Target determination method and device, and computer equipment

Abstract

The present disclosure provides a target determination method, device, and computer equipment. The method includes: obtaining a messenger RNA (mRNA) target sequence, n1 features of each point in the mRNA target sequence, the distance between any two points, and the salinity and humidity that each point can bind to, where n1 is a positive integer; inputting the mRNA target sequence, n1 features of each point in the mRNA target sequence, the distance between any two points, and the salinity and humidity that each point can bind to into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence. The secondary structure model is used to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, and each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure; selecting one or more optimal paths from the secondary structure matrix based on the optimal path search algorithm, and determining all candidate target sequences based on the optimal path. This method does not require biological experiments for each site, is time-consuming, low-cost, and has good results.

Description

Technical Field

[0001] The present disclosure relates to the field of biotechnology, and in particular to a target determination method and device, and a computer device. Background Art

[0002] Ribonucleic acid (RNA)-based therapies require the identification of drug RNA molecules and targeted RNA small molecules. Various forms of RNA can be used to selectively act on proteins, transcripts, and genes that cannot be targeted by conventional small molecules or proteins. RNA-targeted therapies include small interfering RNA (siRNA) and antisense oligonucleotides (ASO). Although these two types of drugs have high specificity, they still face significant delivery challenges outside the limited range of target organs.

[0003] Therefore, selecting RNA targets has become a challenge for current RNA therapeutics. Related technologies typically use biological experiments to determine whether individual sites on an RNA sequence can serve as targets. However, this experimental approach is time-consuming and costly. Summary of the Invention

[0004] The present disclosure provides a target determination method and apparatus, as well as a computer device, that can determine targets by processing RNA data using a computer, thereby reducing time and cost. The technical solution is as follows:

[0005] At least one embodiment of the present disclosure provides a target determination method, the method comprising:

[0006] Obtaining a messenger RNA (mRNA) target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound by each site, where n1 is a positive integer;

[0007] Inputting the mRNA target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence, wherein the secondary structure model is used to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, n1 features of each site in the mRNA sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site, wherein each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure;

[0008] One or more optimal paths are selected from the secondary structure matrix based on an optimal path search algorithm, and all candidate target sequences are determined based on the optimal paths.

[0009] Optionally, the method further includes:

[0010] Obtain a training set and a test set, where each sample in the training set and the test set includes an mRNA sequence, n1 features of the i-th site in the mRNA sequence, the distance between the i-th site and the j-th site, and the salinity and humidity that the i-th site can bind to. The label of the sample is Pij in the secondary structure matrix, where Pij represents the binding probability between the i-th site and the j-th site, and i and j are positive integers.

[0011] The training set and the test set are used to iteratively train the neural network model to obtain the secondary structure model.

[0012] Optionally, the method further comprises: determining the secondary structure matrix corresponding to the mRNA sequences involved in the training set according to the following steps:

[0013] Obtain mRNA sequence and next-generation sequencing NGS data, the length of the mRNA sequence is N, the NGS data is a 1×N sequence matrix, each element of the sequence matrix corresponds to a site in the mRNA sequence, and each element of the sequence matrix indicates that the corresponding site is N in the NGS experiment. 20 the number of times the segment is heterozygous;

[0014] Generate an n1×N sequence feature matrix based on the mRNA sequence, wherein each column element of the sequence feature matrix corresponds to a site of the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site;

[0015] Superimposing the sequence matrix and the sequence feature matrix to obtain a (n1+1)×N primary structure matrix;

[0016] The primary structure matrix is ​​reduced in dimension into a 1×N matrix, and the 1×N matrix is ​​diagonalized to obtain an N×N secondary structure matrix.

[0017] Optionally, the features include at least one of the following:

[0018] Palindrome structure, sequence polymorphism, repetitive sequence, translation region start point, coding region.

[0019] Optionally, the dimensionality reduction processing of the primary structure matrix into a 1×N matrix and diagonalizing the 1×N matrix to obtain an N×N secondary structure matrix includes:

[0020] Adding the elements of each column of the primary structure matrix to obtain the 1×N matrix;

[0021] Calculate each element in the secondary structure matrix according to the formula Pij=Pi×Pj, and fill the calculation results into the secondary structure matrix; wherein Pij is the jth element in the i-th row of the secondary structure matrix, Pi is the i-th element in the 1×N matrix, Pj is the j-th element in the 1×N matrix, and i and j are both positive integers in [1, N].

[0022] Optionally, the selecting one or more optimal paths from the secondary structure matrix based on the optimal path search algorithm, and determining all candidate target sequences based on the optimal paths, comprises:

[0023] Based on n2 selection conditions, n2 N×N selection condition matrices are obtained, where each element of the selection condition matrix represents the score of the corresponding secondary structure under the corresponding condition, and N, n1, and n2 are all positive integers;

[0024] Reducing the dimension of the n2 N×N selection condition matrices into an N×N matrix, and adding them to the N×N secondary structure matrix to obtain an N×N path scoring matrix;

[0025] One or more optimal paths are selected from the path scoring matrix based on an optimal path search algorithm, and all candidate target sequences are determined based on the optimal paths.

[0026] Optionally, the selection condition includes at least one of the following:

[0027] Melting temperature, binding energy, and difficulty of chemical synthesis.

