Evaluation method and evaluation system of metabolic memory

Abstract

To provide a method and a system, which evaluate accumulation (metabolic memory) of past hyperglycemia states in an object by a simple method.SOLUTION: An evaluation method includes: an acquisition step of acquiring amino acid density data related to a density value of amino acid in blood collected from an object; and an evaluation step of predicting a score value showing accumulation of past hyperglycemia states of the object on the basis of density values of a plurality of specified amino acid included in the amino acid density data of the object acquired in the acquisition step and evaluating the accumulation of the past hyperglycemia states of the object on the basis of the score value. An evaluation system is provided using a neural network so as to execute the evaluation method.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a method for predicting metabolic memory in a subject using amino acid concentrations in the blood and a system for carrying out the same. The present invention also relates to a method for evaluating or predicting the risk of diabetic complications in a subject by estimating metabolic memory and a system for carrying out the same. [Background technology]

[0002] In diabetic patients, the concept of "metabolic memory" has been attracting attention in recent years, as it is believed that hyperglycemic states are remembered by the body and have a long-term effect on the onset and progression of diabetic complications. Specifically, when hyperglycemic states continue for a long time, oxidative stress increases in various cells and tissues in the body, causing denaturation and damage to DNA, proteins, lipids, etc., and such changes at the cellular level may not be completely reversed even when normal blood glucose levels are later restored. This is recorded as metabolic memory, and it is said that it continues to increase the risk of diabetes-related complications in the long term. This means the importance of early treatment when diabetes is diagnosed, and it is thought that if early hyperglycemic control is appropriate, the effects of metabolic memory can be minimized, which is expected to reduce the risk of long-term complications in diabetic patients.

[0003] Thus, the existence of metabolic memory and its influence on the onset and progression of complications are well known, but its quantification was reported for the first time by a group including the present inventors (Non-Patent Document 1: Takuro Shigaki et al., "Quantitative Evaluation of Metabolic Memory Using Modeling & Simulation Analysis," presented at the 141st Annual Meeting of the Pharmaceutical Society of Japan, March 26-29, 2021). There, a method for estimating metabolic memory based on the patient's HbA1c over the past 10 years is shown. The accumulation of hyperglycemic states is expressed as the area under the HbA1c curve, and the quantitative relationship with reduced renal function is modeled using a nonlinear mixed effects model (NONMEM). Specifically, a model is shown that calculates the area under the HbA1cM6% curve (AUCHbAlc ≧ 6%) based on the time-course of HbA1c for each patient, and predicts the amount of change in eGFR using each patient factor of gender, age, and diabetic retinopathy as covariates.

[0004] However, this method requires detailed information on the long-term trends in HbA1c, which makes it impractical for clinical use. Therefore, a simpler method for estimating metabolic memory is needed.

[0005] Amino acids, which play a central role in metabolic pathways among metabolites in the body, have been attracting attention as new biomarkers, and disease risk is being predicted based on the amino acid concentrations in blood.

[0006] For example, findings have been reported that predict the onset of diabetes four years later using amino acid concentrations in subjects undergoing health checkups (Non-Patent Document 2: Minoru Yamakado: Health Checkup 32(5):713-725(2018)). In this study, the discriminatory ability of each amino acid was analyzed for subjects who developed diabetes within four years of the test and subjects who had diabetes at the time of the test, and subjects who did not develop diabetes within four years. As a result, subjects who developed diabetes within four years had an amino acid balance similar to that of subjects with diabetes, and therefore a multivariate equation of plasma amino acid concentrations that correlate with visceral fat area, which is an index of visceral fat accumulation, which is a precursor to diabetes, was created as a diabetes risk equation. The diabetes risk equation includes the amino acids asparagine, glycine, alanine, valine, tyrosine, and tryptophan, and it is disclosed that alanine, valine, tyrosine, and tryptophan show significantly higher values ​​in subjects who developed diabetes within four years compared to subjects who did not develop diabetes within four years, while glycine shows significantly lower values. However, this was an assessment of the risk of developing diabetes, and was not a report predicting metabolic memory in patients after they had developed diabetes.

[0007] In addition, a healthcare data analysis method for predicting the risk of developing diabetes based on the change in HbA1c in a subject has been reported (Patent Document 1: JP 2020-135489 A). However, this also evaluates the risk of developing diabetes, and is not a report predicting metabolic memory in patients after they have developed diabetes.

[0008] In addition, in relation to diabetes, for example, the following reports have been made. It has been reported that the concentrations of branched-chain amino acids (isoleucine, leucine, valine) and aromatic amino acids (phenylalanine, tyrosine) are associated with the risk of developing diabetes (Non-Patent Document 3: Ahola-Olli AV, et al. Diabetologia, (2019) 62(12):2298-2309). It has been reported that there is a strong association between branched-chain amino acids and insulin resistance (Non-Patent Document 4: Newgard C, et al. Cel Metab, (2012) 15(5):606-14). It has been reported that tyrosine concentrations are associated with a reduced risk of microangiopathy (Non-Patent Document 5: Welsh P., et al. Diabetologia, (2018) 61:1581-1591).

[0009] Furthermore, for example, a method for relating amino acid concentration to a biological state has been disclosed (Patent Documents 2-4). Also disclosed are a method for evaluating the state of metabolic syndrome using amino acid concentration (Patent Document 5), a method for evaluating the state of visceral fat accumulation using amino acid concentration (Patent Document 6), a method for evaluating the state of impaired glucose tolerance using amino acid concentration (Patent Document 7), a method for evaluating at least one state of apparent obesity, hidden obesity, and obesity defined by BMI (Body Mass Index) and VFA (Visceral Fat Area) using amino acid concentration (Patent Document 8), and a method for evaluating the state of fatty liver disease including at least one of fatty liver, NAFLD (non-alcoholic fatty liver disease), and NASH (non-alcoholic steatohepatitis) using amino acid concentration (Patent Document 9). [Prior art documents] [Patent documents]

[0010] [Patent Document 1] Patent Publication No. 2020-135489 [Patent Document 2] International Publication No. 2004 / 052191 [Patent Document 3] International Publication No. 2006 / 098192 [Patent Document 4] International Publication No. 2009 / 054351 [Patent Document 5] International Publication No. 2008 / 015929 [Patent Document 6] International Publication No. 2009 / 001862 [Patent Document 7] International Publication No. 2009 / 054350 [Patent Document 8] International Publication No. 2010 / 095682 [Patent Document 9] International Publication No. 2013 / 002381 [Non-patent literature]

