Computer system and simulation method

The computer system addresses the inaccuracy of conventional biopolymer stability predictions by region-specific analysis and machine learning, enhancing prediction accuracy through regional stability integration and transformation techniques.

JP2024162488APending Publication Date: 2024-11-21HITACHI LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2023078036
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-10
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Conventional methods for predicting the stability of biopolymers, such as mRNA, have a low correlation coefficient between predicted and actual values, leading to inaccurate stability predictions.

Method used

A computer system that classifies biopolymers into multiple regions, calculates the stability of each region, and uses a representative value to predict the overall stability by integrating region-specific decay rates, employing machine learning with Boxcox transformation to enhance accuracy.

Benefits of technology

The system achieves high-accuracy predictions of biopolymer degradation stability by considering regional variations and applying machine learning, improving correlation coefficients significantly.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024162488000001_ABST
    Figure 2024162488000001_ABST
Patent Text Reader

Abstract

To provide a system capable of predicting the decomposition stability of a biopolymer having an arbitrary array with high precision.SOLUTION: A computer is programmed to: acquire information on an array of biopolymer; classify the array into a plurality of regions; compute respective stabilities of the plurality of regions; and execute predictive simulation of the stability of the biopolymer based upon a representative value obtained by collecting the stabilities of the plurality of regions based upon the result of the computation.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a system and method for simulating the stability of a biopolymer by a computer. [Background technology]

[0002] Biopolymers exist in living organisms and are polymers formed by the polymerization of monomer units. Known biopolymers include polynucleotides, polypeptides, and polysaccharides. Biopolymers are known to have a variety of correlations between their polymerization structure and physiological activity, so when developing biopolymers as pharmaceuticals, the structures and medicinal effects of the biopolymers are often evaluated using computers. For example, it is known that the stability of biopolymers against degradation in the body can be predicted using computers.

[0003] Prior art document 1 describes how the stability of mRNA as a biological polymer is defined as its half-life, and how the half-lives of the mRNAs of a large number of genes are measured by RNA-Seq and then machine-learned to predict the half-life for mRNA of any sequence. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Medina-Munoz, Santiago Gerardo, et al. “Crosstalk between codon optimality and cis-regulatory elements dictates mRNA stability” Genome biology 22.1 (2021):1-23 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the conventional methods described as the prior art have a problem in that the correlation coefficient between the predicted value of mRNA stability and the actual measured value is small. Therefore, the present invention aims to provide a system capable of predicting the degradation stability of a biopolymer having an arbitrary sequence with high accuracy. [Means for solving the problem]

[0006] In order to achieve the above object, the present invention relates to a computer system for simulating the stability of a biopolymer, comprising a controller for executing a program stored in a memory, the controller acquiring information on the sequence of the biopolymer, classifying the sequence into a plurality of regions, calculating the stability of each of the plurality of regions, and executing a predictive simulation of the stability of the biopolymer based on a representative value summarizing the stability of each of the plurality of regions based on the result of the calculation.Furthermore, the present invention relates to a method for simulating the stability of a biopolymer. Effect of the Invention

[0007] According to the present invention, it is possible to provide a system capable of predicting with high accuracy the decomposition stability of a biopolymer having an arbitrary sequence. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram of an embodiment of a computer system for simulating the stability of a biopolymer according to the present invention. [Diagram 2] FIG. 2 is a graph showing a schematic diagram of differences in stability between regions of an mRNA consisting of a given base sequence. [Diagram 3] 1 is a graph comparing correlation coefficients in prediction of mRNA stability between the present invention and a conventional method. [Figure 4] 13 is another graph comparing correlation coefficients in prediction of mRNA stability between the present invention and the conventional method. [Diagram 5]1 is an example of a screen displayed via a computer system for simulating the stability of a biopolymer according to the present invention. [Figure 6] 11 is another example of a screen displayed via the computer system for simulating the stability of a biopolymer according to the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Next, an embodiment of the present invention will be described. FIG. 1 is a block diagram of an embodiment of a computer system for simulating the stability of a biopolymer according to the present invention. The computer system includes a calculation server 120 and a client computer 100 connected to the calculation server. The controller (or processor) of the computer system executes calculations to predict the stability of a biopolymer based on the sequence of a plurality of monomers that constitute the biopolymer, in order to simulate the stability of the biopolymer, based on the program in the memory. Prior to describing the system of FIG. 1, the present inventor's new findings in a simulation system for predicting the stability of a biopolymer will be described.

