KR20080040481A - System, method and program for pharmacokinetic parameter prediction of peptide sequence by mathematical model - Google Patents

Abstract

A system, a method and a program for estimating pharmacokinetic parameters of a peptide sequence by using a mathematical model are provided to estimate whether a peptide sequence passes through the small intestine using a program recording medium without confirming it through an experiment so as to reduce the cost and time required for the experiment. A system for estimating pharmacokinetic parameters of a peptide sequence using a mathematical model includes a microcomputer(10), an input unit(20) and an output unit(30). The microcomputer includes a program recording medium(11), a CPU(Central Processing Unit)(12) and an input/output unit(13). The program recording medium includes a program of converting a peptide sequence inputted by a user into an amino acid descriptor and a program of estimating whether the peptide sequence passes through the small intestine by using a trained mathematical model.

Description

System, method and program for pharmacokinetic parameter prediction of peptide sequence by mathematical model}

1 is a block diagram showing an embodiment of a system for predicting pharmacokinetic properties of a peptide sequence using a mathematical model according to the present invention,

Figure 2 is a flow chart showing an embodiment of a method for predicting the pharmacokinetic properties of the peptide sequence using the mathematical model of the present invention,

Figure 3 is a flow chart showing one embodiment of a method for predicting the pharmacokinetic properties of the peptide sequence using the mathematical model of the present invention,

4 is a flowchart of a method for retraining a pharmacokinetic property prediction model in the present invention.

<Explanation of symbols for main parts of drawing>

10: microcomputer 11: program recording medium

12: CPU 13: input / output unit

20: input means 30: output means

The present invention relates to a system and method for predicting whether a peptide sequence passes through the small intestine using a mathematical model, and a recording medium storing the program.

In general, in developing new drugs or new substances, constructing a model of the correlation between structure and activity is one of the most useful methods for predicting activity while reducing the cost of experiments.

However, while there are programs that use these methods to predict various properties necessary for the development of new drugs such as small intestine passage, solubility, and toxicity for small organic compounds using these methods, to date the peptide sequence There was no program that could predict the characteristics in advance.

In the recent development of new drugs, peptides are one of the most promising research materials for developing new drugs due to their strong efficacy, little toxicity and side effects, and non-existence in the human body. have. However, most peptide drugs have a disadvantage of lowering the absorption rate in the body when administered orally. For this reason, there is an urgent need for the development of a technique capable of predicting whether the small intestine passes through the peptide sequence, which is one of the activities necessary for oral drug delivery or new drug development.

The present invention has been invented in view of the above circumstances, and provides a system and method for predicting whether the peptide sequence passes through the small intestine using a mathematical model and a recording medium storing the program to predict whether the peptide sequence passes through the small intestine. The goal is to present a valid model.

The pharmacokinetic properties prediction system of the peptide sequence using the mathematical model of the present invention for achieving the above object comprises a microcomputer (10) comprising a program recording medium (11), a CPU (12) and an input / output unit (13); Input means 12; Characterized in that the output means 30.

The program recording medium 11 includes a program for converting a peptide sequence to which the user wants to know whether the small intestine has passed or not, and a program for predicting whether the small intestine has passed using a trained mathematical model. Adds a new small intestinal transit peptide sequence that has obtained small intestinal transit activity value using an experimental technique, adds it to the original small intestinal transit peptide set, classifies it, and assigns the expression value and activity value to the added peptide. And a program for training a mathematical model using a training peptide set, a program for predicting whether the small intestine passes through a set of peptides for verification, and a program for verifying a trained mathematical model. .

In addition, the method for predicting the pharmacokinetic properties of the peptide sequence using the mathematical model of the present invention for achieving the above object comprises the steps of obtaining a sample of the peptide sequence through the small intestine using an experimental technique; Obtaining a sample of a sequence of peptides that do not pass through the small intestine based on these sequences; Storing each of the obtained samples as a set and then randomly extracting them in a predetermined ratio to classify them into a mathematical model training set and a mathematical model verification set; Assigning a descriptor value and an activity value to the individual peptide sequence; Training using a training peptide set and a mathematical model; Predicting small intestine passage for the validation peptide set using a trained mathematical model; Verifying a trained mathematical model.

