SiRNA screening method and device based on deep learning

Abstract

The invention relates to the technical field of deep learning, and discloses a siRNA screening method and device based on deep learning, and the method comprises the steps: constructing a siRNA inhibition ratio prediction model; based on the siRNA inhibition rate prediction model, obtaining a preliminary siRNA screening result corresponding to the mRNA sequence of the target spot to be predicted; performing molecular dynamics simulation on the to-be-predicted target mRNA sequence to generate a molecular dynamics trajectory corresponding to the to-be-predicted target mRNA sequence; constructing a neural relation inference model, and generating an interaction relation graph corresponding to the mRNA sequence of the target spot to be predicted; and screening the preliminary siRNA screening result again on the basis of the interaction relation graph to obtain a final siRNA screening result. According to the method, the accuracy of the deep learning model in a real screening task is effectively improved, and the problems of weak generalization ability and the like caused by deviation of a training set when siRNA is screened by the deep learning model are avoided.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a siRNA screening method and device based on deep learning. Background Art

[0002] RNA interference (RNAi), also known as post-transcriptional gene silencing, is a mechanism that can resist parasitic and pathogenic nucleic acids and regulate the expression of specific genes. It can manually regulate gene expression and detect gene function. The regulatory factors of RNAi include small interfering RNA (siRNA) of 21 to 23 nucleotides and microRNA (miRNA) of about 22 nucleotides. Among them, the antisense strand (AS) in the siRNA duplex binds to the target mRNA through Watson-Crick base pairing. If AS completely hybridizes with the target mRNA, it will lead to the cutting of the target mRNA and prevent translation; if only the seed region of AS binds to the target mRNA, it will induce the degradation of the target mRNA and inhibit translation.

[0003] The target mRNA is generally composed of hundreds to thousands of nucleotides, and a large number of candidate siRNAs can be generated using the sliding window method. However, there are significant differences in the inhibition rate of the generated siRNAs, and their composition can also vary. The siRNA inhibition rate is mainly affected by the sequence pattern, binding affinity, and the secondary structure around the binding region. The prediction of siRNA inhibition rate is very difficult, so researchers have been exploring models to accurately predict siRNA inhibition rate.

[0004] The existing siRNA inhibition rate prediction models mainly include expert experience-based methods, machine learning-based methods, and deep learning-based methods. The expert experience-based method uses the design experience and statistical analysis of experts to develop a set of siRNA design principles, which are usually applicable to the case of a small amount of siRNA data. For example, Reynolds et al. summarized some principles for rational design of siRNA by using statistical methods based on 180 siRNAs designed for two genes: GC content is between 0.30 and 0.52; there are at least 3 A or U at positions 15 to 19 of the sense chain; avoid the structure that can form a hairpin loop; the sense chain is A at position 19, A at position 3, U at position 10, non-G or non-C at position 19, and non-G at position 13. The machine learning-based and deep learning-based methods are data-driven, and extract features from the siRNA data set through machine learning and deep learning algorithms to predict the inhibition rate of siRNA. Among them, the machine learning-based method requires manual feature extraction when extracting features, which inevitably leads to errors caused by feature extraction; while the deep learning-based method does not require manual participation when extracting features. In recent years, models such as graph convolutional neural networks, graph neural networks, and Transformer have been used to predict siRNA inhibition rates.

[0005] Although deep learning methods have made good progress in the prediction of siRNA inhibition rate, they are extremely dependent on the quality of the data set. When the distribution of siRNA data in the actual screening task is significantly different from that in the training set, it is difficult for the deep learning model to accurately screen out the results.

[0006] Moreover, in actual application scenarios, the distribution of siRNA data in real screening tasks is often quite different from the distribution of siRNA data in the training set. Summary of the invention

[0007] In view of this, the present invention provides a siRNA screening method and device based on deep learning to solve the problem that the siRNA prediction method based on deep learning in the prior art is overly dependent on the model training data set.

[0008] In a first aspect, the present invention provides a siRNA screening method based on deep learning, the method comprising:

[0009] Construct a siRNA inhibition rate prediction model;

[0010] Based on the siRNA inhibition rate prediction model, the preliminary siRNA screening results corresponding to the target mRNA sequence to be predicted are obtained;

[0011] Perform molecular dynamics simulation on the target mRNA sequence to be predicted, and generate the molecular dynamics trajectory corresponding to the target mRNA sequence to be predicted;

[0012] Construct a neural relationship inference model to generate an interaction relationship diagram corresponding to the target mRNA sequence to be predicted;

[0013] Based on the interaction relationship diagram, the preliminary siRNA screening results were screened again to obtain the final siRNA screening results.

[0014] The present invention greatly improves the accuracy of siRNA screening by constructing an interaction relationship diagram between target mRNA residues, improves the accuracy of the deep learning model in real screening tasks, and effectively avoids the problem of weak generalization ability of the deep learning model when screening siRNA due to the deviation of the training set.