[0028] Optionally, before performing dimensionality reduction processing on the primary structure matrix, the method further includes:

[0029] Each element in the primary structure matrix is ​​normalized so that each element is within the range of [0, 1].

[0030] At least one embodiment of the present disclosure provides a target point determination device, the device comprising:

[0031] An acquisition module is used to obtain a messenger RNA (mRNA) target sequence, n1 features of each point in the mRNA target sequence, the distance between any two points, and the salinity and humidity that can be bound to each point, where n1 is a positive integer;

[0032] A secondary structure matrix module is used to input the mRNA target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence. The secondary structure model is used to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, n1 features of each site in the mRNA sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site, wherein each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure;

[0033] A path search module is used to select one or more optimal paths from the secondary structure matrix based on an optimal path search algorithm, and determine all candidate target sequences based on the optimal paths.

[0034] Optionally, the device further comprises:

[0035] A training data acquisition module is used to obtain a training set and a test set, wherein each sample in the training set and the test set includes an mRNA sequence, n1 features of the i-th site in the mRNA sequence, the distance between the i-th site and the j-th site, and the salinity and humidity that the i-th site can bind to. The label of the sample is Pij in the secondary structure matrix, where Pij represents the binding probability between the i-th site and the j-th site, and i and j are positive integers.

[0036] The training module is used to iteratively train the neural network model using the training set and the test set to obtain the secondary structure model.

[0037] Optionally, the training data acquisition module is used to determine the secondary structure matrix corresponding to the mRNA sequence involved in the training set in the following manner:

[0038] Obtain mRNA sequence and next-generation sequencing NGS data, the length of the mRNA sequence is N, the NGS data is a 1×N sequence matrix, each element of the sequence matrix corresponds to a site in the mRNA sequence, and each element of the sequence matrix indicates that the corresponding site is N in the NGS experiment. 20 the number of times the segment is heterozygous;

[0039] Generate an n1×N sequence feature matrix based on the mRNA sequence, wherein each column element of the sequence feature matrix corresponds to a site of the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site;

[0040] Superimposing the sequence matrix and the sequence feature matrix to obtain a (n1+1)×N primary structure matrix;

[0041] The primary structure matrix is ​​reduced in dimension into a 1×N matrix, and the 1×N matrix is ​​diagonalized to obtain an N×N secondary structure matrix.

[0042] Optionally, the features include at least one of the following:

[0043] Palindrome structure, sequence polymorphism, repetitive sequence, translation region start point, coding region.

[0044] Optionally, the training data acquisition module is used to add the elements of each column of the primary structure matrix to obtain the 1×N matrix; calculate each element in the secondary structure matrix according to the formula Pij=Pi×Pj, and fill the calculation results into the secondary structure matrix; wherein Pij is the jth element in the i-th row of the secondary structure matrix, Pi is the i-th element in the 1×N matrix, Pj is the j-th element in the 1×N matrix, and i and j are both positive integers in [1, N].

[0045] Optionally, the path search module is configured to obtain n2 N×N selection condition matrices based on n2 selection conditions, wherein each element of the selection condition matrix represents a score of a corresponding secondary structure under a corresponding condition, and N, n1, and n2 are all positive integers;

[0046] Reducing the dimension of the n2 N×N selection condition matrices into an N×N matrix, and adding them to the N×N secondary structure matrix to obtain an N×N path scoring matrix;

[0047] One or more optimal paths are selected from the path scoring matrix based on an optimal path search algorithm, and all candidate target sequences are determined based on the optimal paths.

[0048] Optionally, the selection condition includes at least one of the following:

[0049] Melting temperature, binding energy, and difficulty of chemical synthesis.

[0050] Optionally, the training data acquisition module is further configured to normalize each element in the primary structure matrix before performing dimensionality reduction processing on the primary structure matrix, so that each element is within the range of [0, 1].

[0051] At least one embodiment of the present disclosure provides a computer device, comprising a processor and a memory, wherein the memory stores at least one program code, and the program code is loaded and executed by the processor to implement the target determination method as described above.

[0052] At least one embodiment of the present disclosure provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one program code, and the program code is loaded and executed by a processor to implement the target determination method as described in any of the above items.

[0053] The technical solutions provided by the embodiments of the present disclosure have the following beneficial effects:

[0054] In the disclosed embodiments, the mRNA target sequence and its corresponding features are input into a pre-trained secondary structure model to generate a secondary structure matrix, where each element represents the probability of opening the corresponding secondary structure. An optimal path search algorithm is then used to select the optimal path with the highest score from the secondary structure matrix, identifying all candidate target sequences. Furthermore, this method eliminates the need for biological experiments for each site, resulting in a shorter timeframe, lower costs, and improved results. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0056] Figure 1 is a flow chart of a target determination method provided by an embodiment of the present disclosure;

[0057] Figure 2 is a flow chart of a target determination method provided by an embodiment of the present disclosure;

[0058] Figure 3 This is a schematic diagram of a dimensionality reduction process provided by an embodiment of the present disclosure;

[0059] Figure 4 This is a schematic diagram of a path search provided by an embodiment of the present disclosure;

[0060] Figure 5 is a flow chart of a method for obtaining a secondary structure matrix provided by an embodiment of the present disclosure;

[0061] Figure 6 This is a schematic diagram of matrix superposition provided by an embodiment of the present disclosure;

[0062] Figure 7 Schematic diagram of the structure of a target determination device provided by an embodiment of the present disclosure;

[0063] Figure 8 This is a structural block diagram of a computer device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0064] In order to make the objectives, technical solutions and advantages of the present disclosure more clear, the embodiments of the present disclosure will be further described in detail below with reference to the accompanying drawings.