[0010] As mentioned above, biopolymers exist in living organisms and are polymers formed by polymerization of monomer units, and polynucleotides, polypeptides, and polysaccharides are known as biopolymers. In the following, we will explain the biopolymers by taking mRNA, which is a type of polynucleotide, as an example of RNA.

[0011] There are multiple nucleotide sequences, i.e., multiple base sequences, that code for one amino acid. Therefore, there are a large number of mRNA sequences that code for the same protein. In the development of nucleic acid medicines such as mRNA vaccines, it is necessary to select an mRNA molecule that is stable against degradation in the body from a large number of mRNA sequences that code for the target protein.

[0012] The half-life of mRNA described in the conventional method described in the above-mentioned prior art document 1 is actually determined by the decay rate constant (hereinafter, the decay rate constant of the conventional method is referred to as "β g(S) " ), which may be rephrased as the degradation rate constant or stability constant. The decay rate constant β of the mRNA of gene g g (S) is an index of mRNA stability, similar to the half-life, and the decay rate constant β g (S) is shown in the mRNA decay (degradation) reaction model in the following equation (1).

[0013] In formula (1), t is the time, [mRNA g ] is time series data on the amount of intracellular mRNA of gene g at time t. From equation (2) obtained by integrating equation (1), log[mRNA g ] against time to obtain β g (S) can be estimated. g (S) is the integral constant or intercept.

[0014]

number

[0015]

number

[0016] [mRNA g To calculate the amount of mRNA, we use Transcripts Per Million (TPM). TPM is an index of mRNA amount obtained from fastq files, which are experimental data from RNA-Seq. It is the number of reads mapped to genes corrected by the read length, gene length, and total number of reads.

[0017] Specifically, the fastq files were quality checked using Trimmomatic (v0.39), then mapped to the genome using HISAT2 (v2.2.1), the number of reads per gene was tallied using feature count in Subread (v2.0.0), and the TPM was calculated using a custom Python (3.9.15) script.

[0018] The reference files used in mapping and the Gene Transfer Format (GTF) files required for counting the number of reads were the latest versions (Release 109) available from Ensemble (https: / / asia.ensembl.org / index.html) as of January 2023.

[0019] For the RNA-Seq experimental data, we selected human: GSE126520, mouse: GSE99978, zebrafish: SRP072296, and Xenopus GSE65785, and downloaded them from the Gene Expression Omnibus (GEO) (https: / / www.ncbi.nlm.nih.gov / geo / ) of the National Center for Biotechnology Information (NCBI). Through the above operations, we were able to obtain [mRNA g ] and β g (S) was obtained.

[0020] An mRNA molecule is a long chain-like molecule, and the inventors' intensive studies have revealed that the stability of mRNA differs for each region of the nucleotide base sequence (monomer sequence: horizontal axis), as shown in Figure 2. Figure 2 is a graph that shows a schematic diagram of the difference in stability (vertical axis) for each region of an mRNA consisting of a specific base sequence.