[0015] In an optional embodiment, constructing a siRNA inhibition rate prediction model includes:

[0016] Acquire a siRNA data set, the siRNA data set includes: a siRNA sequence, a target mRNA sequence corresponding to the siRNA sequence, and an inhibition rate of the siRNA sequence to the corresponding target mRNA sequence;

[0017] Obtain the thermodynamic property characteristics and expert experience characteristics corresponding to the siRNA sequence;

[0018] Splicing the siRNA sequence in the siRNA data set with the local target mRNA sequence corresponding to the siRNA sequence, and adding the corresponding position code to the spliced ​​sequence to obtain a coding spliced ​​sequence; wherein the local target mRNA sequence is the sequence matching the siRNA sequence plus a preset number of residue sequences before and after the match with the siRNA sequence;

[0019] Input the coding splicing sequence into the pre-built feature network structure for feature extraction to obtain sequence features;

[0020] A preset fully connected neural network is trained based on the inhibition rate of the siRNA sequence on the corresponding target mRNA sequence, sequence characteristics, thermodynamic property characteristics and expert experience characteristics to obtain a siRNA inhibition rate prediction model after training.

[0021] In this embodiment, local sequence splicing and position encoding are used to enhance feature representation, which effectively improves the model's accurate prediction of siRNA inhibition rate. At the same time, combining thermodynamic property characteristics with expert experience characteristics enables the model to more comprehensively capture the interaction between siRNA and target mRNA, effectively improving the reliability and effectiveness of model prediction.

[0022] In an optional embodiment, based on the siRNA inhibition rate prediction model, obtaining preliminary siRNA screening results corresponding to the target mRNA sequence to be predicted includes:

[0023] Constructing the antisense sequence of the siRNA sequence of the target mRNA sequence to be predicted;

[0024] The antisense sequence of the siRNA sequence is input into the siRNA inhibition rate prediction model to obtain the siRNA predicted inhibition rate corresponding to the siRNA sequence one by one;

[0025] Sort the siRNA predicted inhibition rates from high to low;

[0026] Starting from the highest siRNA predicted inhibition rate, siRNA sequences with preset ratios were selected in turn as preliminary siRNA screening results.

[0027] In this embodiment, by constructing the siRNA antisense sequence corresponding to the target mRNA sequence and accurately predicting its inhibition rate through the siRNA inhibition rate prediction model, the siRNA with the best inhibition effect can be effectively screened out.

[0028] In an optional embodiment, constructing an antisense sequence of a siRNA sequence of a target mRNA sequence to be predicted comprises:

[0029] Starting from the 5' end of the target mRNA sequence, the sequence is translated toward the 3' end until the end of the antisense sequence of the siRNA sequence is flush with the 3' end of the target mRNA; the length of the antisense sequence of the siRNA sequence ranges from 19nt to 21nt.

[0030] In this embodiment, by translating from the 5' end to the 3' end of the local target mRNA sequence and matching siRNA antisense sequences of different lengths, the method can comprehensively evaluate multiple binding sites on the mRNA, thereby improving the design flexibility and prediction accuracy of siRNA.

[0031] In an optional embodiment, the neural relationship inference model is constructed based on the variational autoencoder model, and the neural relationship inference model is constructed to generate an interaction relationship graph corresponding to the target mRNA sequence to be predicted, including:

[0032] Obtain node data corresponding to the molecular dynamics trajectory, where the node is the residue in the target mRNA sequence to be predicted, and the node data includes the spatial coordinates and velocity of the node at multiple time steps;

[0033] Input the node data into the encoder structure in the neural relational reasoning model, extract the residue features and generate latent variables based on the residue features;

[0034] Input the latent variables into the decoder structure in the neural relational reasoning model, reconstruct the molecular dynamics trajectory, and calculate the reconstruction error based on the reconstructed molecular dynamics trajectory and the original molecular dynamics trajectory. Back-propagate the reconstruction error to update the encoder parameters, latent variables and decoder parameters, and complete the training after multiple iterations.

[0035] The trained neural relationship inference model is obtained. The latent variable generated by the trained neural relationship inference model is the interaction relationship diagram corresponding to the target mRNA sequence to be predicted.

[0036] In this embodiment, the variational autoencoder structure can effectively extract the characteristics of residues in the molecular dynamics trajectory, capture complex molecular interaction information, and effectively improve the accuracy of siRNA screening.

[0037] In an optional embodiment, the preliminary siRNA screening results are screened again based on the interaction relationship diagram to obtain the final siRNA screening results, including:

[0038] Determine the degree of each node based on the adjacency matrix included in the interaction graph, where the degree is the number of connections between the node and other nodes;

[0039] Generate a graph of the relationship between nodes and degrees, where the horizontal axis is the residue number and the vertical axis is the degree;

[0040] Determine whether the residue number of the target mRNA sequence corresponding to the siRNA sequence in the preliminary siRNA screening result is within the residue number range corresponding to the trough area in the relationship curve graph;

[0041] If it is within the residue number range corresponding to the valley area in the relationship curve graph, the corresponding siRNA sequence is retained;

[0042] If it is not within the residue number range corresponding to the trough area in the relationship curve diagram, the corresponding siRNA sequence will be deleted from the preliminary siRNA screening results.