[0065] Unless otherwise defined, the technical or scientific terms used herein shall have the usual meanings understood by persons of ordinary skill in the field to which the present disclosure belongs. The words “first”, “second”, “third” and similar terms used in the patent application specification and claims of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as “one” or “a” do not indicate a quantity limitation, but rather indicate the presence of at least one. Words such as “include” or “comprising” and similar terms mean that the elements or objects appearing before “include” or “comprising” cover the elements or objects listed after “include” or “comprising” and their equivalents, and do not exclude other elements or objects.

[0066] Figure 1 This is a flow chart of a target determination method provided by an embodiment of the present disclosure. Figure 1 , the method comprising:

[0067] 101: Obtain a messenger RNA (mRNA) target sequence, n1 features of each point in the mRNA target sequence, a distance between any two points, and a salinity and humidity that can be bound to each point, where n1 is a positive integer.

[0068] Among them, the mRNA target sequence is the sequence of the mRNA of the genome to be selected, that is, the sequence where the potential target position is located. The mRNA target sequence generally comes from a genome database.

[0069] 102: Inputting the mRNA target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence. The secondary structure model is used to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, n1 features of each site in the mRNA sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site. Each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure.

[0070] 103: Select one or more optimal paths from the secondary structure matrix based on an optimal path search algorithm, and determine all candidate target sequences based on the optimal paths.

[0071] In the disclosed embodiments, the mRNA target sequence and its corresponding features are input into a pre-trained secondary structure model to generate a secondary structure matrix, where each element represents the probability of opening the corresponding secondary structure. An optimal path search algorithm is then used to select the optimal path with the highest score from the secondary structure matrix, identifying all candidate target sequences. Furthermore, this method eliminates the need for biological experiments for each site, resulting in a shorter timeframe, lower costs, and improved results.

[0072] Figure 2 This is a flow chart of a target determination method provided by an embodiment of the present disclosure. Figure 2 , the method comprising:

[0073] 201: Get the training set and test set.

[0074] Each sample in the training set and the test set includes an mRNA sequence, n1 features of the i-th site in the mRNA sequence, the distance between the i-th site and the j-th site, and the salinity and humidity that the i-th site can bind to. The label of the sample is Pij in the secondary structure matrix, where Pij represents the binding probability between the i-th site and the j-th site, and i and j are positive integers.

[0075] Among them, i and j are variables, ranging from 1 to N, where N is the length of the mRNA sequence. That is, each sample includes the characteristics of each site, the distance between any two points, and the salinity and humidity of each site.

[0076] Among them, salinity and humidity data are derived from DNA / RNA hybrid experimental data and related databases.

[0077] Among them, the characteristics, distance, salinity, humidity, etc. of each site have been characterized as data that can be used for model training.

[0078] Since in this step, elements in the secondary structure matrix need to be used as labels, it is necessary to determine the secondary structure matrix corresponding to each sample.

[0079] In the embodiment of the present disclosure, the secondary structure matrix corresponding to the mRNA sequences involved in the training set and the test set can be determined as follows:

[0080] The first step is to obtain mRNA sequences and next-generation sequencing (NGS) data.

[0081] Wherein, the length of the mRNA sequence is N, and N is a positive integer.

[0082] The NGS data is a 1×N sequence matrix, each element of the sequence matrix corresponds to a site (base) in the mRNA sequence, and each element of the sequence matrix indicates that the corresponding site is N in the NGS experiment. 20 The number of times the fragment is hybridized.

[0083] N 20 Refers to multiple random nucleic acid sequences, each 20 nt in length. NGS experiments involve performing a hybridization experiment on all random nucleic acid sequences with mRNA sequences. After hybridization, the experimental product is hydrolyzed with an enzyme that hydrolyzes single-stranded RNA, so that the sites in the mRNA sequence that have not been hybridized are hydrolyzed, while the sites that have successfully hybridized are retained. An NGS instrument (such as an Illumina sequencer) can then record which sites in the mRNA sequence are hybridized (combined) and the number of times each site has been hybridized. The recorded data can then be used to generate NGS data.

[0084] In addition, NGS experiments are bioinformatics-related experiments. Compared with related art experiments that determine whether each site on an RNA sequence can be used as a target (such as gene walking experiments, which have extremely low data throughput (each time a target is searched, only a fixed and limited length of gene fragment can be searched), NGS experiments can test the entire genome, and the data throughput obtained is extremely large. The possible presence of target binding at a large number of potential sites can be fed back into the data at one time, providing data support for target selection. It can be seen that the NGS experiment used in the present disclosure does not require biological experiments for each site, is time-consuming, and has low cost.

[0085] The second step is to generate an n1×N sequence feature matrix based on the mRNA sequence, where each column element of the sequence feature matrix corresponds to a site in the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site.