[0021] The horizontal axis of FIG. 2 indicates the base sequence, i.e., the i-th base sequence, and the vertical axis indicates the amount of intracellular mRNA of gene g having the same nucleotide at the i-th nucleotide every predetermined time (0-6 hr). The larger the value on the vertical axis, the greater the amount of mRNA molecule, and the smaller the value on the vertical axis, the lower the amount. In other words, the greater the rate of change of the vertical axis value over time, the higher the stability of the mRNA molecule, and the smaller the rate of change of the vertical axis value over time, the lower the stability. From FIG. 2, it can be seen that the stability of the mRNA molecule varies greatly, like an irregular sawtooth, for each region 200 determined by i and w (w is the width of the base sequence).

[0022] The conventional methods described above do not take into account the stability of each region of an mRNA molecule, so when a machine learning learning model is used to predict stability, there is a limit to the accuracy of predicting the stability of an mRNA molecule having an arbitrary base sequence. The computer system of the present invention classifies a biopolymer into multiple regions, and introduces the stability of each region (decay rate of each region) into the calculation process to simulate the stability of the entire biopolymer.

[0023] Stability by region refers to the degree to which the nucleotide sequence of the same region is maintained when comparing mRNA time series data (RNA-Seq data) before and after a time. The stability of each of the multiple regions of a biopolymer is calculated based on the rate at which the sequence of the region decays. The smaller the degree to which the nucleotide sequence of the same region decays (decay rate, etc.), the higher the stability of the region. By obtaining a representative value by synthesizing, integrating, or fusing the stability of one region across the entire region and using the representative value as a factor that affects the stability of the mRNA itself, the stability of even a biopolymer whose stability is region-dependent can be estimated and predicted.

[0024] Furthermore, the computer system of the present invention calculates a representative value (corrected decay rate constant) as the stability of the entire biopolymer by integrating the stability of each region (decay rate of each region), and constructs a machine learning learning model (prediction model) using this as the objective variable. The computer system can accurately predict the stability of the entire biopolymer by applying the sequence of the biopolymer to the learning model as an explanatory variable.

[0025] The modified stability of the mRNA of gene g, the modified decay rate constant (the decay rate constant of the present invention relative to the conventional method, "β g (A) ") is defined by a window size w, which is the width of an arbitrary base sequence of 2 or more, starting from an arbitrary position i, and is a representative value for each region extracted from the mRNA sequence, that is, a representative value based on a statistical function of the decay rate calculated by expanding the reaction order from the mRNA amount of 200 in Figure 2 to a value other than 1. Hereinafter, this region-specific RNA amount is referred to as [mRNA g,i,w The calculated collapse rate is called the collapse rate by area: b g,i,w (A) He said.

[0026] [mRNA g,i,w ], in particular, when considering a first-order reaction model for the entire gene, as in the conventional method, b g,i,w (A) is the decay rate constant β g (S) If the size of gene g is L g If so, the range of the starting point i (monomer position) is 1 to L g -1, the window size w (width of the region) ranges from 2 to L g -i+1, and the computer system has a total of L g (L g -1) / For each of the two regions [mRNA g,i,w ] is calculated.

[0027] The computer system calculated [mRNA g,i,w ] is regressed against time taking into account the appropriate reaction order, g,i,w (A)(Decay rate by region) was calculated, and the obtained, L g (L g -1) / 2 b g,i,w (A) By the statistical function f, β g (A) Integrate into.

[0028] In addition, since it is possible to obtain the amount of mRNA at any point in the mRNA (called "depth") using existing bioinformatics tools, the system can obtain the amount of mRNA in any region ([mRNA g,i,w ]) can be sought.

[0029] The regional decay rate b, which is the time change of [mRNAg,i,w] g,i,w (A) is shown in the decay reaction model formula of formula (3), t is time, [mRNA g,i,w ] is the amount of mRNA in region 200 defined by i and w of gene g at time t (region-specific mRNA amount), p g,i,w is the reaction order when assuming a decay reaction for the region defined by i and w of gene g.

[0030] By integrating equation (3) and regressing the amount on the left side of equation (4) against time, we obtain b g,i,w (A) can be estimated. g,i,w (A) is the integral constant, or intercept.