[0043] In this embodiment, the neural relationship reasoning model is combined with the siRNA inhibition rate prediction model to further improve the accuracy of siRNA screening based on the interaction relationship diagram.

[0044] In a second aspect, the present invention provides a siRNA screening device based on deep learning, the device comprising:

[0045] A construction module for constructing a siRNA inhibition rate prediction model;

[0046] The siRNA inhibition rate prediction module is used to obtain preliminary siRNA screening results corresponding to the target mRNA sequence to be predicted based on the siRNA inhibition rate prediction model;

[0047] A molecular dynamics trajectory generation module is used to perform molecular dynamics simulation on the target mRNA sequence to be predicted and generate a molecular dynamics trajectory corresponding to the target mRNA sequence to be predicted;

[0048] A relationship graph generation module is used to construct a neural relationship inference model and generate an interaction relationship graph corresponding to the target mRNA sequence to be predicted;

[0049] The screening module is used to re-screen the preliminary siRNA screening results based on the interaction relationship diagram to obtain the final siRNA screening results.

[0050] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the deep learning-based siRNA screening method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer instructions, which are used to enable a computer to execute the deep learning-based siRNA screening method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0052] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions, which are used to enable a computer to execute the deep learning-based siRNA screening method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0053] It should be noted that the siRNA screening device based on deep learning, the computer device, the computer-readable storage medium and the program product provided by the present invention correspond to the above-mentioned siRNA screening method based on deep learning. Therefore, for the beneficial effects of the siRNA screening device based on deep learning, the computer device, the computer-readable storage medium and the program product, please refer to the description of the corresponding beneficial effects of the siRNA screening method based on deep learning above, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 is a schematic diagram of a process of a siRNA screening method based on deep learning according to an embodiment of the present invention;

[0056] Figure 2 is a schematic diagram of a process of another siRNA screening method based on deep learning according to an embodiment of the present invention;

[0057] Figure 3 is a schematic diagram of a training process based on a Transformer deep learning model according to an embodiment of the present invention;

[0058] Figure 4 is a schematic diagram of local target mRNA selection according to an embodiment of the present invention;

[0059] Figure 5 is a schematic diagram of the training process of the neural relational reasoning model according to an embodiment of the present invention;

[0060] Figure 6 Schematic diagram of target mRNA residue distribution according to an embodiment of the present invention;

[0061] Figure 7 is a structural block diagram of a siRNA screening device based on deep learning according to an embodiment of the present invention;

[0062] Figure 8 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0064] According to an embodiment of the present invention, an embodiment of a siRNA screening method based on deep learning is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0065] In this embodiment, a siRNA screening method based on deep learning is provided, which can be executed by a server, a terminal, a mobile terminal and other devices. Figure 1 is a flow chart of the siRNA screening method according to an embodiment of the present invention, such as Figure 1 As shown, the process includes the following steps:

[0066] Step S101, constructing a siRNA inhibition rate prediction model. In this embodiment, the siRNA inhibition rate prediction model includes but is not limited to using a convolutional neural network model, a graph neural network model, and a Transformer model.

[0067] Step S102, based on the siRNA inhibition rate prediction model, obtaining the preliminary siRNA screening results corresponding to the target mRNA sequence to be predicted.

[0068] Step S103, molecular dynamics simulation is carried out to the target mRNA sequence to be predicted, and the molecular dynamics trajectory corresponding to the target mRNA sequence to be predicted is generated. Specifically, molecular dynamics simulation software can be used to carry out molecular dynamics simulation to the target mRNA sequence, and the dynamics trajectory of the target mRNA is obtained, wherein, the method of all-atom dynamics simulation should be used for molecular dynamics simulation, and the molecular dynamics simulation software adopted includes but is not limited to Gromacs, Amber, etc. In the present embodiment, molecular dynamics simulation is carried out to the target mRNA sequence using Gromacs molecular dynamics software, wherein, the simulation force field is CHARMM27 force field, and TITP3 is selected for water model, and the target mRNA molecule is placed in a cubic water box and solvated, and the system charge is neutralized by adding chloride ions and sodium ions. The energy of the system initially uses the steepest descent minimization method, minimizes 50000 steps, and then the position of the heavy atom is restricted to carry out system balance. The system can be first subjected to temperature equilibrium, using the Langevin temperature control method for 100 ps, ​​and then subjected to pressure equilibrium, using the isotropic Berendsen pressure control method with a total simulation time of 100 ps, ​​and finally a 50 ns molecular dynamics simulation at 300K.

[0069] Step S104, constructing a neural relationship reasoning model. The neural relationship reasoning model in this embodiment is constructed using a variational autoencoder model. The encoder of the variational autoencoder model extracts features based on the input target mRNA molecule dynamics trajectory and generates hidden variables. Then, the decoder reconstructs the dynamics trajectory of the target mRNA molecule based on these hidden variables, and calculates the reconstruction error based on the reconstructed trajectory and the original trajectory. The reconstruction error is back-propagated to update the encoder parameters, hidden variables and decoder parameters, and the training is completed after multiple iterations; the model after training is obtained, and the hidden variables generated by the model after training are the interaction relationship diagram corresponding to the target mRNA sequence to be predicted.