[0011] [Non-Patent Document 1] Takuro Shigaki et al., "Quantitative evaluation of metabolic memory using modeling and simulation analysis," presented at the 141st Annual Meeting of the Pharmaceutical Society of Japan, March 26-29, 2021 [Non-Patent Document 2] Yamakado, M.: Human Dock 32(5):713-725(2018) [Non-Patent Document 3] Ahola-Olli AV, et al. Diabetologia, (2019) 62(12):2298-2309 [Non-Patent Document 4] Newgard C, et al. Cel Metab, (2012) 15(5):606-14 [Non-Patent Document 5] Welsh P., et al. Diabetologia, (2018) 61:1581-1591 Summary of the Invention [Problem to be solved by the invention]

[0012] An object of the present invention is to provide a method for predicting metabolic memory by a simple method. Another object of the present invention is to provide a method for predicting the risk of diabetic complications by predicting metabolic memory. [Means for solving the problem]

[0013] The present inventors have found a simple method for quantitatively evaluating metabolic memory stored in the body due to the continued hyperglycemic state of a diabetic patient, and have completed the present invention. Specifically, they have found that the accumulation of past hyperglycemic states (metabolic memory) can be predicted by measuring the amino acid concentrations in blood (e.g., plasma or serum) collected from the subject to be evaluated and using the weight ratio of each amino acid in a specific amino acid group. The predicted results of the accumulation of past hyperglycemic states can predict the risk of complications that may develop in the patient, and therefore the present invention is useful in the prognosis and prevention of patients and in determining treatment plans.

[0014] One embodiment of the present invention is a method for assessing the accumulation of a subject's past hyperglycemic state, which uses data on amino acid concentrations in the subject's blood (e.g., plasma or serum), and includes the following embodiments. [1] An acquisition step of acquiring amino acid concentration data relating to concentration values ​​of amino acids in blood collected from a subject; an evaluation step of predicting a score value representing the accumulation of a hyperglycemic state of the subject based on concentration values ​​of a specific plurality of amino acids included in the amino acid concentration data of the subject acquired in the acquisition step, and evaluating the accumulation of a hyperglycemic state of the subject based on the score value; An evaluation method comprising: [2] The score value is the area under the HbA1cM6% curve (AUC HbAlc≧6% ) is a value equivalent to the evaluation method described in [1] above. [3] The evaluation method according to [1] or [2] above, wherein in the evaluation step, the score value is calculated using a neural network. [4] The evaluation method according to [1] or [2] above, wherein in the evaluation step, the score value is calculated using a multiple regression equation consisting of a concentration value of the amino acid and a plurality of variables into which the concentration values ​​of the plurality of amino acids are separately substituted. [5] The evaluation method according to [1] or [2] above, wherein in the evaluation step, the score value is calculated using clustering analysis. [6] The evaluation method according to any one of the above [1] to [5], wherein the subject is a female (preferably a female suffering from diabetes). [7] The evaluation method according to any one of the above [1] to [6], wherein the specific amino acids are valine, arginine, methionine, asparagine, aspartic acid and citrulline. [8] The evaluation method described in [7] above, wherein the specific amino acids further include glutamine, glutamic acid, and tyrosine. [9] The evaluation method according to the above [8], wherein the specific amino acids further include isoleucine.

[10] The evaluation method according to [9] above, wherein the specific amino acids are 20 types of amino acids: alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine and ornithine.

[11] The evaluation method according to any one of the above [7] to

[10] , wherein in the evaluation step, the total concentration value of asparagine and aspartic acid is used as a single concentration value.

[12] The evaluation method according to any one of the above [8] to

[11] , wherein in the evaluation step, the concentration values ​​of glutamine and glutamic acid are expressed as a glutamine / glutamic acid ratio.

[13] The evaluation method according to any one of the above [1] to [5], wherein the subject is a male (preferably a male suffering from diabetes).

[14] The evaluation method according to

[13] above, wherein the specific amino acids are alanine, phenylalanine, glutamine, glutamic acid, histidine and tryptophan.

[15] The evaluation method described in

[14] above, wherein the specific amino acids further include ornithine, valine, and proline.

[16] The evaluation method according to the above

[15] , wherein the specific amino acids further include asparagine and aspartic acid.

[17] The evaluation method according to

[16] above, wherein the specific amino acids are 20 types of amino acids: alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine and ornithine.

[18] The evaluation method according to any one of the above

[14] to

[17] , wherein in the evaluation step, the concentration values ​​of glutamine and glutamic acid are expressed as a glutamine / glutamic acid ratio.

[19] The evaluation method according to any one of the above

[16] to

[18] , wherein in the evaluation step, the total concentration value of asparagine and aspartic acid is used as a single concentration value.

[20] The evaluation method according to any one of [1] to

[18] above, further comprising a prediction step of predicting the risk of developing diabetic complications based on a predetermined evaluation index using a score value representing the accumulation of the past hyperglycemic state in the evaluation step.

[21] The evaluation method according to

[20] above, wherein the complication is a disease selected from the group consisting of nephropathy, retinopathy, coronary artery disease, arteriosclerosis obliterans, cerebral infarction, and neuropathy.

[0015] One aspect of the present invention is a system for evaluating the accumulation of a subject's past hyperglycemic states, comprising a calculation unit that inputs input information about the subject into a neural network to estimate the accumulation of the subject's past hyperglycemic states, the input information including information about concentration value data of amino acids in the subject's blood (e.g., in plasma or serum), the neural network uses the same type of information as the input information (information about the concentration value data of amino acids) as learning data, and generates a training data by machine learning using the results of quantifying the accumulation of past hyperglycemic states as teacher data. The system of the present invention includes the following aspects.

[22] A calculation unit that inputs input information regarding the concentration value of a target amino acid into a neural network to estimate the accumulation of the subject's past hyperglycemic state (metabolic memory), The input information includes information regarding concentration values ​​of specific amino acids in blood collected from a subject, The neural network is an evaluation system that uses the same type of information as the input information as learning data and the results of quantifying the accumulation of past hyperglycemic states as training data, and is generated by machine learning to evaluate the accumulation of past hyperglycemic states in a subject.

[23] The result of quantifying the accumulation of the past hyperglycemic state is the area under the HbA1cM6% curve (AUC HbAlc≧6% ) is a value equivalent to the evaluation system described in

[22] above.

[24] The evaluation system described in

[22] or

[23] above, wherein the subject is a woman (preferably a woman suffering from diabetes).

[25] The evaluation system according to any one of the above

[22] to

[24] , wherein the specific amino acids are valine, arginine, methionine, asparagine, aspartic acid and citrulline.

[26] The evaluation system described in

[25] above, wherein the specific amino acids further include glutamine, glutamic acid, and tyrosine.

[27] The evaluation system according to

[26] above, wherein the specific amino acids further include isoleucine.

[28] The evaluation system described in

[27] above, wherein the specific amino acids are 20 types of amino acids: alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine, and ornithine.

[29] The evaluation system according to any one of

[25] to

[28] above, wherein the calculation unit uses, in the input information, the concentration values ​​of asparagine and aspartic acid as a single concentration value, the sum of these values.