[0031]

number

number

[0032] On the other hand, in this method, the modified decay rate constant β g (A)In the regression calculation at the time of calculation, the residuals (= the difference between the measured value and the predicted value by the regression line) were not normally distributed, and there were many cases where the results of the regression calculation could not be used as an estimated value. Therefore, before the regression analysis, the computer system performs a Box-cox transformation so that the residuals are normally distributed, and then calculates the corrected decay rate constant β g (A) We tried to obtain an estimate of

[0033] In formula (4), 1-p g,i,w = λ g,i,w The left side of the formula (5) is [mRNA g,i,w ] is a Boxcox transformation of the regression. The likelihood when the residuals in the regression after the Boxcox transformation are assumed to follow a normal distribution is given by equation (6). Therefore, we choose λ to maximize the likelihood of equation (6). g,i,w By determining p g,i,w is obtained.

[0034]

number

[0035]

number

[0036]

number

[0037] X is shown in Equation (7). I is a unit matrix. In Equation (7), at time t i The amount of mRNA in y ti , y ti The vector obtained by arranging the Boxcox transformed values ​​of (λ) , y (λ) The dimension of is expressed as n. g,i,w The search range was set to -2 to 2 because there are almost no reactions with a reaction order of 3 or more. The interval was set to 0.2. This search range is the same as that of Python's scipy.

[0038] The half-life in the model of equation (3) is (-1 / (β g (A) λ))*(1-2 -λ ) The system may then allow the user to choose whether to output the decay rate constant or the half-life.

[0039] Decay rates for multiple regions (b g,i,w (A) ) corrected decay rate constant (β g (A) The statistical function f for integrating the decay rate constant (β g (A) The system of the present invention allows the user to select from five methods for calculating the representative value: the average, median, minimum absolute value, or maximum absolute value of the decay rates by region, as well as the conventional method.

[0040] The definition of the statistic function f when the mean value is selected is as shown in the following equations (8) and (9).

[0041]

number

[0042]

number

[0043] When the median is selected, the statistical function f is defined as formula (10), and x k HA g,i,w (A) is a sequence of numbers rearranged in ascending order excluding zero, and N is the same value as in equation (9).

number

[0044] When the absolute minimum is selected, the statistic function f is defined as follows:

[0045]

number

[0046] When the absolute maximum is selected, the statistic function f is defined as follows:

number

[0047] For the conventional method, the decay rate b g,i,w (A) The corrected decay rate constant β g (A) When the conventional method is selected, the statistical function f is defined as follows: L g is the length of gene g.

[0048]

number

[0049] Next, the system according to Fig. 1 will be described again. The controller of the calculation server 120 in Fig. 1 realizes a stability simulation of a biopolymer based on the above-mentioned knowledge, based on a program recorded in the memory. The above-mentioned client computer 100 includes an RNA-Seq data receiving module 102, a machine learning parameter receiving module 104, and a machine learning result viewing module 106. Each of these modules is realized by the controller executing a program recorded in the memory. The module may be replaced with terms such as means, function, unit, and part.

[0050] The RNA-Seq data receiving module 102 receives a request for RNA-Seq data of the mRNA of a given gene g from a user. The machine learning parameter receiving module 104 sets parameters for controlling the deployment of machine learning in the machine learning processing module 134 (described later) of the calculation server 120 according to an input from the user. The machine learning parameter receiving module 104 receives input from the user, for example, via a keyboard, a mouse, a touch panel, a numeric keypad, a scanner, a microphone, or a sensor. Specifically, the machine learning parameter receiving module 104 may receive input via a screen such as that shown in FIG. 5.