The mathematical model is a quantitative structure including regression analysis, machine learning, multiple regression analysis using genetic algorithm, partial least square method using genetic algorithm, partial least square method using principal component analysis, and multiple regression analysis using principal component analysis. A mathematical model method characterized in that it is a characteristic correlation method, and the machine learning method is a neural network, data mining, decision tree, inductive logic, case-based reasoning, pattern recognition, reinforcement learning, Bayesian network, hidden marker model, probability grammar Method, and in particular, a neural network technique.

The pharmacokinetic property of the peptide sequence is absorption in the small intestine, and the presenter value quantitatively represents the molecular structure, amino acid, and peptide, and includes at least one of a binary amino acid presenter, a VHSE amino acid presenter, a Z3 amino acid presenter, and a Z5 amino acid presenter. It is characterized by including any one.

The data collected to build the machine learning model is in vivo , ex vivo is, data obtained from at least any one of the in vitro experiments, especially in using the phage display technique experiment vivo , ex vivo , in It is characterized in that the data obtained from at least one of the in vitro experiments. The peptide sequence is a sequence consisting of 2 to 12 peptides, more preferably 3 to 7 peptides, and the species to which the method of predicting pharmacokinetic properties of the peptide sequence using the mathematical model of the present invention are mammals, particularly humans. It is done.

In addition, the recording medium storing the program for predicting the pharmacokinetic properties of the peptide sequence using the mathematical model of the present invention includes the process of obtaining a sample of the peptide sequence through the small intestine using an experimental technique; Obtaining a sample of a sequence of peptides that do not pass through the small intestine based on these sequences; Storing each of the obtained samples as a set and then randomly extracting them in a predetermined ratio to classify them into a mathematical model training set and a mathematical model verification set; Assigning a descriptor value and an activity value to the individual peptide sequence; Training using training peptide sets and mathematical models; A processor for predicting whether the small intestine passes through the set of peptides for verification using a trained mathematical model; And a processor for verifying the trained mathematical model.

The objects, features and advantages of the present invention will be more readily understood by reference to the accompanying drawings and the following detailed description.

Hereinafter, a system and method for predicting pharmacokinetic properties of a peptide sequence using the mathematical model of the present invention and a recording medium storing the program will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing an embodiment of a system for predicting pharmacokinetic properties of a peptide sequence using a mathematical model according to the present invention, Figure 2 is an embodiment of a method for predicting pharmacokinetic properties of a peptide sequence using the mathematical model of the present invention As an example flow chart, as shown in FIG. 2, a sample (number) of peptide sequences passing through the small intestine is first collected by the phage display experiment technique (step S1). Here, the length of the peptide sequence means the number of amino acids in one peptide, and the peptide sequence length 3 indicates a peptide having three amino acids. The number of peptide sequences collected is shown in Table 1 below. In the case of the peptide sequence consisting of three amino acids in Table 1, the number of peptide sequences passing through the small intestine obtained by phage display experiment technique is 4252.

In addition, the phage display peptide library used in step S1 is 'ph.D.-C7C ^TM (New England BioLab.)', Which is a constant of coat protein in the genome of M13 bacteriophage. More than hundreds of millions of different species obtained by infecting E. coli by artificially inserting gene sequences to express peptides of 7 random amino acid sequences at the ends of pIII-producing genes It consists of recombinant bacteriophage expressing peptides. On the other hand, the seven random amino acid sequences introduced into the M13 phage are designed to have cysteine residues on both sides to form a disulfide bond in the expression of the peptide, thereby forming a loop shape, thereby allowing the loop protein to form a loop shape. It is designed to induce strong bonding. The oral phage display technique was followed by oral injection of 1.2 X 1012 pfu of phage peptide library (~ 1,000 copies of individual recombinant phage clones) into mice overnight, and after 1 hour, representative internal organ tissues (liver, lung, kidney, And then pass through the mucous membrane of the small intestine, and collect and quantify the phages distributed in each organ along the bloodstream.