[0086] Each column element of the sequence feature matrix corresponds to a site in the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site, where n1 is a positive integer. Each element is a numerical value obtained by digitizing a feature of the corresponding site.

[0087] Exemplarily, the features include at least one of the following:

[0088] Palindrome structure, sequence polymorphism, repetitive sequence, translation region start point, coding region.

[0089] A palindrome is a fragment of an mRNA sequence (including sites) where the sequence read from 5' to 3' on one strand is identical to the sequence read from the same 5' to 3' direction on its complementary strand. The 5' and 3' positions represent the endpoints of the RNA carbon ring (position 5 and position 3) of a nucleic acid sequence / gene sequence, and are also the starting and ending points of a gene sequence.

[0090] Sequence polymorphism refers to the location of a site in a gene, which is divided into gene coding regions, gene non-coding regions, gene intergenic regions (regions between genes), etc.

[0091] A repetitive sequence refers to a sequence in which each base is repeated, such as AAATTTGGGCCC.

[0092] The translation start point refers to the region outside the untranslated region (UTR). In molecular genetics, the untranslated region (UTR) refers to any region located at either end of the coding sequence of an mRNA strand. If it is located at the 5' end, it is called the 5' untranslated region (or "leader sequence"). If it is located at the 3' end, it is called the 3' untranslated region (or "trailer sequence").

[0093] The coding region refers to the part of RNA that encodes proteins and is composed of exons.

[0094] In the embodiment of the present disclosure, the above features can be quantified in the following manner to form a sequence feature matrix:

[0095] Quantify each feature based on whether it promotes heterozygosity (complementary base pairing). For example, for a site:

[0096] The numerical value of the palindrome structure feature is as follows: the value corresponding to the absence of the palindrome structure is greater than the value corresponding to the presence of the palindrome structure;

[0097] The numerical value of sequence polymorphism characteristics is as follows: the value corresponding to the gene interval region is greater than the value corresponding to the gene coding region and the gene non-coding region.

[0098] The numerical value of the repeated sequence feature is as follows: the value corresponding to the absence of repeated sequences is greater than the value corresponding to the presence of repeated sequences;

[0099] The numerical value of the starting point feature of the translation region is as follows: the value corresponding to the absence of the untranslated region is greater than the value corresponding to the presence of the untranslated region;

[0100] The numerical value of the coding region feature is as follows: the value corresponding to the absence of the coding region is greater than the value corresponding to the presence of the coding region.

[0101] When quantifying each of the above features, a larger and smaller value are assigned based on the feature category. A smaller value indicates a lower binding probability, meaning it is less likely to be a potential target. Conversely, a larger value indicates a higher binding probability, meaning it is more likely to be a potential target. For example, the larger value can be 1, and the smaller value can be 0. Alternatively, any other set of values ​​with a significant relationship can be used, without limitation.

[0102] In the embodiments disclosed herein, NGS data can reflect the binding probability of sites in the sequence, but it is only a necessary but not sufficient probability and is not sufficient as an absolutely reliable probability for screening targets. By combining the characteristics of each site, the binding probability of each site can be better explained, making the selection of targets more accurate.

[0103] In a possible implementation, the step of generating an n1×N sequence feature matrix from the mRNA sequence can be automatically performed by a computer using an existing algorithm.

[0104] In another possible implementation, the step of generating an n1×N sequence feature matrix from the mRNA sequence can be completed by computer and manual interaction. For example, the computer provides an input table for each site in turn, and the quantitative value corresponding to the feature of each site is manually input.

[0105] The third step is to perform matrix superposition on the sequence matrix and the sequence feature matrix to obtain a (n1+1)×N primary structure matrix.

[0106] The superposition here means arranging the elements in the sequence matrix and the elements in the sequence feature matrix corresponding to the same site into a row.

[0107] The fourth step is to reduce the dimension of the primary structure matrix into a 1×N matrix, and diagonalize the 1×N matrix to obtain an N×N secondary structure matrix.

[0108] Among them, each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure, that is, the binding probability of two sites in the primary structure matrix, that is, the probability of whether a target can be formed between the two sites. The opening probability is also the possibility of the base being hybridized into a double strand, representing the probability of forming an ASO at this base position (site). The secondary structure is also the structure formed by the combination of bases at two sites.

[0109] The samples used in this scheme use the elements in the secondary structure matrix as labels. The process of generating the secondary structure matrix not only considers the data of the NGS experiment, but also considers the characteristics of each site, so that the binding rate of the selected sites is high. Ultimately, the secondary structure matrix generated by the trained model is used to select sites with a high binding rate.

[0110] It is worth noting that, before the embodiment of the present disclosure, the target sequence selection can be achieved by combining the first to fourth steps with step 204. When the process generates enough secondary structure matrices to be used as samples, the model is retrained and subsequently used to determine the secondary structure matrix, thereby simplifying the secondary structure matrix generation process and improving efficiency.

[0111] 202: Iteratively train the neural network model using the training set and the test set to obtain the secondary structure model.

[0112] Among them, the secondary structure model is used to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, n1 features of each point in the mRNA sequence, the distance between any two points, and the salinity and humidity that can be bound to each point. Each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure.