[0051] The parameters include the selection of the decay rate constant or half-life as an index of stability of the biopolymer, the presence or absence of Boxcox transformation when regressing the previously described [mRNAg,i,w] (the amount of mRNA in the region defined by i,w of gene g at time) with time, the lower limit of the Ratio (described later), the upper limit of the standard deviation (described later), and the selection of the previously described type of statistical function (average value, median value, etc.). The client computer 100 allows the user to set parameters in order to enable machine learning according to the characteristics of genes and biopolymers (mRNA). The machine learning result viewing module 106 refers to the machine learning result database 124 to enable the user to view the results of machine learning. The machine learning result viewing module 106 outputs the results, for example, via a display, a printer, or a speaker (FIG. 6).

[0052] The calculation server 120 executes the biopolymer stability simulation system in cooperation with a database. The database includes an RNA-Seq database 122 and a machine learning result database 124. The machine learning result database 124 stores summary data 12401 related to machine learning and a learning model 12402. The database is, for example, a ROM (Read Only Memory), a RAM (Random Access Memory), a HDD (Hard Disk Drive), or a flash memory.

[0053] The calculation server 120 includes a depth calculation processing module 126, a region-specific mRNA amount calculation processing module 128, a region-specific decay rate calculation processing module 130, a corrected decay rate constant calculation processing module 132, and a machine learning processing module 134. These modules are also realized by the controller of the calculation server 120 executing a program. The combination of these modules constitutes a biopolymer stability simulation module.

[0054] The Depth calculation processing module 126 refers to the RNA-Seq database 122 based on the gene mRNA accepted by the RNA-Seq data accepting module 102, and performs data processing to create the graph of Figure 2 based on the time series data of the mRNA amount (Depth) at any point in the mRNA sequence.

[0055] The region-specific mRNA amount calculation processing module 128 calculates the amount of mRNA (see the above [mRNA g,i,w ]: the amount of mRNA in a region defined by i and w of gene g at time t). The corrected decay rate constant calculation processing module 132 calculates the decay rates (b g,i,w (A) ) corrected decay rate constant (β g (A) ) to be integrated.

[0056] The machine learning processing module 134 uses the logarithm of the codon frequency and CDS length of the mRNA molecule, and the 3'UTR length and a squared term consisting of any combination of these as explanatory variables, and constructs a prediction model by the Least Absolute Shrinkage and Selection Operator (LASSO). Other prediction models such as logistic regression, ridge regression, and elastic net regression may also be used. The machine learning processing module 134 randomly assigns training data and test data, and performs operations of creating a model and calculating a correlation coefficient multiple times, takes the average, and outputs the average as the final correlation coefficient. Specifically, the operation of calculating the correlation coefficient may be performed 1000 times, and the coefficient obtained by calculating the average may be used as the correlation coefficient.

[0057] The machine learning processing module 134 calculates the amount of explanatory variables based on the GTF file using a custom Python script. Python's scikit learn (v1.2.1) is used to build a predictive model using LASSO.

[0058] The inventors tested the following two methods for selecting the training gene set used to construct a learning model and the test gene set used to evaluate performance. The first method was the same as the conventional method, with 65,977 training genes and 7,376 test genes. g (S) A predictive model was constructed by setting the above as the objective variable, and the correlation coefficient was 0.43.

[0059] The second was for the purpose of performance verification using Bootstrap. A random 10% of the total 97,804 genes were used as test genes to build a predictive model 1000 times. The average correlation coefficient obtained was 0.12.

[0060] The vector of training gene names is g train Let g train For each element g, the decay rate constant β is calculated according to equations (1) and (2) and function f. g (S)The vector obtained by deriving and arranging is β train (S)(true) These are used as the objective variables for machine learning learning data. The correlation coefficient is calculated as β train (S)(true) and β train (A)(true) Since there is no change in train (x)(train) It is written as follows.

[0061] Next, the machine learning processing module 134 generates a feature matrix Q (train) =p(G train ) is calculated. Q i,j is a matrix that contains the value of the jth feature for the i-th gene, and is used as an explanatory variable for the machine learning training data. p is a function that takes a set of gene names as an argument and returns a matrix Q.