TABLE 1 Number of Peptide Sequences

Length of peptide sequence             Number of Peptide Sequences    Small intestine  Small intestine  For training   Verification           3      4252     4252   6786   1718           4      3402     3402   5428   1376           5      2552     2552   4078   1026           6      1702     1702   2748    656           7       852      852   1400    304

In addition, using a program to select any amino acid, three amino acids of length 3 of the small intestine pass peptide sequence were extracted and compared to the small intestine pass peptide set obtained in the experiment. Classify into a set of sequences (step S2). The program for selecting any amino acid here uses a well-known program.

Next, classify a set of peptide sequences for machine learning training (step S3). This step (step S3) includes a process of making the population equal between the two sets because the number of peptide sequence sets that pass through the small intestine is less than the number of peptide sequence sets that do not pass through the small intestine. In this step (step S3), in the case of peptide sequence length 3, 4252 small intestinal non-penetrating peptides were obtained as shown in Table 1.

Subsequently, approximately 80% of the random peptide sequences are extracted from the small intestine-penetrating peptide set, approximately 80% of the random peptide sequences are extracted from the small intestine non-penetrating peptide set, and the two are collected and classified into a set of machine learning training peptides (step S4). ).

As in step S4, approximately 20% of the small intestine passing peptide set and the remaining 20% of the small intestine passing peptide set are collected and classified into a set of machine learning verification peptides (step S5).

As a result, as shown in Table 1, in the case of length 3 of the peptide sequence, the number of machine learning peptides was 6786, and the number of machine learning verification peptides was 1718.

Next, using the machine learning method, the step of training and acquiring the small intestine passage prediction model using the machine learning training set obtained in step S4 is performed (step S10). That is, a step of arbitrarily changing the order in which the small intestine passing peptide set is inputted, and the small intestine passing peptide sequence and the small intestine non-passing peptide sequence are alternated in equal proportions, so that the machine learning training set can be entered as an input value. The order is adjusted and input as an input value for training the machine learning model (step S11).

Thereafter, individual peptide sequences included in the machine learning training set are converted into amino acid descriptor values (step S12). Wherein the amino acid presenter value includes at least one of a binary amino acid presenter, a VHSE amino acid presenter, a Z3 amino acid presenter, and a Z5 amino acid presenter, and the binary amino acid presenter for an amino acid is defined as 19 "0s for an amino acid. It is expressed by the number of 20 digits which consist of "and one" 1 ", and it sets so that the order in which the value of" 1 "may be located may differ from each other for 20 amino acids. In the case of peptide sequence length 3, it consists of 60 markers, and the small intestinal transit activity value is 0.9 for small intestinal transit peptide and 0.1 for small intestinal non-penetrating peptide.

As such, conversion of individual peptide sequences included in the machine learning training set to the presenter value can be converted using the VHSE amino acid presenter, and the values defined therefor are shown in Table 3 below. VHSE descriptors consist of eight markers for an amino acid, which are known to represent the hydrophobicity, electronic and steric properties of the amino acids, and 24 inputs for peptide sequence length 3. It consists of.