[0113] In the embodiment of the present disclosure, the above-mentioned neural network model can be a convolutional neural network (CNN) or a deep neural network (DNN).

[0114] The VGG16 neural network is used as an example, representing a deep neural network model for model training and prediction. VGG16 contains 13 convolutional layers and 3 fully connected layers. These 13 convolutional layers progressively extract the characteristic features of each dimension of the secondary structure matrix, until the rules are extracted and filtered. In other words, during training, VGG16 automatically extracts the various distributed feature representations required for the task. The subsequent fully connected layers map the learned "distributed feature representations" to the sample label space. VGG16 uses the Rectified Linear Unit (ReLU) activation function and the categorical crossentropy loss function.

[0115] During training, multiple rounds of iterative learning and training of training set-test set-validation set are performed, and the model is output and saved.

[0116] Each round of iterative training involves optimizing multiple loss functions using mini-batch stochastic gradient descent with momentum, using backpropagation. Weight decay is used for regularization, with dropout (p = 0.5) applied to the first two fully connected layers. The learning rate is 0.01, decaying by a factor of 10 when the validation set stops improving.

[0117] The samples are randomly divided into training and test sets using an 80%-20% method. Each round of model iteration outputs a trend curve of the loss function. Model parameters are adjusted based on changes in the loss function, and further model iterations are performed.

[0118] 203: Input the mRNA target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence.

[0119] The mRNA target sequence is used as the input of the secondary structure model, and the secondary structure model outputs the corresponding secondary structure matrix.

[0120] 204: Based on the n2 selection conditions, obtain n2 N×N selection condition matrices.

[0121] Each element of the selection condition matrix represents the score of the corresponding secondary structure under the corresponding condition.

[0122] Each element of the selection condition matrix represents the score of the corresponding secondary structure under the corresponding condition, and n2 is a positive integer. Each N×N selection condition matrix here can be obtained according to the method of step 104, that is, a primary condition matrix for each condition is obtained, that is, a 1×N primary condition matrix, each element of which is the score of the selection condition corresponding to the site. The 1×N primary condition matrix is ​​diagonalized to obtain an N×N selection condition matrix, each element of which represents the score of the combination of two sites in the primary condition matrix, which serves as the judgment condition for whether a target can be formed.

[0123] The purpose of this step is to consider not only the probability of opening the target when selecting the target, but also the physicochemical and sequence characteristics of the target. For example, based on the selection conditions, targets that are easy to synthesize and have stable properties can be selected as much as possible.

[0124] Exemplarily, the selection condition includes at least one of the following:

[0125] Melting temperature, binding energy, and difficulty of chemical synthesis.

[0126] Melting temperature refers to the temperature at which absorbance reaches half of its maximum value. Melting temperatures for individual sites can be obtained using existing tools. Generally, within a normal range, higher melting temperatures correspond to higher scores, while lower melting temperatures correspond to lower scores.

[0127] Binding energy refers to the energy released when two or more particles combine to form a larger particle, or the energy required to dissociate the corresponding particles into their original form. Binding energy at each site can be calculated using existing tools. Generally, within a normal range, higher binding energies correspond to higher scores, while lower binding energies correspond to lower scores.

[0128] Chemical synthesis difficulty refers to the difficulty of synthesis in a synthesis experiment. Chemical synthesis difficulty can be scored based on experience, for example, the lower the difficulty, the higher the score.

[0129] Exemplarily, each value in the condition matrix may be a negative value, that is, each site is penalized by the above conditions, so that the site selection takes the above conditions into consideration.

[0130] 205: Reduce the dimension of the n2 N×N selection condition matrices into an N×N matrix, and add them to the N×N secondary structure matrix to obtain an N×N path scoring matrix.

[0131] Here, matrix addition refers to adding elements in the same position. The addition can be performed in a weighted manner. For example, in the initial case, the N×N matrix A and the N×N secondary structure matrix B are added in a 1:1 ratio. Of course, this is only an example and the present disclosure is not limited to this.

[0132] Figure 3 This is a schematic diagram of a dimensionality reduction process provided by an embodiment of the present disclosure. Figure 3 , add the elements at the same position of each selection condition matrix, and the addition can be performed in a weighted manner. Initially, the weights of each element can be the same.

[0133] 206: Select one or more optimal paths from the secondary structure matrix based on an optimal path search algorithm, and determine all candidate target sequences based on the optimal paths.

[0134] The optimal path search algorithm is used to select a path with the highest score. The combination of the two sites (bases) corresponding to each point in the path can be used as a target point. The path corresponds to a sequence with one site of the mRNA sequence as the starting point and another site as the end point. The base sequence of the selected fragment of the sequence "can" complementarily pair with the targeted drug, and this sequence is the candidate target sequence.

[0135] In combination with the application, the site of the mRNA sequence is the selected point to be treated, and the targeted drug is bound to it.

[0136] For example, the point with the largest value in the secondary structure matrix is ​​taken as the starting point, the point with the highest value adjacent to the starting point is selected as the second point, and the point with the highest value adjacent to the second point is selected as the third point, and so on. The points selected each time are not repeated, so that a path is obtained, corresponding to a candidate target sequence.

[0137] Since there may be points with the same value at the same time in the process of selecting the next point in each step, different paths will be generated at this time. If the sum of the values ​​of the generated different paths is the same or the difference is less than the threshold, multiple paths may be obtained.