[0062] Next, the machine learning processing module 134 train (X)(true) =f LASS0 (x) (Q (train) ) LASS0 (x) Finally, the vector of test gene names is denoted by g test Then, the decay rate constant β is calculated according to equations (1) and (2) or equations (3) to (7) and the statistical function f. test (x)(true) On the other hand, the feature quantity Q (test) =p(G test ) and calculate β test (X)(pred) =f LASS0 (x) (Q (test) ) is calculated. The correlation coefficient (r) is expressed by the following formula (14), where Cov(·,·) is the covariance and V(·) is the variance.

[0063]

number

[0064] In order to improve the accuracy of machine learning, the machine learning processing module 134 uses reliable biopolymers (mRNA) and does not use genes that are not reliable in the training data and test data. Therefore, the system calculates the ratio (Ratio) of the region where the value of the vertical axis in Figure 2 is not 0 (non-zero region) to the total region, which is equal to or greater than a predetermined threshold value, and calculates the decay rate (b wi (A) We decided to use mRNAs whose standard deviation of ) was below a certain threshold for machine learning.

[0065] Next, the standard deviation (SD)-correlation coefficient change characteristics calculated by the conventional method and the present invention will be compared for each of the above-mentioned ratios. However, the allocation of training data and test data was performed randomly, and the operation of creating a model and calculating the correlation coefficient was performed 1000 times, and the final correlation coefficient was calculated by taking the average.

[0066] As a result of machine learning, the correlation coefficient of the present invention was improved compared to the conventional method. Figure 3 (A) shows the change characteristics of the correlation coefficient when the statistical function f is set to the "average value" and the standard deviation (SD) mentioned above for each ratio. Figure 3 (B) shows the change characteristics of the correlation coefficient when the statistical function f is set to the "median" and the standard deviation (SD) mentioned above. Figure 3 (C) shows the change characteristics in the conventional method, where SD is the β mentioned above. g (S) 3A, 3B, and 3C, it can be seen that the correlation coefficient of the present invention is higher than that of the conventional method.

[0067] FIG. 4 shows the results of a further comparison between the correlation coefficient of the present invention and that of the conventional method. FIG. 4 differs from FIG. 3 in that a Box-cox transformation was performed during regression analysis. FIG. 4(A) shows the change characteristics of the correlation coefficient when the statistical function f is set to the "average value" and the standard deviation (SD) mentioned above for each ratio. FIG. 4(B) shows the change characteristics of the correlation coefficient when the statistical function f is set to the "median" and the standard deviation (SD) mentioned above. FIG. 4(C) shows the change characteristics of the conventional method. Comparing FIG. 4(A), (B), and (C), it can be seen that the correlation coefficient of the present invention is further improved over that of the conventional method.

[0068] Next, we will explain the operation of the mRNA stability simulation using the learning model 12402. The user inputs a given nucleotide into the client computer 100. This information is sent to the calculation server 120, and the simulation analysis unit of the calculation server 120 refers to the machine learning result database 124 to determine multiple nucleotide sequence candidates that have identical amino acid sequences encoded by the nucleotide sequences.

[0069] The simulation analysis unit inputs sequence information of each of the multiple candidates as explanatory variables to the learning model 12402. The learning model 12402 calculates a target variable based on the explanatory variables of each of the multiple candidates, and outputs the multiple candidates in a ranked form in descending order of the decay rate constant. Based on this ranking, the user can determine the mRNA having the sequence with the best degradation stability.

[0070] According to the above-mentioned embodiment, there is provided a first computer system for simulating the stability of a biopolymer, comprising a controller for executing a program stored in a memory, the controller acquiring information on the sequence of the biopolymer, classifying the sequence into a plurality of regions, calculating the stability of each of the plurality of regions, and executing a prediction simulation of the stability of the biopolymer itself based on a representative value summarizing the stability of each of the plurality of regions based on the result of the calculation. According to the first computer system, it is possible to predict the decomposition stability of a biopolymer having an arbitrary sequence with high accuracy.