[30] The evaluation system according to any one of

[26] to

[29] above, wherein the calculation unit uses a glutamine / glutamic acid ratio as a single concentration value for the concentration values ​​of glutamine and glutamic acid in the input information.

[31] The evaluation system described in

[22] or

[23] above, wherein the subject is a male (preferably a male suffering from diabetes).

[32] The evaluation system according to

[31] above, wherein the specific amino acids are alanine, phenylalanine, glutamine, glutamic acid, histidine and tryptophan.

[33] The evaluation system described in

[32] above, wherein the specific amino acids further include ornithine, valine, and proline.

[34] The evaluation system according to

[33] above, wherein the specific amino acids further include asparagine and aspartic acid.

[35] The evaluation system according to the above

[34] , wherein the specific amino acids are 20 types of amino acids: alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine and ornithine.

[36] The evaluation system according to any one of

[32] to

[35] above, wherein the calculation unit uses a glutamine / glutamic acid ratio as a single concentration value for the concentration values ​​of glutamine and glutamic acid in the input information.

[37] The evaluation system according to any one of

[34] to

[36] above, wherein the calculation unit uses, in the input information, the concentration values ​​of asparagine and aspartic acid as a single concentration value, the sum of these values.

[38] The evaluation system according to any one of

[22] to

[37] above, and a prediction unit for predicting a risk of developing diabetic complications in a subject based on a predetermined evaluation index, using a value indicating an accumulation of past hyperglycemic states acquired in the evaluation system; and a disease risk prediction system comprising:

[39] The system for predicting the risk of developing a complication described in

[38] above, wherein the complication is a disease selected from the group consisting of nephropathy, retinopathy, coronary artery disease, arteriosclerosis obliterans, cerebral infarction, and neuropathy. Effect of the Invention

[0016] The evaluation method and / or evaluation system of the present invention can evaluate the accumulation of past hyperglycemic states in a subject, thereby making it possible to predict the risk of diabetic complications in the subject. [Brief description of the drawings]

[0017] [Figure 1] This shows a model for quantifying the accumulation of past hyperglycemic states reported by a group including the present inventor. It shows the transition of HbA1c in subjects (diabetic patients) during the observation period (past 10 years), and the colored part shows the area under the curve for HbA1c ≧ 6% (AUCHbA1c ≧ 6%). [Diagram 2] The figure shows a conceptual diagram of a prediction device constituting the evaluation system of the present invention. [Diagram 3] 1 shows the correlation between AUCHbAlc ≥ 6% and decreased renal function in diabetic patients (left: male, right: female). [Figure 4]FIG. 2 shows the correlation between the area under the HbA1cM6% curve (AUCHbAlc≧6%) (actual AUC value) in the HbA1c transition over the past 10 years for the subject shown in FIG. 1 and the score value (predicted AUC value by amino acid) estimated by the method of the present invention using concentration data of 20 types of amino acids and a neural network based on the amino acid concentration data, and shows the results for female patients. [Diagram 5] Similar to Fig. 4, this figure shows the correlation between the actual measurement value (AUC) and the predicted value (predicted value) of AUCHbAlc ≥ 6% by the method of the present invention using 20 kinds of amino acid concentration data, and is the result for female patients. However, for asparagine and aspartic acid, the sum of the two values, and for glutamine and glutamic acid, the ratio of the two values, were used as one amino acid concentration data, respectively. [Figure 6] Similar to Fig. 4, this figure shows the correlation between the actual measurement value (AUC) and the predicted value (predicted value) of AUCHbAlc ≥ 6% by the method of the present invention using 20 kinds of amino acid concentration data, and is the result for male patients. However, for asparagine and aspartic acid, the sum of the two values, and for glutamine and glutamic acid, the ratio of the two values, were used as one amino acid concentration data, respectively. [Figure 7] 1 shows the results of clustering analysis performed by the method of the present invention using concentration data of 20 types of amino acids. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] The present invention will be described below with exemplary embodiments, together with preferred methods, materials, and devices and systems that can be used in the practice of the present invention, but the present invention is not limited to the embodiments described below. Unless otherwise specified in the text, all technical and scientific terms used in this specification have the same meaning as commonly understood by those skilled in the art to which the present invention belongs. In addition, any materials, methods, and devices and systems equivalent or similar to those described in this specification can be used in the practice of the present invention. In addition, all publications and patents cited in this specification in relation to the present invention are cited in this specification and constitute a part of this specification, for example, as showing methods, materials, devices, systems, etc. that can be used in the present invention.

[0019] A group including the present inventors has reported a method for estimating metabolic memory based on the patient's past changes in HbA1c over time (Non-Patent Document 1). Specifically, as shown in FIG. 1, the area under the curve for HbA1c ≧ 6% (AUC HbA1c≧6% ) was calculated.

[0020] And the accumulation of hyperglycemic states (AUC HbA1c≧6% The quantitative relationship between the mean mean and renal function decline was modeled using a nonlinear mixed-effects model (NONMEM). AUC HbA1c≧6% In addition, the patient's eGFR baseline The following model was constructed to predict the change in eGFR using the following factors as covariates: diabetic retinopathy, hypertension, dyslipidemia, sex, and age. max indicates the maximum decrease in eGFR, and AUC50 indicates the maximum decrease in eGFR. max AUC at 50% of the mean HbA1c≧6% Shows.

[0021]

number

[0022] In this model, the AUC HbA1c≧6%The eGFR decreased nonlinearly with increasing blood glucose level, indicating the importance of the degree and duration of hyperglycemia exposure. HbA1c≧6% The change in the relationship between AUC and eGFR reflects the progress of clinical practice. HbA1c≧6% However, the evaluation method using this model requires detailed data on HbA1c levels over a long period of time (10 years).

[0023] eGFR (estimated glomerular filtration rate) is an indicator of kidney function, and predicting the change in eGFR can be used to predict kidney health, such as the progression of kidney disease, the effectiveness of kidney disease treatment, and the risk of kidney-related complications (e.g., cardiovascular disease).

[0024] In this specification, "metabolic memory", "accumulation of past hyperglycemic states" and "accumulation of hyperglycemic states" are used as interchangeable terms. In this specification, when "accumulation of hyperglycemic states" is mentioned, unless it is clear that a different meaning is meant from the context, it means the accumulation of hyperglycemic states in the past (preferably over the past 10 years) of a subject (particularly a diabetic patient), and is a numerical value expressed as the area under the curve for HbA1c≧6% calculated based on the time course of HbA1c in a subject (patient) over the past 10 years (AUC HbA1c≧6% For example, the AUC for patients who maintained HbA1c at 7% for 10 years was HbA1c≧6% is 10.0%·year.

[0025] As used herein, "diabetic complications" or "complications" refers to diseases that develop in diabetic patients when they are not properly managed over the long term, and examples of such diseases include, but are not limited to, diabetic retinopathy, diabetic nephropathy, diabetic neuropathy, cardiovascular disease, as well as lower limb complications, foot complications, dental complications, skin complications, and stomach complications (e.g., gastroparesis).