[0138] For example, when, among the selected paths, the scores of multiple paths with higher scores have a small difference, these multiple paths may be selected as the optimal paths.

[0139] Figure 4 This is a schematic diagram of a path search provided by an embodiment of the present disclosure. Figure 4 , based on the optimal path search algorithm to find the path with the highest score, T, G, A, and C in this matrix refer to bases, that is, the sequence of bases. Figure 4 As shown, the base sequence in the rows and columns of this matrix is ​​the same, obtained by diagonalizing a 1×N sequence matrix. The optimal path search algorithm does not necessarily select every column. Starting with the point with the largest value in the secondary structure matrix, the point with the highest value adjacent to the starting point is selected as the second point. The point with the highest value adjacent to the second point is selected as the third point, and so on. The selected points are unique each time, resulting in a path. This path corresponds to a sequence starting from one site in the mRNA sequence and ending at another site. The base sequence of the selected fragment is "capable" of complementary pairing with a targeted drug, and this sequence is considered a candidate target sequence. For example, the sequence from the value 13 to the value 0 in the figure is shown.

[0140] When selecting this sequence, the direction is single, and this single direction is related to the position of the starting point. A single direction can be from the starting point to the center. For example, if the starting point is in the lower right corner, the single direction is from the lower right to the upper left. For example, starting from the number 13 in the figure, selecting other points in the direction from the lower right to the upper left means selecting only upward and leftward, not downward and rightward. Therefore, when selecting the adjacent maximum value for the number 13, 10 is selected instead of 11, and the same applies to subsequent selections.

[0141] exist Figure 4 After selection, the resulting sequence is CATTGT when viewed horizontally and GCATTGT when viewed vertically. The only difference between the two is that the vertical sequence has one more site than the horizontal one. The final result can be either of the two, for example, the horizontal sequence, the vertical sequence, or the longer or shorter sequence.

[0142] This approach not only considers the binding probability of each site when selecting targets, but also the target's physicochemical and sequence characteristics. This results in selected sites with high binding rates, ease of synthesis, and stability. Furthermore, this approach eliminates the need for biological experiments for each site, resulting in shorter time, lower costs, and better results.

[0143] Figure 5 This is a flow chart of a method for obtaining a secondary structure matrix provided by an embodiment of the present disclosure. Figure 5 , the method comprising:

[0144] 301: Obtain mRNA sequence and NGS data.

[0145] The length of the mRNA sequence is N, the NGS data is a 1×N sequence matrix, each element of the sequence matrix corresponds to a site in the mRNA sequence, and each element of the sequence matrix indicates that the corresponding site is N in the NGS experiment. 20 The number of times the fragment is hybridized.

[0146] 302: Generate an n1×N sequence feature matrix based on the mRNA sequence.

[0147] Each column element of the sequence feature matrix corresponds to a site in the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site.

[0148] 303: Perform matrix superposition on the sequence matrix and the sequence feature matrix to obtain a (n1+1)×N primary structure matrix.

[0149] Figure 6 This is a schematic diagram of matrix superposition provided by an embodiment of the present disclosure. Figure 6, the 1×N sequence matrix on the left and the n1×N sequence feature matrix on the right are superimposed to obtain a (n1+1)×N primary structure matrix. After superposition, the elements of the 1×N sequence matrix are located in the first row of the (n1+1)×N primary structure matrix. Of course, this is only one possible implementation method. In other implementation methods, the elements of the 1×N sequence matrix can also be located in the second row, third row, and so on.

[0150] Figure 6 The description is made for example with n1 being 3, but the present disclosure is not limited thereto.

[0151] 304: Normalize each element in the primary structure matrix so that each element is within the range of [0, 1].

[0152] You can first determine the maximum and minimum values ​​of each element, set the maximum value to 1, and the minimum value to 0. Then, scale the other values ​​proportionally to the difference between the maximum value and the minimum value. For example, if a value has the same difference between the maximum and minimum values, you can simply scale it to 0.5.

[0153] 305: Add the elements of each column of the primary structure matrix to obtain the 1×N matrix.

[0154] by Figure 6 Taking the primary structure matrix shown in the figure as an example, the elements of each column are summed to form an element of the 1×N matrix. This summing can be performed using a weighted approach. Initially, the weights of the elements in a column can be the same. Alternatively, the weights of the elements in a column corresponding to the sequence matrix can be greater than the weights of the elements in the sequence feature matrix.

[0155] 306: Calculate each element in the secondary structure matrix according to the formula Pij=Pi×Pj, and fill the calculation results into the secondary structure matrix.

[0156] Among them, Pij is the jth element in the i-th row in the secondary structure matrix, Pi is the i-th element in the 1×N matrix, Pj is the j-th element in the 1×N matrix, i and j are both positive integers in [1, N], and each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure.

[0157] For example, the element P34 in the secondary structure matrix is ​​equal to P3×P4. That is, the fourth element in the third row of the secondary structure matrix is ​​equal to the product of the third and fourth elements in the 1×N matrix.

[0158] Figure 7 Schematic diagram of a target determination device provided by an embodiment of the present disclosure. Figure 7The device includes: an acquisition module 501, a secondary structure matrix module 502 and a path search module 503.