TABLE 2 VHSE Amino Acid Descriptors

 Amino acids VHSE 1 VHSE 2 VHSE 3 VHSE 4 VHSE 5 VHSE 6 VHSE 7 VHSE 8  Ala   A   0.15  -1.11  -1.35  -0.92   0.02  -0.91   0.36  -0.48  Arg   R  -1.47   1.45   1.24   1.27   1.55   1.47   1.30   0.83  Asn   N  -0.99   0.00  -0.37   0.69  -0.55   0.85   0.73  -0.80  Asp   D  -1.15   0.67  -0.41  -0.01  -2.68   1.31   0.03   0.56  Cys   C   0.18  -1.67  -0.46  -0.21   0.00   1.20  -1.61  -0.19  Gln   Q  -0.96   0.12   0.18   0.16   0.09   0.42  -0.20  -0.41  Glu   E  -1.18   0.40   0.10   0.36  -2.16  -0.17   0.91   0.02  Gly   G  -0.20  -1.53  -2.63   2.28  -0.53  -1.18   2.01  -1.34  His   H  -0.43  -0.25   0.37   0.19   0.51   1.28   0.93   0.65  Ile   I   1.27  -0.14   0.30  -1.80   0.30  -1.61  -0.16  -0.13  Leu   L   1.36   0.07   0.26  -0.80   0.22  -1.37   0.08  -0.62  Lys   K  -1.17   0.70   0.70   0.80   1.64   0.67   1.63   0.13  Met   M   1.01  -0.53   0.43   0.00   0.23   0.10  -0.86  -0.68  Phe   F   1.52   0.61   0.96  -0.16   0.25   0.28  -1.33  -0.20  Pro   P   0.22  -0.17  -0.50   0.05  -0.01  -1.34  -0.19   3.56  Ser   S  -0.67  -0.86  -1.07  -0.41  -0.32   0.27  -0.64   0.11  Thr   T  -0.34  -0.51  -0.55  -1.06  -0.06  -0.01  -0.79   0.39  Trp   W   1.50   2.06   1.79   0.75   0.75  -0.13  -1.01  -0.85  Tyr   Y   0.61   1.61   1.17   0.73   0.53   0.25  -0.96  -0.52  Val   V   0.76  -0.92  -0.17  -1.91   0.22  -1.40  -0.24  -0.03

Subsequently, machine learning training is performed using the experimental value of whether the small intestine passes through the set of peptides for machine learning training and the expression value of the peptide sequence as input values (step S13). The methods for machine learning are Neural Network, Data Mining, Decision Tree, Case Based Reasoning, Pattern Recognition, Reinforcement Learning For example, in the case of using a feed forward neural network, feed forward neural network learning training is performed using a set of machine learning training peptides. The structure of the feed forward neural network for machine learning is composed of an input layer, a hidden layer, and an output layer. The input layer is composed of input nodes, and the number of input nodes is determined by multiplying the number of peptide sequence lengths by the number of components of the presenter value, and one input node is a real or integer value of one presenter value component. One hidden layer consists of 0 to 2 hidden nodes, and an output layer consists of output nodes, and the number of output nodes is one. In the case of using the 20-digit binary amino acid descriptor for the peptide sequence length 3, the structure of the feed forward neural network is composed of 60 input nodes, and the input values of each node are 60 expression values "0" or "1" created in step S12. "to be. The structure of the feedforward neural network for all peptide sequence lengths may consist of an output layer with just one output node without using a hidden layer.

Next, a small intestine passage prediction model capable of predicting whether the peptide sequence passes through the small intestine is obtained through appropriate machine learning training in the step S13 (step S14).

Subsequently, the small intestine passage prediction model obtained in step S14 and the machine learning verification set obtained in step S5 are obtained, and the predicted small intestine passage prediction model is verified and evaluated by comparing with the experimental values (S20). step). This step S20 consists of steps S21 to S24, that is, first prepares input values for the machine learning model verification (step S21). In step S21, the machine learning verification set obtained in step S5 is used as it is.

Subsequently, the individual peptide sequences included in the machine learning verification set are converted into the presenter values (step S22). In this case, the presenter uses the same presenter that the small intestine passage prediction model acquired in step S14 is identical to the presenter used in the training process of step S13.

Subsequently, an amino acid descriptor value for the peptide sequence is used as an input value as a set of machine learning verification peptides, and a small intestine passage prediction model obtained in step S14 is obtained to predict whether the small intestine passes (step S23).

Then, using the machine learning verification set to obtain the predicted value for passing the small intestine using the small bowel prediction model obtained in step S23 and using the results are shown in Table 3 below (step S24).

In step S22, using the presenter as a 20-digit binary amino acid presenter, a machine learning model was trained to execute step S24, and the results are shown in Table 3 below.