[0026] Complications whose risk of development can be predicted using the evaluation method of the present invention are not particularly limited as long as they are diseases that develop in diabetic patients when management is inappropriate over the long term, but preferred examples include nephropathy, retinopathy, cardiovascular disease, in particular coronary artery disease, arteriosclerosis obliterans, cerebral infarction, and neuropathy.

[0027] In this specification, various amino acids may be represented by abbreviations, but their official names are as follows: (Official name) (Abbreviation) Glycine...Gly Alanine...Ala Serine...Ser Threonine...Thr Asparagine...Asn Glutamine...Gln Aspartic acid...Asp Glutamic acid...Glu Lysine...Lys Arginine...Arg Histidine...His Valine...Val Leucine...Leu Isoleucine...Ile Tyrosine...Tyr Phenylalanine...Phe Tryptophan...Trp Proline...Pro Methionine...Met Cystine...Cys Citrulline...Cit Ornithine...Ornithine

[0028] The subject (to be evaluated) in the present invention may be a human or a non-human mammal, including a cat, a dog, a cow, or a horse, and is preferably a human. When the term "patient" is used herein, it is synonymous with a human subject, unless otherwise specified.

[0029] In the present invention, concentration data of amino acids in blood (e.g., plasma, serum, etc.) collected from the subject to be evaluated is used. Although any of whole blood, plasma, and serum can be used as blood, plasma or serum is preferable, and plasma is more preferable. In the evaluation method of the present invention, amino acid concentration data relating to the concentration values ​​of amino acids in blood collected from the subject to be evaluated is obtained (obtaining step). In the evaluation system of the present invention, amino acid concentration data relating to the concentration values ​​of amino acids in blood collected from the subject to be evaluated is input as input information to a calculation unit.

[0030] Amino acid concentration data may be data on the concentration of all amino acids contained in blood, or may be concentration data on a part of specific amino acids.For example, amino acids may include alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine, ornithine, serine and glycine.In the evaluation method or evaluation system of the present invention, the amino acid concentration data most preferably includes the concentration data of 20 kinds of amino acids, namely alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine and ornithine. When the subject is a woman, the amino acid concentration data preferably includes the concentration data of valine, arginine, methionine, asparagine, aspartic acid and citrulline, more preferably includes the concentration data of glutamine, glutamic acid and tyrosine, and more preferably includes the concentration data of isoleucine.When the subject is a man, the amino acid concentration data preferably includes the concentration data of alanine, phenylalanine, glutamine, glutamic acid, histidine and tryptophan, more preferably includes the concentration data of ornithine, valine and proline, and more preferably includes the concentration data of asparagine and aspartic acid.

[0031] In the present invention, concentration data is used for each individual amino acid, but preferably, for asparagine and aspartic acid, the sum of these is used as one amino acid concentration data, and for glutamine and glutamic acid, the glutamine / glutamic acid ratio is used as one amino acid concentration data.

[0032] The amino acid concentration data used in the acquisition step in the evaluation method of the present invention may be amino acid data of a subject that has already been measured by a company that provides amino acid concentration measurement as a service, or may be acquired by measuring blood collected from the subject to be evaluated, for example, by the measurement method shown below. Here, the unit of the amino acid concentration value may be, for example, molar concentration or weight concentration, or may be obtained by adding, subtracting, multiplying or dividing these concentrations by any constant. Measurement method 1: Plasma or serum is separated from the collected blood sample and frozen at -80°C until the time of measurement of amino acid concentration values. After thawing, protein is removed, pre-column derivatization is performed using a labeling reagent (3-aminopyridyl-N-hydroxysuccinimidyl carbamate), and amino acid concentration values ​​are analyzed by liquid chromatography mass spectrometry (LC / MS). Measurement method 2: Plasma or serum is separated from the collected blood sample and frozen at -80°C until the time of measurement of amino acid concentration values. After thawing, protein is removed and the amino acid concentration values ​​are analyzed using an amino acid analyzer based on the post-column derivatization method using ninhydrin reagent. In the evaluation system of the present invention, the above amino acid concentration data is used as input information.

[0033] Next, a score value representing the accumulation of the past hyperglycemic state of the subject is predicted based on the concentration values ​​of the specific amino acids included in the amino acid concentration data acquired in the acquisition step. The specific amino acids are most preferably 20 types of amino acids, namely, Ala, Phe, Pro, Val, Thr, Leu, Ile, Asn, Asp, Gln, Glu, Lys, Met, His, Arg, Cit, Tyr, Trp, Cys, and Orn. ​​When the amino acid concentration data is used, data such as missing values ​​and outliers may be removed from the acquired amino acid concentration data.

[0034] The amino acid concentration data used in the evaluation method or system preferably includes Val, Arg, Met, Asn, Asp, and Cit, more preferably includes Gln, Glu, and Tyr, and more preferably includes Ile, when the subject is a woman. Preferably, the amino acid concentration data includes Ala, Phe, Gln, Glu, His, and Trp, more preferably includes Orn, Val, and Pro, and more preferably includes Asn and Asp, when the amino acid concentration data includes Asn and Asp, it is preferable to use the sum of them as one amino acid concentration data. When the amino acid concentration data includes Gln and Glu, it is preferable to use the ratio (Gln / Glu) between them as one amino acid concentration data.

[0035] In one embodiment of the present invention, in the evaluation step, a machine learning algorithm can be used to predict a score value that represents the accumulation of the past hyperglycemic state of the subject to be evaluated, using the concentration value of a specific amino acid. The machine learning algorithm can include a supervised learning algorithm. Examples of supervised learning algorithms can include average-one dependent estimator (AODE), neural network (e.g., backpropagation), Bayesian statistics (e.g., naive Bayesian classifier, Bayesian network, Bayesian knowledge base), case-based reasoning, decision tree, inductive logic programming, Gaussian process regression, group method of data processing (GMDH), learning automaton, learning vector quantization, minimum message length (decision tree, decision graph, etc.), lazy learning, instance-based learning nearest neighbor algorithm, analogical modeling, probabilistic approximately correct learning (PAC) learning, ripple down rule, knowledge acquisition methodology, symbolic machine learning algorithm, sub-symbolic machine learning algorithm, support vector machine, random forest, ensemble of classifiers, bootstrap aggregation (bagging), and boosting. Supervised learning may include ordinal classification such as regression analysis and information fuzzy networks (IFNs). Alternatively, supervised learning methods may include statistical classification such as AODE, linear classifiers (e.g., Fisher's linear discriminant, logistic regression, naive Bayes classifier, perceptrons, and support vector machines), quadratic classifiers, k-nearest neighbors, boosting, decision trees (e.g., C4.5, random forests), Bayesian networks, and hidden Markov models. Reinforcement learning concepts may include temporal difference learning, Q-learning, learning automata, and SARSA. Deep learning concepts may include deep belief networks, deep Boltzmann machines, deep convolutional neural networks, deep recurrent neural networks, and hierarchical temporal memories.