[0159] The acquisition module 501 is used to obtain the messenger RNA (mRNA) target sequence, n1 features of each point in the mRNA target sequence, the distance between any two points, and the salinity and humidity that can be combined with each point, where n1 is a positive integer.

[0160] A secondary structure matrix module 502 is configured to input the mRNA target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence. The secondary structure model is configured to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, the n1 features of each site in the mRNA sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site. Each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure.

[0161] The path search module 503 is configured to select one or more optimal paths from the secondary structure matrix based on an optimal path search algorithm, and determine all candidate target sequences based on the optimal paths.

[0162] Optionally, the device further comprises:

[0163] A training data acquisition module 504 is used to obtain a training set and a test set, where each sample in the training set and the test set includes an mRNA sequence, n1 features of the i-th site in the mRNA sequence, the distance between the i-th site and the j-th site, and the salinity and humidity that the i-th site can bind to. The label of the sample is Pij in the secondary structure matrix, where Pij represents the binding probability between the i-th site and the j-th site, and i and j are positive integers.

[0164] The training module 505 is used to iteratively train the neural network model using the training set and the test set to obtain the secondary structure model.

[0165] Optionally, the training data acquisition module 504 is configured to determine the secondary structure matrix corresponding to the mRNA sequence involved in the training set in the following manner:

[0166] Obtain mRNA sequence and next-generation sequencing NGS data, the length of the mRNA sequence is N, the NGS data is a 1×N sequence matrix, each element of the sequence matrix corresponds to a site in the mRNA sequence, and each element of the sequence matrix indicates that the corresponding site is N in the NGS experiment. 20 the number of times the segment is heterozygous;

[0167] Generate an n1×N sequence feature matrix based on the mRNA sequence, wherein each column element of the sequence feature matrix corresponds to a site of the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site;

[0168] Superimposing the sequence matrix and the sequence feature matrix to obtain a (n1+1)×N primary structure matrix;

[0169] The primary structure matrix is ​​reduced in dimension into a 1×N matrix, and the 1×N matrix is ​​diagonalized to obtain an N×N secondary structure matrix.

[0170] Optionally, the features include at least one of the following:

[0171] Palindrome structure, sequence polymorphism, repetitive sequence, translation region start point, coding region.

[0172] Optionally, the training data acquisition module 504 is used to add the elements of each column of the primary structure matrix to obtain the 1×N matrix; calculate each element in the secondary structure matrix according to the formula Pij=Pi×Pj, and fill the calculation results into the secondary structure matrix; wherein Pij is the jth element in the i-th row of the secondary structure matrix, Pi is the i-th element in the 1×N matrix, Pj is the j-th element in the 1×N matrix, and i and j are both positive integers in [1, N].

[0173] Optionally, the path search module 503 is configured to obtain n2 N×N selection condition matrices based on n2 selection conditions, where each element of the selection condition matrix represents a score of a corresponding secondary structure under a corresponding condition, and N, n1, and n2 are all positive integers;

[0174] Reducing the dimension of the n2 N×N selection condition matrices into an N×N matrix, and adding them to the N×N secondary structure matrix to obtain an N×N path scoring matrix;

[0175] One or more optimal paths are selected from the path scoring matrix based on an optimal path search algorithm, and all candidate target sequences are determined based on the optimal paths.

[0176] Optionally, the selection condition includes at least one of the following:

[0177] Melting temperature, binding energy, and difficulty of chemical synthesis.

[0178] Optionally, the training data acquisition module 504 is further configured to normalize each element in the primary structure matrix before performing dimensionality reduction processing on the primary structure matrix, so that each element is within the range of [0, 1].

[0179] It should be noted that the target determination device provided in the above embodiment is merely illustrated by the division of the aforementioned functional modules when selecting a target. In actual applications, the aforementioned functions can be assigned to different functional modules as needed, i.e., the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the target determination device provided in the above embodiment and the target determination method embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0180] Figure 8 6 is a block diagram of a computer device provided by an embodiment of the present disclosure. Generally, the computer device includes: a processor 601 and a memory 602.

[0181] Processor 601 may include one or more processing cores, such as a quad-core processor or an octa-core processor. Processor 601 may be implemented using at least one of the following hardware forms: a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), or a PLA (Programmable Logic Array). Processor 601 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state.

[0182] The memory 602 may include one or more computer-readable storage media, which may be non-transitory. The memory 602 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 602 is used to store at least one instruction, which is executed by the processor 601 to implement the target determination method provided in the method embodiment of the present application.

[0183] The present disclosure also provides a computer program product, which includes computer program code. When the computer program code is executed by a computer device, the computer device can execute the target determination method provided in the method embodiment of the present application.

[0184] Those skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware, or by a program to instruct the relevant hardware, and the program may be stored in a computer-readable storage medium, which may be a read-only memory, a disk, or an optical disk, etc.

[0185] The above description is merely an optional embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the scope of protection of the present disclosure.