[0071] Further, according to the above-mentioned embodiment, in the first computer system, a second computer system is provided, which extracts each of the plurality of regions from the sequence of the biopolymer by specifying the respective regions by the position and width of the sequence, and calculates the stability of each of the plurality of regions based on the time series data of the amount of the biopolymer. The second computer system can obtain the stability of each of the plurality of regions.

[0072] Further, according to the above-mentioned embodiment, there is provided a third computer system, which is the second computer system, and calculates the stability of each of the plurality of regions based on the rate at which the sequence of the region collapses. With the third computer system, the stability of each of the plurality of regions can be reliably obtained.

[0073] Furthermore, according to the above-mentioned embodiment, a fourth computer system can be provided in which, in the first computer system, machine learning is executed using the stability of the biopolymer itself as a response variable and sequence information constituting the biopolymer as an explanatory variable, and a prediction model for the stability of the biopolymer itself is constructed by the machine learning. According to the fourth computer system, a learning model capable of reliably predicting the stability of the biopolymer can be provided.

[0074] Furthermore, according to the above-mentioned embodiment, there is provided a fifth computer system, which is the fourth computer system, and which newly acquires information on the sequence of a biopolymer and calculates the stability of the biopolymer itself based on the prediction model. According to the fifth computer system, even if a new biopolymer is received as an input, the stability of the newly received biopolymer can be calculated with high accuracy based on the constructed prediction model.

[0075] According to the above-mentioned embodiment, there is provided a sixth computer system, which is the third computer system, and calculates the rate at which the sequence of the region decays from the change in the amount of the biopolymer having the sequence of the region based on the time series data of the biopolymer. According to the sixth computer system, the stability of the region can be reliably calculated.

[0076] Furthermore, according to the above-described embodiment, there is provided a seventh computer system in which the fifth computer system is configured to accept as an input a method for calculating the representative value. According to the seventh computer system, the method for calculating the representative value can be appropriately changed by the user depending on the biopolymer.

[0077] According to the above-mentioned embodiment, the first computer system receives time series data of mRNA as the biopolymer, calculates the modification stability of each region for the sequence of the mRNA from the time series data, calculates the modification stability as a target variable in machine learning based on the stability, performs machine learning based on the modification stability and the time series data, generates a learning model, and outputs the learning model. According to this eighth computer system, the prediction accuracy of the stability of mRNA can be improved.

[0078] Further, according to the above-mentioned embodiment, there is provided a ninth computer system, which calculates the stability of each of the plurality of regions by regressing the molecular weight of a biopolymer having a sequence of the region with time in the second computer system. According to the ninth computer system, the stability of each of the regions can be estimated by regression analysis.

[0079] Furthermore, according to the above-mentioned embodiment, there is provided a tenth computer system, which performs a Box-cox transformation on the amount of a biopolymer having the sequence of the region in time regression in the fifth computer system. According to the tenth computer system, the user can select whether or not to perform the Box-cox transformation depending on the attributes of the biopolymer.

[0080] Further, according to the above-mentioned embodiment, there can be provided an eleventh computer system for simulating the stability of a biopolymer, the computer system including a controller for executing a program stored in a memory, the controller receiving time series data of mRNA as the biopolymer, classifying the sequence into a plurality of regions, calculating the stability of each of the plurality of regions, and outputting the stability. According to the eleventh computer system, according to the first computer system, the degradation stability of an mRNA having an arbitrary sequence can be predicted with high accuracy.

[0081] According to the embodiment described above, there is provided a simulation system method for simulating the stability of a biopolymer, in which the computer acquires information on the sequence of the biopolymer, classifies the sequence into a plurality of regions, calculates the stability of each of the plurality of regions, and performs a predictive simulation of the stability of the biopolymer itself based on a representative value that summarizes the stability of each of the plurality of regions based on the results of the calculation.