[0036] Machine learning algorithms may also include unsupervised learning algorithms. Examples of unsupervised learning algorithms may include neural networks, data clustering, expectation maximization algorithms, self-organizing maps, radial basis function networks, vector quantization, generative terrain maps, information bottleneck methods, and IBSEAD. Unsupervised learning may also include association rule learning algorithms such as Apriori algorithm, Eclat algorithm, and FP-growth algorithm. Hierarchical clustering such as single linkage clustering and conceptual clustering may also be used. Linkage methods such as shortest distance method, longest distance method, group average method, and Ward's method may be used in hierarchical clustering. Unsupervised learning may also include partial (or non-hierarchical) clustering such as K-means algorithm and fuzzy clustering.

[0037] The machine learning algorithm preferably used in the present invention is a neural network, particularly a supervised learning neural network. Examples of neural networks that can be used in the present invention include a multilayer perceptron, a ConvLSTM network that combines a CNN (Convolutional Neural Network) and a LSTM (Long short-term memory), a convolutional network composed of a CNN, and a composite neural network that combines a ConvLSTM network and a convolutional network.

[0038] The evaluation system 1 of the present invention includes a calculation unit 35 for estimating the accumulation of past hyperglycemic states (metabolic memory) of the evaluation subject, a recording unit 34 including a control program, a control unit 32 for controlling it, an input unit 31 for inputting input information, an output unit 33 for outputting prediction results, and a prediction unit device 3. The evaluation system 1 of the present invention may also include a terminal device 2 that can be electrically connected to the input unit 31. The evaluation system 1 of the present invention can constitute the onset prediction system of the present invention together with an onset prediction unit 4. The conceptual configuration of the prediction device 3 that constitutes the evaluation system 1 of the present invention is shown in FIG. 2.

[0039] The prediction device 3 has an input unit 31, a control unit 32, an output unit 33, and a storage unit 34. The input unit 31, the output unit 33, the control unit 32, and the storage unit 34 are electrically connected to one another. Input information I is input to the input unit from the terminal device 2. The control unit 32 executes a control program to calculate, in the calculation unit 35, a score value of the accumulation (metabolic memory) of the subject's past hyperglycemic state (area under the HbA1cM6% curve (AUC HbAlc≧6% ) can be estimated. The output unit 33 can output the estimation result O estimated by the calculation unit 35. The storage unit 34 stores a control program and various data, parameters, etc. required for control.

[0040] The prediction device 3 of the present invention has a plurality of electronic components and circuits. The plurality of electronic components may be active elements such as transistors or diodes, or passive elements such as capacitors, and may be formed by a conventional method.

[0041] The input unit receives input information I. The input unit may have a communication unit so that the input information I acquired by a terminal device 2 described later is directly input from the terminal device 2. The input unit may also include an input device capable of inputting the input information I or other information. The input device may be, for example, a keyboard, a touch panel, or a mouse.

[0042] The controller may manage the operation of the prediction device 3 by controlling the other components of the prediction device 3. The controller is a control device or control circuit that includes at least one processor, e.g., a CPU, to provide control and processing power to perform various functions.

[0043] The storage unit includes a non-transitory recording medium that can be read by the CPU of the control unit, such as a ROM and a RAM. The storage unit stores a control program for controlling the prediction device 3, and various functions of the control unit are realized by executing the control program.

[0044] The control unit executes a control program in the memory unit, thereby making it possible to estimate the prediction result O. The onset prediction unit 4 includes, for example, a neural network. In addition to the control program, the memory unit stores learned parameters related to the neural network, estimation data (hereinafter also referred to as "input information"), learning data, and teacher data. The learning data and teacher data are data used when training the neural network. The learned parameters and estimation data are data used when the trained neural network estimates a score value of the accumulation (metabolic memory) of the subject's past hyperglycemic states.

[0045] The calculation unit can estimate a score value of the accumulation (metabolic memory) of the subject's past hyperglycemic state from the input information I input to the input unit. The calculation unit has AI (Artificial Intelligence). The AI ​​may be, for example, a neural network.

[0046] The calculation unit has undergone a learning process in advance. That is, by applying machine learning to the calculation unit using the learning data and the teacher data, the calculation unit can calculate the estimation result O from the input information I. Note that the learning data or the teacher data may be data corresponding to the input information I input to the prediction device 3 and the estimation result O output from the prediction device 3.

[0047] The terminal device 2 can be electrically connected to the prediction device 3. The terminal device 2 may be, for example, a measurement device for acquiring amino acid concentration data in the blood of a subject, or may be a device for receiving amino acid concentration data of a subject provided by a company that provides an amino acid concentration measurement service. The amino acid concentration data of a subject acquired by the terminal device is sent to an input unit of the prediction device 3 that is electrically connected thereto.

[0048] The evaluation system of the present invention may be composed of only the prediction device 3, or may further include a terminal device 2. The evaluation system of the present invention can be electrically connected to an onset prediction unit 4 that predicts the risk of developing a complication, which will be described later, and can be combined with the onset prediction unit 4 to form an onset prediction system.

[0049] <Examples of input information, learning data, and teacher data> The input information used in the evaluation system of the present invention is amino acid concentration data related to the concentration of amino acids contained in the blood of a subject. Examples of the amino acid concentration data include concentration data of 20 kinds of amino acids, including alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine, and ornithine. The type of amino acid in the amino acid concentration data can be selected arbitrarily and can be appropriately selected according to the gender of the subject, and the preferred combination of types when the subject is female and when the subject is male is as described above. The amino acid concentration data may be subjected to any processing in order to be used as input information. The amino acid concentration data can be subjected to any processing when used as input information I. For example, but not limited to, standardization in which the average value is converted to 0 and the variance to 1, and normalization in which the value is converted to the range of 0 to 1 (0% to 100%) can be performed.

[0050] The learning data is the same data as the input information I, and may be acquired using the terminal device 2, or already acquired information may be input from the input unit. The learning information is the area under the HbA1cM6% curve (AUC HbAlc≧6% ) are amino acid concentration data for subjects for which values ​​are available.

[0051] The training data is the area under the HbA1cM6% curve (AUC HbAlc≧6% The training data includes the AUC HbAlc≧6% value.

[0052] <Neural network learning example> The calculation unit is optimized by machine learning using learning data and teacher data so that it can calculate the estimation result O from the input information I. That is, the control unit calculates the estimation result O from the input information I based on an approximation formula (calculation unit) optimized by machine learning, and the control unit calculates the estimation result O from the input information I based on the approximation formula (calculation unit), and machine learning is performed by adjusting parameters in the calculation unit so that the difference between the pseudo estimation result calculated from the learning data input to the input unit and output from the output unit and the teacher data is small. As a result, the calculation unit can perform calculations on the input information I based on the learned parameters and output the estimation result O.