[Table 3] Validation results of the small intestine pass prediction model

Peptide sequence length                 Responder Activity Score      Randomize input order         5 sets of whole sets  For training (80%)  Verification (20%)  For training (80%)  Verification (20%)     3 0.8885 ± 0.0014 0.8876 ± 0.0056 0.8894 ± 0.0035 0.8855 ± 0.0152     4 0.7203 ± 0.0065 0.6907 ± 0.0068 0.7242 ± 0.0047 0.7059 ± 0.0173     5 0.7475 ± 0.0047 0.7212 ± 0.0140 0.7471 ± 0.0032 0.7279 ± 0.0070     6 0.7813 ± 0.0068 0.7444 ± 0.0244 0.7870 ± 0.0033 0.7447 ± 0.0136     7 0.8228 ± 0.0457 0.7707 ± 0.0209 0.8452 ± 0.0060 0.7884 ± 0.0412

As can be seen from Table 3, the verification was performed five times by arbitrarily changing the order of input values of the feedforward neural network, and as a result, the responder action characteristic score for the peptide sequence length 3 was 0.8885 ± 0.0014 for the training set and 0.8876 ± for the validation set. It was 0.0056. The whole set was divided into 5 parts, 4 parts were used for training, and the remaining 1 part was used for verification, and the results were verified by exchanging training equal parts and verification parts alternately. In case of 3, the training set was 0.8894 ± 0.0035 and the verification set was 0.8855 ± 0.0152.

In step S22, using the presenter as a VHSE amino acid presenter to train the machine learning model to perform the step S24 and the results are shown in Table 4 below.

[Table 4] Validation results of the small intestine passing prediction model

Peptide sequence length                  Responder Activity Score      Randomize input order        5 sets of whole sets   For training (80%)   Verification (20%)   For training (80%)   Verification (20%)        3 0.8371 ± 0.0025 0.8305 ± 0.0121 0.8358 ± 0.0024 0.8321 ± 0.0098        4 0.6937 ± 0.0069 0.6828 ± 0.0099 0.7032 ± 0.0040 0.6930 ± 0.0148        5 0.7129 ± 0.0071 0.6833 ± 0.0158 0.7149 ± 0.0031 0.7014 ± 0.0128        6 0.7460 ± 0.0080 0.7184 ± 0.2445 0.7537 ± 0.0032 0.7299 ± 0.0156        7 0.7964 ± 0.0074 0.7497 ± 0.0170 0.7999 ± 0.0062 0.7605 ± 0.0220

As can be seen from Table 4, the verification was performed five times by arbitrarily changing the order of input values of the feed forward neural network. 0.0121, and the whole set was divided into 5 parts, 4 parts were used for training, and the remaining 1 part was used for verification. In the case of 3, the training set was 0.8358 ± 0.0024 and the verification set was 0.8321 ± 0.0098.

Next, in step S24, when the small intestine passthrough peptide set is replaced with a randomly selected set of the small intestine passing peptides, and the feedforward neural network model is trained using the same, the feedforward neural network model is the small intestine passing peptide sequence and the small intestine. In order to verify whether the non-passing peptide sequences were accidentally distinguished or an accurate learning model was made, five times were verified in the example using a binary descriptor for amino acids and the results are shown in Table 5 below.

[Table 5] Validation results of the small intestine pass prediction model

   Peptide sequence length           Responder Activity Score       For training (80%)      Verification (20%)             3     0.5705 ± 0.0024     0.4935 ± 0.0079             4     0.5745 ± 0.0070     0.4970 ± 0.0244             5     0.5947 ± 0.0021     0.4989 ± 0.0114             6     0.6156 ± 0.0096     0.4849 ± 0.0353             7     0.6959 ± 0.0105     0.4969 ± 0.0216

As can be seen from Table 5, the responder functional characteristic score for the peptide sequence length 3 was low in the training set of 0.5705 ± 0.0024 and the validation set in 0.4935 ± 0.0079.

In addition, the verification was performed five times in the example using the VHSE amino acid descriptor in the step S24 and the results are shown in Table 6 below.

