing the methods described above in this specification. In some possible embodiments, various aspects of this disclosure may also be implemented as a program product including program code that, when run on an electronic device, causes the electronic device to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure. This program product may be a portable compact disc read-only memory (CD-ROM) including program code and may run on an electronic device, such as a personal computer. However, the program product of this disclosure is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0297] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0298] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0299] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0300] Program code for performing the operations of this disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0301] Exemplary embodiments of this disclosure also provide an electronic device capable of implementing the above-described method. Referring below... Figure 9 To describe an electronic device 900 according to such an exemplary embodiment of the present disclosure. Figure 9 The electronic device 900 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0302] like Figure 9 As shown, the electronic device 900 can be represented as a general-purpose computing device. The components of the electronic device 900 may include, but are not limited to: at least one processing unit 910, at least one storage unit 920, a bus 930 connecting different system components (including the storage unit 920 and the processing unit 910), and a display unit 940.

[0303] The storage unit 920 stores program code that can be executed by the processing unit 910, causing the processing unit 910 to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure. For example, the processing unit 910 can execute... Figures 2 to 7 Any one or more of the method steps.

[0304] Storage unit 920 may include readable media in the form of volatile storage units, such as random access memory (RAM) 921 and / or cache memory 922, and may further include read-only memory (ROM) 923.

[0305] The storage unit 920 may also include a program / utility 924 having a set (at least one) of program modules 925, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0306] Bus 930 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0307] Electronic device 900 can also communicate with one or more external devices 1000 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 900, and / or with any device that enables electronic device 900 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 950. Furthermore, electronic device 900 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 960. As shown, network adapter 960 communicates with other modules of electronic device 900 via bus 930. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 900, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0308] In some embodiments, the RNA-protein interaction prediction method described herein can be executed by the processing unit 910 of an electronic device. In some embodiments, the RNA sequence and protein sequence to be predicted, as well as training datasets for training various models, can be input through the input interface 950. For example, the RNA sequence and protein sequence to be predicted, as well as training datasets for training various models, can be input through the user interface of the electronic device. In some embodiments, the prediction results of the interaction between the RNA sequence and protein sequence to be predicted can be output to an external device 1000 for user viewing through the output interface 950.

[0309] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the method according to the exemplary embodiments of this disclosure.

[0310] Furthermore, the above figures are merely illustrative representations of the processes included in the methods according to exemplary embodiments of this disclosure, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0311] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0312] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for predicting RNA-protein interactions, characterized in that, include: Obtain the RNA and protein sequences to be predicted; The RNA sequence to be predicted is encoded to obtain an RNA vector sequence; The protein sequence to be predicted is encoded to obtain a protein vector sequence; Construct a matching feature matrix based on the RNA vector sequence and protein vector sequence; Feature extraction is performed on the matching feature matrix, and the interaction between the RNA sequence and protein sequence to be predicted is determined based on the extracted matching features.

2. The method for predicting RNA-protein interactions according to claim 1, characterized in that, The process of encoding the RNA sequence to be predicted to obtain an RNA vector sequence includes: The RNA sequence to be predicted is converted into an N-base k-mer subsequence; The RNA vector sequence is obtained by vectorizing each base k-mer subsequence.

3. The method for predicting RNA-protein interactions according to claim 2, characterized in that, The vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes: Each base k-mer subsequence is encoded to obtain a first vector of N base k-mer subsequences, and the first vector of the N base k-mer subsequences is used to form the RNA vector sequence.

4. The method for predicting RNA-protein interactions according to claim 2, characterized in that, The vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes: Encode each k-mer subsequence to obtain the first vector of N k-mer subsequences; The first vector of the N base k-mer subsequences is sequentially input into a pre-trained recurrent neural network, which outputs N base k-mer vectors, and the N base k-mer vectors form the RNA vector sequence.