[0053] For example, the backpropagation algorithm is used as a parameter adjustment method. The parameters include, for example, weighting coefficients used in the ConvLSTM layer, the convolution layer, and the fully connected layer.

[0054] As described above, the evaluation system 1 outputs the calculation result O from the input information I, thereby obtaining the area under the HbA1cM6% curve (AUC HbAlc≧6% ) can be estimated.

[0055] <Example using a linear regression model> In one embodiment of the present invention, in the evaluation step, a score value representing the accumulation of a hyperglycemic state in the past of the subject can be predicted by using a linear regression model (preferably a multiple regression model), i.e., using an equation including a concentration value of a specific amino acid and a variable into which the concentration value of the specific amino acid is substituted, to calculate the value of the equation. Based on the predicted score value, the accumulation of a hyperglycemic state in the past of the subject can be evaluated. For example, when 20 types of amino acid concentration data are used, the formula is expressed as follows: a1×AA1+b1×AA2+c1×AA3+d1×AA4+e1×AA5+f1×AA6+g1...t1×AA20...(Formula 1) In formula 1, a1, b1, c1, d1, e1, f1...t1 are any real numbers other than zero, and AA1, AA2,...AA20 are individual amino acids used as amino acid data.

[0056] The model formula may be any one of a logistic regression formula, a fractional formula, a linear discriminant formula, a multiple regression formula, a formula created by a support vector machine, a formula created by the Mahalanobis distance method, a formula created by canonical discriminant analysis, and a formula created by a decision tree.

[0057] In one embodiment, the present invention is a method for obtaining information for evaluating the accumulation of hyperglycemic states (metabolic memory) in the blood of a female subject, comprising: obtaining concentration values ​​of valine, arginine, methionine, asparagine, aspartic acid and citrulline in the blood of the subject, wherein the concentration values ​​of asparagine and aspartic acid are calculated by using the sum of the amino acids as a concentration value of one amino acid, and a multiple regression equation consisting of the five variables into which the concentration values ​​of the five amino acids are separately substituted, and using the value of the regression equation to obtain information for evaluating the accumulation of hyperglycemic states (metabolic memory) in the past for the subject.Preferably, the method further includes obtaining concentration values ​​of glutamine, glutamic acid and tyrosine, wherein the concentration values ​​of glutamine and glutamic acid are calculated by using the glutamine / glutamic acid ratio as a concentration value of one amino acid, and using the value of the regression equation to obtain information for evaluating the metabolic memory of the subject, wherein the concentration values ​​of glutamine and glutamic acid are calculated by using the multiple regression equation consisting of the seven variables into which the concentration values ​​of the seven amino acids are separately substituted, and using the value of the regression equation to obtain information for evaluating the metabolic memory of the subject. More preferably, the method further comprises obtaining a concentration value of isoleucine, and using the value of the regression equation calculated using a multiple regression equation consisting of the eight variables into which the concentration values ​​of the eight amino acids are separately substituted, to obtain information for evaluating metabolic memory of the subject to be evaluated.Even more preferably, the method further comprises obtaining concentration values ​​of alanine, phenylalanine, proline, threonine, leucine, lysine, histidine, tryptophan, cystine and ornithine, and using the value of the regression equation calculated using a multiple regression equation consisting of the 18 variables into which the concentration values ​​of the 18 amino acids are separately substituted, to obtain information for evaluating metabolic memory of the subject to be evaluated.

[0058] In another aspect, the present invention is an acquisition method comprising the steps of: acquiring concentration values ​​of alanine, phenylalanine, glutamine, glutamic acid, histidine, tryptophan, ornithine, valine, proline, asparagine and aspartic acid in the blood of the subject to be evaluated when the subject is male; wherein the concentration values ​​of asparagine and aspartic acid are defined as the sum of the amino acids as a single amino acid concentration value, and the concentration values ​​of glutamine and glutamic acid are defined as the glutamine / glutamic acid ratio as a single concentration value; and acquiring information for evaluating metabolic memory for the subject to be evaluated using values ​​of the regression equation calculated using a multiple regression equation consisting of the nine variables into which the concentration values ​​of the nine amino acids are separately substituted. More preferably, the method further comprises obtaining concentration values ​​of threonine, leucine, isoleucine, lysine, methionine, arginine, citrulline, tyrosine and cystine, and obtaining information for evaluating metabolic memory in the subject to be evaluated using the value of the regression equation calculated using a multiple regression equation consisting of the 18 variables into which the concentration values ​​of the 18 amino acids are separately substituted.

[0059] In one embodiment, the present invention is a calculation method characterized by comprising: obtaining concentration values ​​of valine, arginine, methionine, asparagine, aspartic acid and citrulline in the blood of the subject when the subject is a woman, wherein the concentration values ​​of asparagine and aspartic acid are the sum of the amino acids, and a multiple regression equation for evaluating the metabolic memory of the subject in the past, which is composed of the five variables into which the concentration values ​​of the five amino acids are separately substituted, to calculate the value of the regression equation.Preferably, the calculation method further includes obtaining concentration values ​​of glutamine, glutamic acid and tyrosine, wherein the concentration values ​​of glutamine and glutamic acid are the glutamine / glutamic acid ratio, and a multiple regression equation for evaluating the metabolic memory of the subject, which is composed of the seven variables into which the concentration values ​​of the seven amino acids are separately substituted, to calculate the value of the regression equation. More preferably, the method further comprises obtaining a concentration value of isoleucine, and calculating the value of the regression equation using a multiple regression equation for evaluating metabolic memory in the subject, the multiple regression equation consisting of the eight variables into which the concentration values ​​of the eight amino acids are separately substituted.More preferably, the method further comprises obtaining concentration values ​​of alanine, phenylalanine, proline, threonine, leucine, lysine, histidine, tryptophan, cystine and ornithine, and calculating the value of the regression equation using a multiple regression equation for evaluating metabolic memory in the subject, the multiple regression equation consisting of the eighteen variables into which the concentration values ​​of the eighteen amino acids are separately substituted.

[0060] In one aspect, the present invention is a calculation method characterized by including a calculation step of obtaining concentration values ​​of alanine, phenylalanine, glutamine, glutamic acid, histidine, tryptophan, ornithine, valine, proline, asparagine and aspartic acid in the blood (preferably serum) of the subject to be evaluated when the subject is a male, wherein the concentration values ​​of asparagine and aspartic acid are defined as the sum of the amino acids, and the concentration values ​​of glutamine and glutamic acid are defined as the glutamine / glutamic acid ratio, and a multiple regression equation for evaluating metabolic memory in the subject to be evaluated, which is composed of nine variables into which the concentration values ​​of the nine amino acids are separately substituted, to calculate the value of the regression equation. More preferably, the method further comprises obtaining concentration values ​​of threonine, leucine, isoleucine, lysine, methionine, arginine, citrulline, tyrosine and cystine, and calculating the value of the regression equation using a multiple regression equation for evaluating metabolic memory in the subject, the multiple regression equation consisting of 18 variables into which the concentration values ​​of the 18 amino acids are separately substituted.