[0082] Although the embodiment has been described above, the technical scope of the present invention is not limited to the scope described in the above embodiment. It is clear from the claims that various modifications and improvements to the above embodiment are also included in the technical scope of the present invention.

[0083] For example, in the above-described embodiment, mRNA has been used as an example of a biopolymer, but the present invention can also be applied to biopolymers of a type whose stability differs depending on the region of the monomer sequence, such as proteins or polysaccharides whose stability differs depending on the three-dimensional structure. [Explanation of symbols]

[0084] 100: client computer, 102: RNA-Seq data reception module, 104: machine learning parameter reception module, 106: machine learning result viewing module, 120: calculation server, 122: RNA-Seq database, 124: machine learning result database, 126: depth calculation processing module, 128: region-specific mRNA amount calculation processing module, 130: region-specific decay rate calculation processing module, 132: corrected decay rate constant calculation processing module, 134: machine learning processing module, 200: region, 12401: machine learning summary data, 12402: learning model

Claims

1. A computer system for simulating the stability of a biopolymer, comprising: A controller for executing a program stored in a memory, The controller: Obtaining information on the sequence of the biopolymer; Classifying the sequence into regions; Calculating the stability of each of the plurality of regions; executing a prediction simulation of the stability of the biopolymer based on a representative value summarizing the stability of each of the plurality of regions based on the result of the calculation; Computer system.

2. The controller: extracting each of the plurality of regions from the sequence of the biopolymer by identifying each of the regions by a position and a width of the sequence; Calculating the stability of each of the plurality of regions based on time series data of the amount of the biopolymer.

2. The computer system of claim 1.

3. The controller: Calculating the stability of each of the plurality of regions based on the rate at which the sequence of the region decays; 3. The computer system of claim 2.

4. The controller: Executing machine learning using the stability of the biopolymer as a response variable and sequence information constituting the biopolymer as an explanatory variable; A predictive model of the stability of the biopolymer itself is constructed by the machine learning.

2. The computer system of claim 1.

5. The controller: Obtaining new information on the sequence of biopolymers, Calculating the stability of the biopolymer itself based on the prediction model; 5. The computer system of claim 4.

6. The controller: Calculating the rate at which the sequence of the region decays from the change in the amount of the biopolymer having the sequence of the region based on the time series data of the biopolymer; 4. The computer system of claim 3.

7. The controller: Accepting as input a method for calculating the representative value; 7. The computer system of claim 6.

8. The controller: receiving time series data of mRNA as the biopolymer; Calculating the stability of each region of the mRNA sequence from the time series data; Calculating a modified stability as a target variable in machine learning based on the stability; Performing machine learning based on the correction stability and the time series data to generate a learning model; Outputting the learning model.

2. The computer system of claim 1.

9. The controller: Calculating the stability of each of the plurality of regions by regressing the molecular weight of a biopolymer having a sequence of the region over time.

3. The computer system of claim 2.

10. The controller: When the amount of a biopolymer having the sequence of the region is regressed over time, the amount of the biopolymer is subjected to a Boxcox transformation based on a user's input.

7. The computer system of claim 6.

11. A computer system for simulating the stability of a biopolymer, comprising: A controller for executing a program stored in a memory, The controller: receiving time series data of mRNA as the biopolymer; Classifying the sequence into regions; Calculating the stability of each of the plurality of regions; outputting said stability; Computer system.

12. A system and method for simulating the stability of a biopolymer, comprising: Obtaining information on the sequence of the biopolymer; classifying the sequence of the biopolymer into a plurality of regions; Calculating the stability of each of the plurality of regions; A simulation is performed to predict the stability of the biopolymer itself based on a representative value obtained by summarizing the stability of each of the plurality of regions based on the results of the calculation. Simulation method.