5. The method for predicting RNA-protein interactions according to claim 2, characterized in that, The vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes: Encode each k-mer subsequence to obtain the first vector of N k-mer subsequences; The first vector of the N base k-mer subsequences is processed using the first mapping matrix to obtain the second vector of the N base k-mer subsequences, and the second vector of the N base k-mer subsequences is used to form the RNA vector sequence.

6. The method for predicting RNA-protein interactions according to claim 2, characterized in that, The vectorization of each base k-mer subsequence to obtain the RNA vector sequence includes: Encode each k-mer subsequence to obtain the first vector of N k-mer subsequences; The first vector of the N base k-mer subsequences is obtained by performing operations on the first vector of the N base k-mer subsequences using the first mapping matrix; The second vectors of the N base k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs N base k-mer vectors, and the N base k-mer vectors form the RNA vector sequence.

7. The method for predicting RNA-protein interactions according to claim 1, characterized in that, The process of encoding the protein sequence to be predicted to obtain a protein vector sequence includes: The protein sequence to be predicted is converted into an M-amino acid k-mer subsequence; The protein vector sequence is obtained by vectorizing each amino acid k-mer subsequence.

8. The method for predicting RNA-protein interactions according to claim 7, characterized in that, The vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes: Each amino acid k-mer subsequence is encoded to obtain a first vector of M amino acid k-mer subsequences, and the first vector of the M amino acid k-mer subsequences is used to form the protein vector sequence.

9. The method for predicting RNA-protein interactions according to claim 7, characterized in that, The vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes: Encode each amino acid k-mer subsequence to obtain the first vector of M amino acid k-mer subsequences; The first vectors of the M amino acid k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs M amino acid k-mer vectors, and the protein vector sequence is composed of the M amino acid k-mer vectors.

10. The method for predicting RNA-protein interactions according to claim 7, characterized in that, The vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes: Encode each amino acid k-mer subsequence to obtain the first vector of M amino acid k-mer subsequences; The first vector of the M amino acid k-mer subsequences is processed using the second mapping matrix to obtain the second vector of the M amino acid k-mer subsequences, and the second vector of the M amino acid k-mer subsequences is used to form the protein vector sequence.

11. The method for predicting RNA-protein interactions according to claim 7, characterized in that, The vectorization of each amino acid k-mer subsequence to obtain the protein vector sequence includes: Encode each amino acid k-mer subsequence to obtain the first vector of M amino acid k-mer subsequences; The second vector of the M amino acid k-mer subsequences is obtained by operating on the first vector of the M amino acid k-mer subsequences using the second mapping matrix; The second vectors of the M amino acid k-mer subsequences are sequentially input into a pre-trained recurrent neural network, which outputs M amino acid k-mer vectors, and the protein vector sequence is composed of the M amino acid k-mer vectors.

12. The method for predicting RNA-protein interactions according to claim 1, characterized in that, The step of constructing a matching feature matrix based on the RNA vector sequence and the protein vector sequence includes: Calculate the matching degree between the base k-mer vector in the RNA vector sequence and the amino acid k-mer vector in the protein vector sequence; The calculated matching score is used as an element of the matching feature matrix to construct the matching feature matrix.

13. The method for predicting RNA-protein interactions according to claim 12, characterized in that, The calculation of the matching degree between the base k-mer vector in the RNA vector sequence and the amino acid k-mer vector in the protein vector sequence includes: according to: Calculate the first RNA vector sequence. i k-mer vector of bases With the protein vector sequence of the first j amino acid k-mer vector Matching degree between ;in, express Length, express The length.

14. The method for predicting RNA-protein interactions according to claim 1, characterized in that, The step of constructing a matching feature matrix based on the RNA vector sequence and the protein vector sequence includes: The matching feature matrix is ​​obtained by performing a dot product operation between the RNA vector sequence and the protein vector sequence.

15. The method for predicting RNA-protein interactions according to claim 1, characterized in that, The feature extraction of the matching feature matrix includes: The matching features are obtained by using a feature extraction network to extract features from the matching feature matrix.