[Table 6] Validation results of the small intestine pass prediction model

  Peptide sequence length        Responder Activity Score        For training (80%)       Verification (20%)            3      0.5523 ± 0.0037     0.5171 ± 0.0142            4      0.5521 ± 0.0080     0.4968 ± 0.0197            5      0.5564 ± 0.0041     0.4807 ± 0.0213            6      0.5727 ± 0.0050     0.4750 ± 0.0234            7      0.6265 ± 0.0094     0.4926 ± 0.0155

As can be seen from Table 6, the responder functional characteristic score for the peptide sequence length 3 was low in the training set of 0.5523 ± 0.0037 and the validation set of 0.5171 ± 0.0142. In this embodiment using two different presenters, the use of a false small pass peptide as an input means that a machine learning model is not created, and a feedforward artificial composed of an input layer, a hidden layer, and an output layer for the peptide sequence. It can be seen that the neural network model distinguished the actual small intestinal pass-through peptide sequence from the small intestine pain-free peptide sequence.

Figure 3 is a flow chart showing an embodiment of a method for predicting the pharmacokinetic properties of the new peptide sequence using the machine learning method of the present invention, first input the peptide sequence to know the pharmacokinetic properties through the input means 20 to record the program The medium 11 is stored in the medium 11 (step S101).

Next, the individual peptide sequences inputted are converted into the expression values required by the trained prediction model (step S23) through the process of FIG. 2 (step S102).

Thereafter, it is applied to the pharmacokinetic property prediction model composed of the trained prediction model (step S23).

In order to know the pharmacokinetic properties, whether the small intestine passes through the new peptide sequence input by the user is output (step S104).

4 is a flowchart illustrating a method for retraining a pharmacokinetic property prediction model according to the present invention. First, a means for inputting a new small intestine passing peptide sequence and a small intestine non-passing peptide sequence obtained by an experimental technique of small intestine passage activity (20) Stored in the program recording medium 11 through step S201.

Subsequently, the machine learning model is trained by performing the steps S3 to S5 and the steps S10 and S20 of FIG. First, it is determined whether the newly input peptide sequences are the same as the already designated sequences, and these sequences are added and stored in the small intestine passing peptide set and the small intestine passing peptide set according to the activity value (step S211).

Next, by adding the newly input peptide sequence to the previously stored peptide sequence to classify the peptide sequence set for the machine learning training of step S3 of FIG. 2, and obtains the peptide set for machine learning training of step S4 Obtain a set of peptides for verifying machine learning in step S5, train the small intestine passage prediction model using the machine learning method in step S10, and verify the small intestine passage prediction model using the machine learning method in step S20 (step S212). ).

Thereafter, the responder function characteristic score of the small intestine passage prediction model stored previously is compared with the responder function characteristic score of the small intestine passage prediction model obtained in step S212 (step S213).

Subsequently, the responder action characteristic score calculated in step S213 is output to the user, and based on this, the newly trained small intestine passage prediction model is stored (step S202).

This allows the user to retrain and validate mathematical models based predictive models using newly acquired small intestine peptide sequences through experiments.

Although the present invention has been described as a preferred embodiment so far, the present invention is not limited thereto and can be variously modified and implemented within the scope not departing from the gist of the invention.

As described above, according to the present invention, it is possible to predict in advance by using a recording medium storing a program without always confirming whether the small intestine passes through the small intestine, which is one of the characteristics required for drug delivery by oral administration of a peptide drug. There is a special advantage that can reduce the cost and time required for the experiment.