Claims

1. A target determination method, characterized in that: The method comprises: Obtaining a messenger RNA (mRNA) target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound by each site, where n1 is a positive integer; Inputting the mRNA target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence, wherein the secondary structure model is used to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, n1 features of each site in the mRNA sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site, wherein each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure; Based on n2 selection conditions, n2 N×N selection condition matrices are obtained, where each element of the selection condition matrix represents the score of the corresponding secondary structure under the corresponding condition, and N, n1, and n2 are all positive integers; Reducing the dimension of the n2 N×N selection condition matrices into an N×N matrix, and adding them to the N×N secondary structure matrix to obtain an N×N path scoring matrix; Selecting one or more optimal paths from the path scoring matrix based on an optimal path search algorithm, and determining all candidate target sequences based on the optimal paths; The method further comprises: determining the secondary structure matrix corresponding to the mRNA sequence involved in the training set in the following manner: Obtain mRNA sequence and next-generation sequencing NGS data, the length of the mRNA sequence is N, the NGS data is a 1×N sequence matrix, each element of the sequence matrix corresponds to a site in the mRNA sequence, and each element of the sequence matrix indicates that the corresponding site is N in the NGS experiment. 20 the number of times the segment is heterozygous; Generate an n1×N sequence feature matrix based on the mRNA sequence, wherein each column element of the sequence feature matrix corresponds to a site of the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site; Superimposing the sequence matrix and the sequence feature matrix to obtain a (n1+1)×N primary structure matrix; The primary structure matrix is ​​reduced in dimension into a 1×N matrix, and the 1×N matrix is ​​diagonalized to obtain an N×N secondary structure matrix.

2. The method according to claim 1, characterized in that The method further comprises: Obtain a training set and a test set, where each sample in the training set and the test set includes an mRNA sequence, n1 features of the i-th site in the mRNA sequence, the distance between the i-th site and the j-th site, and the salinity and humidity that the i-th site can bind to. The label of the sample is Pij in the secondary structure matrix, where Pij represents the binding probability between the i-th site and the j-th site, and i and j are positive integers. The training set and the test set are used to iteratively train the neural network model to obtain the secondary structure model.

3. The method according to claim 1 or 2, characterized in that The features include at least one of the following: Palindrome structure, sequence polymorphism, repetitive sequence, translation region start point, coding region.

4. The method according to claim 1 or 2, characterized in that The dimensionality reduction process of the primary structure matrix into a 1×N matrix and diagonalizing the 1×N matrix to obtain an N×N secondary structure matrix includes: Adding the elements of each column of the primary structure matrix to obtain the 1×N matrix; Calculate each element in the secondary structure matrix according to the formula Pij=Pi×Pj, and fill the calculation results into the secondary structure matrix; wherein Pij is the jth element in the i-th row of the secondary structure matrix, Pi is the i-th element in the 1×N matrix, Pj is the j-th element in the 1×N matrix, and i and j are both positive integers in [1, N].

5. The method according to claim 1 or 2, characterized in that The selection condition includes at least one of the following: Melting temperature, binding energy, and difficulty of chemical synthesis.

6. A target determination device, characterized in that: The device comprises: An acquisition module is used to obtain a messenger RNA (mRNA) target sequence, n1 features of each point in the mRNA target sequence, the distance between any two points, and the salinity and humidity that can be bound to each point, where n1 is a positive integer; A secondary structure matrix module is used to input the mRNA target sequence, n1 features of each site in the mRNA target sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site into a secondary structure model to obtain a secondary structure matrix corresponding to the mRNA target sequence. The secondary structure model is used to determine the secondary structure matrix corresponding to the mRNA sequence based on the mRNA sequence, n1 features of each site in the mRNA sequence, the distance between any two sites, and the salinity and humidity that can be bound to each site, wherein each element of the secondary structure matrix represents the opening probability of the corresponding secondary structure; A path search module is used to obtain n2 N×N selection condition matrices based on n2 selection conditions, where each element of the selection condition matrix represents the score of the corresponding secondary structure under the corresponding condition, and N, n1, and n2 are all positive integers; Reducing the dimension of the n2 N×N selection condition matrices into an N×N matrix, and adding them to the N×N secondary structure matrix to obtain an N×N path scoring matrix; Selecting one or more optimal paths from the path scoring matrix based on an optimal path search algorithm, and determining all candidate target sequences based on the optimal paths; The training data acquisition module is used to determine the secondary structure matrix corresponding to the mRNA sequences involved in the training set in the following manner: Obtain mRNA sequence and next-generation sequencing NGS data, the length of the mRNA sequence is N, the NGS data is a 1×N sequence matrix, each element of the sequence matrix corresponds to a site in the mRNA sequence, and each element of the sequence matrix indicates that the corresponding site is N in the NGS experiment. 20 the number of times the segment is heterozygous; Generate an n1×N sequence feature matrix based on the mRNA sequence, wherein each column element of the sequence feature matrix corresponds to a site of the mRNA sequence, and each column element of the sequence feature matrix represents n1 features of the corresponding site; Superimposing the sequence matrix and the sequence feature matrix to obtain a (n1+1)×N primary structure matrix; The primary structure matrix is ​​reduced in dimension into a 1×N matrix, and the 1×N matrix is ​​diagonalized to obtain an N×N secondary structure matrix.

7. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores at least one program code, and the program code is loaded and executed by the processor to implement the target determination method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores at least one program code, and the program code is loaded and executed by a processor to implement the target determination method according to any one of claims 1 to 5.

Images