16. The method for predicting RNA-protein interactions according to claim 15, characterized in that, The step of using a feature extraction network to extract features from the matching feature matrix to obtain the matching features includes: The matching feature matrix is ​​used to extract features to obtain the original features; The original features are processed using the third mapping matrix to obtain the matching features.

17. The method for predicting RNA-protein interactions according to claim 1, characterized in that, The step of determining the interaction between the RNA sequence and protein sequence to be predicted based on the extracted matching features includes: The predicted interaction values ​​between the RNA sequence and the protein sequence to be predicted are obtained based on the matching features. The interaction between the RNA sequence and the protein sequence to be predicted is determined based on the interaction prediction value.

18. The method for predicting RNA-protein interactions according to claim 17, characterized in that, The step of obtaining the predicted interaction value between the RNA sequence and the protein sequence to be predicted based on the matching features includes: The matching features are input into the classifier, which outputs the probability that there is an interaction between the RNA sequence and the protein sequence to be predicted.

19. The method for predicting RNA-protein interactions according to claim 18, characterized in that, The probability that there is an interaction between the RNA sequence and the protein sequence to be predicted is: in, This represents the RNA sequence to be predicted. This represents the protein sequence to be predicted. This represents the first feature value in the matching features. This represents the second feature value in the matching features.

20. The method for predicting RNA-protein interactions according to claim 17, characterized in that, Determining the interaction between the RNA sequence and the protein sequence to be predicted based on the interaction prediction value includes: If the predicted interaction value meets a preset threshold condition, it is determined that there is an interaction between the RNA sequence and the protein sequence to be predicted.

21. The method for predicting RNA-protein interactions according to any one of claims 4, 6, 9, and 11, characterized in that, The method further includes: The recurrent neural network and the feature extraction network are trained.

22. The method for predicting RNA-protein interactions according to claim 21, characterized in that, The training of the recurrent neural network and the feature extraction network includes: Obtain a training dataset, which includes positive RNA-protein pairs and negative RNA-protein pairs; The recurrent neural network and feature extraction network are used to determine the predicted interaction values ​​for each RNA-protein pair in the training dataset; The loss function is used to calculate the predicted interaction value and label value of each RNA-protein pair in the training dataset to obtain the corresponding loss value; The model parameters of the recurrent neural network and the feature extraction network are adjusted based on the loss value.

23. The method for predicting RNA-protein interactions according to claim 22, characterized in that, The loss function is: in, Represents the first in the training dataset i RNA sequences, Represents the first in the training dataset i A protein sequence, Represents the first in the training dataset i The tag value of each RNA-protein pair Represents the first in the training dataset i The predicted value for the existence of interactions between RNA-protein pairs, Represents the first in the training dataset i The predicted value for the absence of interaction between RNA-protein pairs. K This represents the total number of RNA-protein pairs in the training dataset.

24. The method for predicting RNA-protein interactions according to claim 22, characterized in that, The step of adjusting the model parameters of the recurrent neural network and the feature extraction network based on the loss value includes: Based on the loss value, the model parameters of the recurrent neural network and the feature extraction network are iteratively updated using the stochastic gradient descent algorithm. When the iteration termination condition is met, the training of all model parameters is completed.

25. The method for predicting RNA-protein interactions according to claim 1, characterized in that, The method further includes: Output the predicted results of the interaction between the RNA sequence and the protein sequence to be predicted.

26. An RNA-protein interaction prediction device, characterized in that, include: The data acquisition module is used to acquire the RNA and protein sequences to be predicted. The first data encoding module is used to encode the RNA sequence to be predicted to obtain an RNA vector sequence; The second data encoding module is used to encode the protein sequence to be predicted to obtain a protein vector sequence. A feature matrix construction module is used to construct a matching feature matrix based on the RNA vector sequence and the protein vector sequence; An interaction prediction module is used to extract features from the matching feature matrix and determine the interaction between the RNA sequence and protein sequence to be predicted based on the extracted matching features.

27. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method according to any one of claims 1-25.

28. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to execute the method of any one of claims 1-25 by executing the executable instructions.

Images