Claims (16)

A microcomputer 10 comprising a program recording medium 11, a CPU 12, and an input / output unit 13; Input means 20; System for predicting pharmacokinetic properties of a peptide sequence using a mathematical model, characterized in that the output means (30). 2. The program according to claim 1, wherein the program recording medium 11 converts the peptide sequence to a presenter when the user inputs a peptide sequence to know whether the small intestine passes, and a program for predicting whether the small intestine passes using a trained mathematical model. If the user adds a new small intestinal transit peptide sequence obtained in the small intestine passthrough activity value by the experimental technique, the program to add it to the original small intestinal transit peptide set and then classify, the expression value and activity in the added peptide A program for assigning values, a program for training a mathematical model using a training peptide set, a program for predicting passage of a small intestine for a test peptide set, and a program for validating a trained mathematical model. Peptide sequence using a mathematical model, characterized in that Pharmacokinetic Properties Prediction System. Obtaining a sample of peptide sequence through the small intestine using experimental techniques; Obtaining a sample of a sequence of peptides that do not pass through the small intestine based on these sequences; Storing each of the obtained samples as a set and randomly extracting them in a predetermined ratio to classify them into a mathematical model training set and a mathematical model verification set; Assigning a descriptor value and an activity value to the individual peptide sequence; Training using a training peptide set and a mathematical model; Predicting small intestine passage for the validation peptide set using a trained mathematical model; A method for predicting the pharmacokinetic properties of a peptide sequence using a mathematical deduction model, characterized in that the step of verifying the trained mathematical model. 4. The mathematical model of claim 3, wherein the mathematical model includes regression analysis, machine learning, multiple regression analysis using genetic algorithm, partial least square method using genetic algorithm, partial least square method using principal component analysis, and multiple regression using principal component analysis. A method for predicting pharmacokinetic properties of a peptide sequence using a mathematical model, characterized in that it is a quantitative structure-characteristic correlation method including an analysis method. The method of claim 3, wherein the machine learning method is a neural network, data mining, decision tree, inductive logic, case-based reasoning, pattern recognition, reinforcement learning, Bayesian network, hidden Markov model, mathematical model characterized in that the probability grammar method Method for predicting pharmacokinetic properties of peptide sequences used. 5. The method of claim 4, wherein the machine learning method is a neural network technique. The method of claim 3, wherein the pharmacokinetic properties of the peptide sequence is absorption in the small intestine. 4. The method of claim 3, wherein the presenter value is a quantitative representation of molecular structure, amino acids, and peptides. 9. The pharmacokinetic property of a peptide sequence using a mathematical model according to claim 8, wherein the presenter value comprises at least one of a binary amino acid presenter, a VHSE amino acid presenter, a Z3 amino acid presenter, and a Z5 amino acid presenter. Prediction Method. The method of claim 3, wherein the data collected to build the mathematical model is in vivo, ex vivo , in Method for predicting the pharmacokinetic properties of a peptide sequence using a mathematical model, characterized in that the data obtained from at least one of the in vitro experiments. According to claim 3, wherein the data is collected in order to build the mathematical model, using the phage display technique in vivo , ex vivo , in Method for predicting the pharmacokinetic properties of a peptide sequence using a mathematical model, characterized in that the data obtained from at least one of the in vitro experiments. The method of claim 3, wherein the peptide sequence is a sequence consisting of 2 to 12 peptides. The method of claim 3, wherein the peptide sequence is a sequence consisting of 3 to 7 peptides. A method for predicting the pharmacokinetic properties of a peptide sequence using a mathematical model, characterized in that the species is a mammal applying the method for predicting the pharmacokinetic properties of the peptide sequence. A method for predicting the pharmacokinetic properties of a peptide sequence using a mathematical model, characterized in that the species is a human applying the method for predicting the pharmacokinetic properties of the peptide sequence. Obtaining a sample of peptide sequence through the small intestine using experimental techniques; Obtaining a sample of a sequence of peptides that do not pass through the small intestine based on these sequences; Storing each of the obtained samples as a set and then randomly extracting them in a predetermined ratio to classify them into a mathematical model training set and a mathematical model verification set; Assigning a descriptor value and an activity value to the individual peptide sequence; Training using training peptide sets and mathematical models; A process of predicting whether the small intestine passes through a set of peptides for verification using a trained mathematical model; A recording medium storing a program for predicting pharmacokinetic properties of a peptide sequence using a mathematical model, characterized in that it comprises a process of verifying a trained mathematical model.

Images