[0061] <Example using clustering analysis> In one embodiment of the present invention, in the evaluation step, a score value representing the accumulation of a hyperglycemic state in the past of the subject can be predicted using clustering analysis. The accumulation of a hyperglycemic state in the past of the subject can be evaluated based on the predicted score value. AUC HbAlc≧6% Using the above method, various amino acids can be clustered, and the concentration of a specific number of amino acids is estimated using groups classified by clustering analysis. Clustering analysis can also be used in combination with subgroup analysis. The divided groups can also be classified according to the type of amino acid and factors associated with it. Examples of factors include, but are not limited to, dietary habits, coexisting diseases, and currently administered therapeutic drugs.

[0062] The score value representing the accumulation of the past hyperglycemic state of the subject calculated in the evaluation method or evaluation system of the present invention is preferably the area under the HbA1cM6% curve (AUC HbAlc≧6% The AUC HbAlc≧6% is shown to represent metabolic memory (accumulation of past hyperglycemic states) in a subject (Non-Patent Document 1). The score value is expressed as AUC HbAlc≧6% It may be a specific value of or may indicate the degree of the metabolic memory (for example, very high, high, medium, low, etc.). Therefore, by calculating the metabolic memory using the present invention, the risk of developing complications in a subject can be predicted.

[0063] In one aspect, the present invention is also an evaluation method including a prediction step of predicting the risk of developing diabetic complications based on a predetermined evaluation index using a value (score value) indicating the accumulation of past hyperglycemic conditions in the evaluation step.

[0064] The evaluation index is the accumulation of hyperglycemia in diabetic patients (e.g., AUC HbAlc≧6% ) and complications that diabetic patients develop. For example, the present inventors have found that the relationship between renal function decline and AUC HbAlc≧6% It has been reported that there is a correlation between the increase in the value of the score and the risk of developing diabetes complications in a subject using the score calculated in the evaluation step of the present invention based on the predetermined evaluation index in this way.

[0065] In another aspect, the present invention is an onset prediction system that includes the above-mentioned evaluation system and a prediction unit that uses the accumulated value of past hyperglycemic states obtained by the evaluation system and predicts the risk of onset of diabetic complications in a subject based on a predetermined evaluation index.

[0066] In the present invention, the subject for evaluating or predicting the risk of developing diabetic complications is preferably a patient suffering from diabetes. EXAMPLES

[0067] The present invention will be specifically described below with reference to examples, but the present invention is not limited to the following examples.

[0068] Example 1 Based on the "Guidelines for Diabetes Diagnosis" in the Diabetes Treatment Guidelines 2019 published by the Japan Diabetes Society, 182 patients (average age 66.2 years; 135 male patients and 47 female patients) were diagnosed with type 2 diabetes. The amount of 20 types of serum amino acids (alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine, and ornithine) was measured using a capillary electrophoresis-mass spectrometry, and the weight ratio of each amino acid to the total amino acid amount was calculated. The sample used for amino acid measurement was prepared by first adding chloroform, methanol, and ultrapure water to 100 μL of patient serum, separating it into three phases: aqueous layer, protein layer, and organic layer, and extracting the aqueous layer. The aqueous layer was then deproteinized and evaporated to dryness under reduced pressure using an evaporator, after which it was dissolved in 30 μL of ultrapure water to prepare the sample for measurement. Capillary electrophoresis and mass spectrometry were performed according to standard methods and the protocols provided by the manufacturers.

[0069] (Example 2) Estimation using amino acid concentrations of 20 amino acids Next, we used a neural network, a machine learning method, to examine whether it is possible to predict AUCHbA1c ≥ 6% from the blood amino acid weight ratio, and calculated the coefficient of determination (R 2 Specifically, for the 182 type 2 diabetes patients who were subjects in Example 1, the area under the HbA1cM6% curve (AUC HbAlc≧6%) was calculated. For patients who had been previously controlled for glycemic control by medication (intensive therapy), the accumulation of past hyperglycemia (AUC HbAlc≧6% ) indicates a low value.

[0070] In the analysis using the neural network, the subjects were divided in a 7:3 ratio, and a prediction model was constructed using machine learning for 70% of the subjects (male: 96 in the training data; female: 34 in the training data). The predictive ability of the constructed model was then evaluated for all subjects. For the neural network, a method called multilayer perceptron was used. As the amino acid composition of the subjects differed between males and females, analysis was performed separately for each gender. As a result of the neural network, the AUC HbA1c≧6% The predictability of 2 =0.591). The results are shown in Figure 4 below. On the other hand, the predictive value for male patients was not high, but a certain correlation was suggested (R 2 =0.078).

[0071] (Example 3) Improvement of Amino Acid Concentration Data To improve the predictive power of metabolic memory, we made the following changes to the use of amino acid concentration data and used a neural network to calculate the AUC HbA1c≧6% We made predictions on the following. (Change 1) The amounts of asparagine and aspartic acid were added together and included in the calculation. This is because asparagine is easily converted to aspartic acid. (Change 2) Glutamine / glutamic acid (division) was added to the calculation. This is because previous reports have shown that the glutamine / glutamic acid ratio is related to the onset of diabetic retinopathy. By changing the above two points, predictions were made using a total of 18 amino acid amounts or ratios: alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine, ornithine, (asparagine + aspartic acid), and (glutamine / glutamic acid ratio).HbA1c≧6% was even more predictive (R 2 =0.708). The results are shown in Figure 5 below.

[0072] The importance of each amino acid was calculated for female patients, and the results are shown in the table below. The importance indicates how much the prediction results change when the weight ratio of each amino acid is changed, and the importance was converted to 0-100% by normalization for comparison.

[0073] [Table 1]

[0074] For female patients, amino acid indices with high importance were selected and predictions were performed again. The results are shown below. Total 18 types: R 2 =0.708 Top 5 important types (valine, arginine, methionine, asparagine + aspartic acid, citrulline): R 2 =0.467 Top 7 important types (valine, arginine, methionine, asparagine + aspartic acid, citrulline, glutamine / glutamic acid ratio, tyrosine): R 2 =0.543 Top 8 important types (valine, arginine, methionine, asparagine + aspartic acid, citrulline, glutamine / glutamic acid ratio, tyrosine, isoleucine): R 2 =0.672 The above results show that for women, the top eight amino acid indices showed a high degree of predictability that was almost equal to that of using all the indices. A certain degree of predictability was also confirmed for the top five and seven indices.