[0070] Step S105, based on the interaction relationship diagram, the preliminary siRNA screening results are screened again to obtain the final siRNA screening results.

[0071] The interaction relationship diagram includes the interactions between various residues. According to the residue characteristics in the interaction relationship diagram, the siRNA sequences initially screened can be screened again to determine the final screening results.

[0072] The present invention provides a virtual screening method based on a neural relationship inference model to improve the inhibition rate of siRNA screening by a deep learning model, referring to Figure 2 As shown, preliminary siRNA screening results are screened out according to the siRNA inhibition rate prediction model; molecular dynamics simulation software is used to simulate the molecular dynamics of the target mRNA to obtain the molecular dynamics trajectory of the target mRNA; the molecular dynamics trajectory is reconstructed using the neural relationship reasoning model, and the interaction relationship diagram of the target mRNA is obtained; finally, the target mRNA interaction relationship diagram obtained based on the neural relationship reasoning model is used to further screen siRNA based on the preliminary siRNA screening results. The present invention greatly improves the accuracy of siRNA screening by constructing the interaction relationship diagram between target mRNA residues, improves the accuracy of the deep learning model in the real screening task, and effectively avoids the problem that the deep learning model has weak generalization ability due to the deviation of the training set when screening siRNA.

[0073] In some optional embodiments, the siRNA inhibition rate prediction model is established by the following steps:

[0074] The siRNA data set is obtained, and the siRNA data set includes: siRNA sequence, target mRNA sequence corresponding to the siRNA sequence, and inhibition rate of the siRNA sequence to the corresponding target mRNA sequence.

[0075] Obtain the thermodynamic property characteristics and expert experience characteristics corresponding to the siRNA sequence.

[0076] The siRNA sequence in the siRNA data set is spliced ​​with the local target mRNA sequence corresponding to the siRNA sequence, and the corresponding position code is added to the spliced ​​sequence to obtain a coding spliced ​​sequence; wherein the local target mRNA sequence is a sequence matching the siRNA sequence plus a preset number of residue sequences before and after the match with the siRNA sequence.

[0077] The coding splicing sequence is input into the pre-built feature network structure for feature extraction to obtain sequence features.

[0078] Based on the inhibition rate of siRNA sequence on the corresponding target mRNA sequence, sequence characteristics, siRNA thermodynamic properties and expert experience characteristics, a preset fully connected neural network is trained to obtain a siRNA inhibition rate prediction model after training.

[0079] Before constructing the siRNA inhibition rate prediction model, it is first necessary to construct an siRNA data set, which should include three data items: siRNA sequence, mRNA target corresponding to siRNA, and inhibition rate of target corresponding to siRNA. In this embodiment, the siRNA data experimentally determined by Huesken, Reynolds et al. can be used as a data set. The collected siRNA inhibition rate data is processed to a total of 2816 items. The siRNA data set is divided into a training set, a validation set, and a test set at a ratio of 7:2:1 using a random partitioning method.

[0080] Further, a deep learning prediction model for siRNA inhibition rate is constructed, and the deep learning prediction model for siRNA inhibition rate is trained according to the siRNA data set. The trained model can be used to predict the inhibition rate of new siRNA. In this embodiment, a deep learning prediction model for siRNA inhibition rate is constructed based on the Transformer model. Figure 3 The figure shows a schematic diagram of the deep learning prediction model for siRNA inhibition rate based on the Transformer model. The input of the model includes two parts, the sequence of siRNA and the sequence of local target mRNA. The sequence of local target mRNA is the sequence that matches the siRNA plus the sequence of 20 residues before and after the match. Figure 4As shown. In addition, not all local target mRNA sequences can be matched with siRNA and truncated by 20 residues before and after. For those that cannot be truncated, the filling character "0" is used to fill. Before inputting the Transformer encoder, the siRNA and the local target mRNA sequence are first spliced, and then the position code is added to the spliced ​​sequence. In this embodiment, the Transformer encoder has two layers. After the Transformer encoder extracts the features, the linear layer is used to change the dimension of the feature vector, and then the feature vector is expanded by the expansion layer; in addition, the model also integrates the characteristics of the thermodynamic properties of siRNA and the characteristics of expert experience, that is, the expert experience thermodynamic characteristics, and the two characteristics are input into the fully connected neural network together with the inhibition rate and sequence characteristics of the siRNA sequence to the corresponding target mRNA sequence, and the training continues. The siRNA inhibition rate prediction model after training can be used to predict the inhibition rate of siRNA and output.

[0081] In this embodiment, the feature representation is enhanced by splicing and position encoding of siRNA sequences and local target mRNA sequences, which effectively improves the model's accurate prediction of siRNA inhibition rate. At the same time, combining thermodynamic property characteristics with expert experience characteristics enables the model to more comprehensively capture the interaction relationship between siRNA and target mRNA, effectively improving the reliability and effectiveness of model prediction.

[0082] In some optional embodiments, based on the siRNA inhibition rate prediction model, obtaining preliminary siRNA screening results corresponding to the target mRNA sequence to be predicted includes:

[0083] Construct the antisense sequence of the siRNA sequence of the target mRNA sequence to be predicted.

[0084] The antisense sequence of the siRNA sequence is input into the siRNA inhibition rate prediction model to obtain the siRNA predicted inhibition rate corresponding to the siRNA sequence one by one.

[0085] Sort the siRNA predicted inhibition rates from high to low.

[0086] Starting from the highest siRNA predicted inhibition rate, siRNA sequences with preset ratios were selected in turn as preliminary siRNA screening results.

[0087] In this embodiment, the antisense sequence of the siRNA sequence of the target mRNA is constructed according to the principle of mutual base matching. The 5' end of the target mRNA should be used as the starting point during construction, and the 3' end should be translated in sequence until the end of the antisense sequence of the siRNA sequence is flush with the 3' end of the target mRNA, and the antisense sequence length of the siRNA sequence should be 19nt-21nt. The antisense sequence of the constructed siRNA sequence is then input into the constructed siRNA inhibition rate deep learning prediction model to predict the inhibition rate. Finally, the siRNA can be sorted from high to low according to the predicted siRNA inhibition rate, for example, the top 10% of siRNA is selected. In this embodiment, a total of 702 siRNAs were constructed, and 70 siRNAs were obtained after screening by the siRNA inhibition rate deep learning model.

[0088] In this embodiment, by constructing the siRNA antisense sequence corresponding to the target mRNA sequence and accurately predicting its inhibition rate through the siRNA inhibition rate prediction model, the siRNA with the best inhibition effect can be effectively screened out.

[0089] In some optional embodiments, constructing the antisense sequence of the siRNA sequence of the target mRNA sequence to be predicted includes:

[0090] Starting from the 5' end of the target mRNA sequence, the sequence is translated toward the 3' end until the end of the antisense sequence of the siRNA sequence is flush with the 3' end of the target mRNA; the length of the antisense sequence of the siRNA sequence ranges from 19nt to 21nt.

[0091] In this embodiment, by translating from the 5' end to the 3' end of the local target mRNA sequence and matching siRNA antisense sequences of different lengths, this method can comprehensively evaluate multiple binding sites on mRNA, thereby improving the design flexibility and prediction accuracy of siRNA.

[0092] In some optional embodiments, the neural relational inference model is constructed based on a variational autoencoder model, and the molecular dynamics trajectory is reconstructed based on the neural relational inference model to generate an interaction relationship diagram corresponding to the target mRNA sequence to be predicted, including:

[0093] The node data corresponding to the molecular dynamics trajectory is obtained. The node is the residue in the target mRNA sequence to be predicted. The node data includes the spatial coordinates and velocity of the node at multiple time steps.

[0094] The node data is input into the encoder structure in the neural relational inference model, the residue features are extracted and the latent variables are generated based on the residue features.

[0095] Input the latent variables into the decoder structure in the neural relational reasoning model, reconstruct the molecular dynamics trajectory, and calculate the reconstruction error based on the reconstructed molecular dynamics trajectory and the original molecular dynamics trajectory. Back-propagate the reconstruction error to update the encoder parameters, latent variables and decoder parameters, and complete the training after multiple iterations.

[0096] The trained neural relationship inference model is obtained. The latent variable generated by the trained neural relationship inference model is the interaction relationship diagram corresponding to the target mRNA sequence to be predicted.

[0097] In this embodiment, the target mRNA sequence to be predicted is taken as an example of the EGFP target. The target mRNA is simulated by molecular dynamics using Gromacs molecular dynamics software, and the molecular dynamics trajectory is reconstructed using the neural relationship inference model to obtain the interaction relationship diagram of the target mRNA. Figure 5 The above is a schematic diagram of the target relationship inference model, which is constructed using a variational autoencoder model. The encoder of the variational autoencoder model extracts features based on the input target mRNA molecular dynamics trajectory and generates hidden variables. Then, the decoder reconstructs the dynamics trajectory of the target mRNA molecule based on these hidden variables. Subsequently, by analyzing the reconstructed trajectory, an interaction relationship diagram between the target mRNA molecule residues is generated.

[0098] Specifically, the input target mRNA dynamic trajectory consists of N nodes, and the feature vector of each node consists of the position and velocity in the x, y, and s dimensions. represents the feature of node i at time t, so the features of all nodes at time t can be expressed as The trajectory of node i can be expressed as Where T is the number of time steps. The neural relational inference model learns the interaction graph of target mRNA and reconstructs the future trajectory of the target mRNA dynamic system based on the interaction graph in an unsupervised manner. The interaction between node i and node j is in the form of a hidden variable z i,j ∈{1,2,…,K}, where K is the number of interaction types modeled, and in this embodiment, K = 2. These interaction types do not have any predefined meanings, but the model learns to assign meanings to each type.

[0099] The target relationship inference model is constructed by a variational autoencoder model, which maximizes the evidence lower bound (ELBO) through the following formula:

[0100]

[0101] Among them, Φ and θ are trainable parameters, q Φ (z|x) is the encoder return value for the hidden variable z, p θ(x|z) is a decoder that is used to reconstruct the dynamical system given a set of latent variables z. θ (z) is the prior probability distribution, which is the uniform independent classification distribution of each relationship variable.

[0102] Furthermore, the encoder is as follows:

[0103] q Φ (z ij |x)=softmax(f enc,Φ (x) ij,1:K )

[0104] Among them, f enc,Φ (x) is a graph neural network without self-connection, which is used to predict potential target mRNA interaction graphs. The distribution of z can be learned according to the encoder. The decoder is shown in the following formula:

[0105]

[0106] It reconstructs the dynamical system trajectory p based on the hidden variable z of the interaction relationship graph of the given potential target mRNA θ (x t+1 |x t ,…,x 1 ,z).

[0107] The interaction relationship graph of the target mRNA contains an N×N adjacency matrix, where N is the number of residues in the target mRNA, and the elements in the adjacency matrix represent the interaction scores of the corresponding two residues.

[0108] In this embodiment, the variational autoencoder structure can effectively extract the features of residues in the molecular dynamics trajectory, capture complex molecular interaction information, and effectively improve the accuracy of siRNA screening.

[0109] In some optional embodiments, the preliminary siRNA screening results are screened again based on the interaction relationship diagram to obtain the final siRNA screening results, including:

[0110] Based on the adjacency matrix included in the interaction graph, the degree of each node is determined, where the degree is the number of connections between the node and other nodes.

[0111] Generate a relationship curve graph between nodes and degrees, where the horizontal axis is the residue number and the vertical axis is the degree.

[0112] It is determined whether the residue number of the target mRNA sequence corresponding to the siRNA sequence in the preliminary siRNA screening results is within the residue number range corresponding to the trough area in the relationship curve diagram.

[0113] If it is within the residue number range corresponding to the valley area in the relationship curve graph, the corresponding siRNA sequence is retained.

[0114] If it is not within the residue number range corresponding to the trough area in the relationship curve diagram, the corresponding siRNA sequence will be deleted from the preliminary siRNA screening results.

[0115] Specifically, the N×N adjacency matrix in the target mRNA interaction relationship graph can be added row by row or column by column to obtain the degree of each node, and a curve diagram of the relationship between the node and the degree can be drawn. Figure 6 Then, the residues corresponding to the trough area of ​​the relationship curve diagram are selected to further screen the siRNAs screened by the siRNA inhibition rate prediction model. If the residue number of the target mRNA corresponding to the siRNA is in the trough area of ​​the relationship curve diagram, it is retained, otherwise it is removed, and the retained siRNA is the siRNA obtained by the final screening.

[0116] In this embodiment, the neural relationship reasoning model is combined with the siRNA inhibition rate prediction model to further improve the accuracy of siRNA screening based on the interaction relationship diagram.

[0117] In addition, in the present invention, the accuracy of EGFP target siRNA screening using only a deep learning model and a deep learning model and a neural relationship reasoning model is also compared. In the screening results, if the inhibition rate of the screened siRNA is greater than 0.7, the screening result is considered positive, otherwise it is negative. Furthermore, in order to explore the ability of the neural relationship reasoning model to be applied in the screening of various deep learning models and to improve the ability of the deep learning model to screen siRNA, the present invention also uses the AttSiOff model (a gene expression inhibition prediction model that combines attention mechanism and siRNA interference) to screen siRNA for the EGFP target. The results are shown in Table 1:

[0118] Table 1 Accuracy of screening results by different methods

[0119] Model Neural Relational Reasoning Model (None) Neural Relational Reasoning Model (with) The present invention 58.6％ 68.6％ AttSiOff 62.8％ 78.9％

[0120] As shown in Table 1, if only the siRNA inhibition rate deep learning prediction model constructed by the present invention is used to screen the siRNA of the EGFP target, the accuracy rate is 58.6%. If the neural relationship reasoning model is added for further screening, the accuracy rate can be improved by 10%. In the screening of the AttSiOff model, the accuracy rate is 62.8%, but on this basis, the neural relationship reasoning model is added for further screening, and the accuracy rate can reach 78.9%, an increase of 16.1%.

[0121] It can be seen that the screening method based on the neural relationship reasoning model provided by the present invention can improve the accuracy of deep learning model screening siRNA; the screening method can also be used for a variety of deep learning models, not limited to the deep learning model provided in this embodiment.

[0122] In this embodiment, a siRNA screening device based on deep learning is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0123] This embodiment provides a siRNA screening device based on deep learning, such as Figure 7 As shown, the device comprises:

[0124] Construction module 201, used to construct a siRNA inhibition rate prediction model;

[0125] The siRNA inhibition rate prediction module 202 is used to obtain the preliminary siRNA screening results corresponding to the target mRNA sequence to be predicted based on the siRNA inhibition rate prediction model;

[0126] A molecular dynamics trajectory generation module 203 is used to perform molecular dynamics simulation on the target mRNA sequence to be predicted, and generate a molecular dynamics trajectory corresponding to the target mRNA sequence to be predicted;

[0127] A relationship graph generation module 204 is used to construct a neural relationship reasoning model to generate an interaction relationship graph corresponding to the target mRNA sequence to be predicted;

[0128] The screening module 205 is used to re-screen the preliminary siRNA screening results based on the interaction relationship diagram to obtain the final siRNA screening results.

[0129] In some optional embodiments, the device further comprises:

[0130] A construction module is used to obtain a siRNA data set, wherein the siRNA data set includes: a siRNA sequence, a target mRNA sequence corresponding to the siRNA sequence, and an inhibition rate of the siRNA sequence to the corresponding target mRNA sequence;

[0131] Thermodynamic property characteristics and expert experience characteristics corresponding to the siRNA sequence are obtained; the siRNA sequence in the siRNA data set is spliced ​​with the local target mRNA sequence corresponding to the siRNA sequence, and the corresponding position code is added to the spliced ​​sequence to obtain a coding spliced ​​sequence; wherein the local target mRNA sequence is a sequence matching the siRNA sequence plus a preset number of residue sequences before and after the match with the siRNA sequence; the coding spliced ​​sequence is input into a pre-constructed feature network structure for feature extraction to obtain sequence features; based on the inhibition rate of the siRNA sequence to the corresponding target mRNA sequence, sequence features, siRNA thermodynamic property characteristics and expert experience features, a preset fully connected neural network is trained to obtain a siRNA inhibition rate prediction model after the training is completed.

[0132] In some optional embodiments, the siRNA inhibition rate prediction module 202 includes:

[0133] Construct the antisense sequence of the siRNA sequence of the target mRNA sequence to be predicted; input the antisense sequence of the siRNA sequence into the siRNA inhibition rate prediction model to obtain the siRNA predicted inhibition rate corresponding to the siRNA sequence; sort the siRNA predicted inhibition rate from high to low; starting from the highest siRNA predicted inhibition rate, select the siRNA sequences with a preset ratio as the preliminary siRNA screening results. Among them, starting from the 5' end of the local target mRNA sequence, move to the 3' end in turn until the end of the antisense sequence of the siRNA sequence is flush with the 3' end of the target mRNA; the length of the antisense sequence of the siRNA sequence ranges from 19nt to 21nt.

[0134] In some optional implementations, the relationship diagram generation module 204 includes:

[0135] A generation unit is used to obtain node data corresponding to the molecular dynamics trajectory, where the node is the residue in the target mRNA sequence to be predicted, and the node data includes the spatial coordinates and velocity of the node at multiple time steps; the node data is input into the encoder structure in the neural relational reasoning model, the residue features are extracted and hidden variables are generated based on the residue features; the hidden variables are input into the decoder structure in the neural relational reasoning model to generate an interaction relationship diagram.

[0136] In some optional implementations, the screening module 205 includes:

[0137] The screening unit is used to determine the degree of each node based on the adjacency matrix included in the interaction relationship graph, wherein the degree is the number of nodes connected to other nodes; generate a relationship curve graph between the node and the degree, wherein the abscissa of the relationship curve graph is the residue number and the ordinate is the degree; determine whether the residue number of the target mRNA sequence corresponding to the siRNA sequence in the preliminary siRNA screening result is within the residue number range corresponding to the trough area in the relationship curve graph; if it is within the residue number range corresponding to the trough area in the relationship curve graph, retain the corresponding siRNA sequence; if it is not within the residue number range corresponding to the trough area in the relationship curve graph, delete the corresponding siRNA sequence from the preliminary siRNA screening result.

[0138] The deep learning-based siRNA screening device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0139] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0140] The embodiment of the present invention also provides a computer device having the above Figure 7 The deep learning-based siRNA screening device shown.

[0141] See also Figure 8 , Figure 8 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 8 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 A processor 10 is taken as an example.

[0142] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0143] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0144] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0145] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0146] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0147] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0148] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0149] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A siRNA screening method based on deep learning, characterized in that: The method comprises: Construct a siRNA inhibition rate prediction model; Based on the siRNA inhibition rate prediction model, a preliminary siRNA screening result corresponding to the target mRNA sequence to be predicted is obtained; Performing molecular dynamics simulation on the target mRNA sequence to be predicted to generate a molecular dynamics trajectory corresponding to the target mRNA sequence to be predicted; Constructing a neural relationship inference model to generate an interaction relationship diagram corresponding to the target mRNA sequence to be predicted; Based on the interaction relationship diagram, the preliminary siRNA screening results are screened again to obtain the final siRNA screening results.

2. The method according to claim 1, characterized in that The construction of the siRNA inhibition rate prediction model comprises: Acquire a siRNA data set, wherein the siRNA data set includes: a siRNA sequence, a target mRNA sequence corresponding to the siRNA sequence, and an inhibition rate of the siRNA sequence to the corresponding target mRNA sequence; Obtain the thermodynamic property characteristics and expert experience characteristics corresponding to the siRNA sequence; Splicing the siRNA sequence in the siRNA data set with the local target mRNA sequence corresponding to the siRNA sequence, and adding the corresponding position code to the spliced ​​sequence to obtain a coding spliced ​​sequence; wherein the local target mRNA sequence is a sequence matching the siRNA sequence plus a preset number of residue sequences before and after the siRNA sequence; Inputting the coding splicing sequence into a pre-constructed feature network structure for feature extraction to obtain sequence features; A preset fully connected neural network is trained based on the inhibition rate of the siRNA sequence on the corresponding target mRNA sequence, the sequence characteristics, the thermodynamic property characteristics and the expert experience characteristics to obtain the siRNA inhibition rate prediction model after the training is completed.

3. The method according to claim 1, characterized in that The method of obtaining a preliminary siRNA screening result corresponding to the target mRNA sequence to be predicted based on the siRNA inhibition rate prediction model comprises: Constructing the antisense sequence of the siRNA sequence of the target mRNA sequence to be predicted; Inputting the antisense sequence of the siRNA sequence into the siRNA inhibition rate prediction model to obtain the siRNA predicted inhibition rate corresponding to the siRNA sequence one by one; The predicted inhibition rates of the siRNAs are sorted from high to low; Starting from the highest siRNA predicted inhibition rate, siRNA sequences with a preset ratio are selected in sequence as the preliminary siRNA screening results.

4. The method according to claim 3, characterized in that: The method of constructing the antisense sequence of the siRNA sequence of the target mRNA sequence to be predicted comprises: Starting from the 5' end of the target mRNA sequence, the sequence is translated toward the 3' end in sequence until the end of the antisense sequence of the siRNA sequence is flush with the 3' end of the target mRNA; the length of the antisense sequence of the siRNA sequence ranges from 19nt to 21nt.

5. The method according to claim 1, characterized in that The neural relationship inference model is constructed based on a variational autoencoder model, and the neural relationship inference model is constructed to generate an interaction relationship graph corresponding to the target mRNA sequence to be predicted, including: Obtaining node data corresponding to the molecular dynamics trajectory, wherein the node is a residue in the target mRNA sequence to be predicted, and the node data includes spatial coordinates and velocities of the node at multiple time steps; Inputting the node data into an encoder structure in the neural relational reasoning model, extracting residue features and generating latent variables based on the residue features; Input the latent variables into the decoder structure in the neural relational reasoning model, reconstruct the molecular dynamics trajectory, calculate the reconstruction error based on the reconstructed molecular dynamics trajectory and the original molecular dynamics trajectory, back-propagate the reconstruction error to update the encoder parameters, latent variables and decoder parameters, and complete the training after multiple iterations; The trained neural relationship inference model is obtained, and the latent variable generated by the trained neural relationship inference model is the interaction relationship diagram corresponding to the target mRNA sequence to be predicted.

6. The method according to claim 5, characterized in that The method of rescreening the preliminary siRNA screening results based on the interaction relationship diagram to obtain the final siRNA screening results comprises: Determine the degree of each node based on the adjacency matrix included in the interaction relationship graph, wherein the degree is the number of other nodes connected to the node; generating a relationship curve graph between the node and the degree, wherein the abscissa of the relationship curve graph is the residue number and the ordinate is the degree; Determine whether the residue number of the target mRNA sequence corresponding to the siRNA sequence in the preliminary siRNA screening result is within the residue number range corresponding to the trough area in the relationship curve graph; If it is within the residue number range corresponding to the valley area in the relationship curve diagram, the corresponding siRNA sequence is retained; If it is not within the residue number range corresponding to the valley area in the relationship curve diagram, the corresponding siRNA sequence is deleted from the preliminary siRNA screening results.

7. A siRNA screening device based on deep learning, characterized in that: The device comprises: A construction module for constructing a siRNA inhibition rate prediction model; An siRNA inhibition rate prediction module, used to obtain preliminary siRNA screening results corresponding to the target mRNA sequence to be predicted based on the siRNA inhibition rate prediction model; A molecular dynamics trajectory generation module is used to perform molecular dynamics simulation on the target mRNA sequence to be predicted, and generate a molecular dynamics trajectory corresponding to the target mRNA sequence to be predicted; A relationship graph generation module is used to construct a neural relationship reasoning model to generate an interaction relationship graph corresponding to the target mRNA sequence to be predicted; The screening module is used to re-screen the preliminary siRNA screening results based on the interaction relationship diagram to obtain the final siRNA screening results.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the deep learning-based siRNA screening method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the deep learning-based siRNA screening method described in any one of claims 1-6.

10. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the deep learning-based siRNA screening method according to any one of claims 1 to 6.

Images