[0075] AUC by blood amino acid weight ratio in male patients HbA1c≧6% The predictability of was found to be improved by the changes (Change 1 and Change 2) (R 2 = 0.273). The results are shown in Figure 6.

[0076] Similarly, when the importance of each amino acid was calculated for male patients, the results are shown in the table below.

[0077] [Table 2]

[0078] For male patients, amino acid indices with high importance were selected and predictions were performed again. The results are shown below. Total 18 types: R 2 =0.273 Top 5 important types (alanine, phenylalanine, glutamine / glutamic acid ratio, histidine, tryptophan): R 2 =0.186 Top 8 important types (alanine, phenylalanine, glutamine / glutamic acid ratio, histidine, tryptophan, ornithine, valine, proline): R 2 =0.196 Top 9 important nutrients (alanine, phenylalanine, glutamine / glutamic acid ratio, histidine, tryptophan, ornithine, valine, proline, asparagine + aspartic acid): R 2 =0.235 The above results show that for men, the top nine amino acid indices showed predictive ability similar to that of using all the indices. A certain degree of predictive ability was also confirmed with the top five and eight indices.

[0079] (Example 4) Prediction by multiple regression analysis In multiple regression analysis, the dependent variable was the area under the HbA1cM6% curve (AUC HbAlc≧6% ) and select the weight ratio of each amino acid as the explanatory variable (forced input method), and AUC HbAlc≧6% A model was constructed to predict the following: The model was constructed using all cases, and its predictability was examined. As a result, in the same way as with the neural network, the AUC HbA1c≧6% The predictability of 2 =0.755), whereas in male patients, a certain degree of predictability was confirmed (R 2 =0.186).

[0080] (Example 5) Prediction by clustering analysis Targeting female patients, clustering analysis was performed using the K-means method using the concentration data of 20 types of amino acids measured in Example 1. The results of the clustering analysis are shown in FIG. As a result, the AUC HbAlc≧6% was high.

[0081] The above results show that by measuring the amino acid balance in a sample from a subject, it is possible to estimate the metabolic memory of the subject and predict the risk of diabetic complications in the subject. In particular, it was found that the metabolic memory of female type 2 diabetes patients can be estimated with high accuracy. Female type 2 diabetes patients are known to be at high risk of complications, and understanding past hyperglycemic states based on blood amino acids has been shown to be useful in predicting the prognosis and treatment of patients.

[0082] The above detailed description is merely illustrative of the objects and subject matter of the present invention and is not intended to limit the scope of the appended claims. Various modifications and substitutions to the described embodiments will be apparent to those skilled in the art from the teachings set forth herein without departing from the scope of the appended claims. [Industrial Applicability]

[0083] The present invention provides a simple method for estimating metabolic memory. Because metabolic memory affects the onset and progression of diabetic complications, the present invention is expected to contribute to reducing the risk of these complications. [Explanation of symbols]

[0084] 1. Rating System 2. Terminal Device 3. Prediction device 4. Onset Prediction Department 31. Input section 32. Control Unit 33. Output section 34.Memory part 35. Calculation section

Claims

1. An acquisition step of acquiring amino acid concentration data relating to concentration values ​​of amino acids in blood collected from a subject; an evaluation step of predicting a score value representing the accumulation of a hyperglycemic state of the subject based on concentration values ​​of a specific plurality of amino acids included in the amino acid concentration data of the subject acquired in the acquisition step, and evaluating the accumulation of a hyperglycemic state of the subject based on the score value; An evaluation method comprising:

2. The score value is the area under the HbA1cM6% curve (AUC) in the past time course of HbA1c in a subject. ＨｂＡｌｃ≧６％ 2. The evaluation method according to claim 1, wherein the value corresponds to the value of

3. The evaluation method according to claim 1 , wherein in the evaluation step, the score value is calculated using a neural network.

4. 3. The evaluation method according to claim 1, wherein in the evaluation step, the score value is calculated using a multiple regression equation consisting of the concentration value of the amino acid and a plurality of variables into which the concentration values ​​of the plurality of amino acids are separately substituted.

5. The evaluation method according to claim 1 , wherein in the evaluation step, the score value is calculated using a clustering analysis.

6. 3. The evaluation method according to claim 1 or 2, wherein the specific amino acids are 20 types of amino acids: alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine and ornithine.

7. 7. The evaluation method according to claim 6, wherein in the evaluation step, the concentration values ​​of asparagine and aspartic acid are calculated by using their sum as a single concentration value, and the concentration values ​​of glutamine and glutamic acid are calculated by using a glutamine / glutamic acid ratio as a single concentration value.

8. 3. The method according to claim 1, wherein the subject is a female, the specific amino acids are 10 types of amino acids, namely, valine, arginine, methionine, asparagine, aspartic acid, citrulline, glutamine, glutamic acid, tyrosine, and isoleucine, and in the evaluation step, the concentration values ​​of asparagine and aspartic acid are calculated by using a sum of these values ​​as a single concentration value, and the concentration values ​​of glutamine and glutamic acid are calculated by using a glutamine / glutamic acid ratio as a single concentration value.

9. 3. The evaluation method according to claim 1, wherein the subject is a male, the specific amino acids are alanine, phenylalanine, glutamine, glutamic acid, histidine, tryptophan, ornithine, valine, proline, asparagine and aspartic acid, and in the evaluation step, the concentration values ​​of asparagine and aspartic acid are calculated by using a sum of the concentration values ​​thereof as a single concentration value, and the concentration values ​​of glutamine and glutamic acid are calculated by using a glutamine / glutamic acid ratio as a single concentration value.

10. The evaluation method according to claim 1 or 2, further comprising a prediction step of predicting a risk of developing diabetic complications based on a predetermined evaluation index using a score value representing the accumulation of the past hyperglycemic state in the evaluation step.

11. a calculation unit that inputs input information on the concentration value of a target amino acid into a neural network to estimate the accumulation of the past hyperglycemic state (metabolic memory) of the target, The input information includes information regarding concentration values ​​of specific amino acids in blood collected from a subject, The neural network is an evaluation system that uses the same type of information as the input information as learning data and the results of quantifying the accumulation of past hyperglycemic states as training data, and is generated by machine learning to evaluate the accumulation of past hyperglycemic states in a subject.

12. The evaluation system according to claim 11 , wherein the result of quantifying the accumulation of the past hyperglycemic state is a value equivalent to an area under the HbA1cM6% curve (AUCHbAlc≧6%) in the past transition of HbA1c over time.

13. 13. The evaluation system according to claim 12, wherein the specific amino acids are 20 types of amino acids: alanine, phenylalanine, proline, valine, threonine, leucine, isoleucine, asparagine, aspartic acid, glutamine, glutamic acid, lysine, methionine, histidine, arginine, citrulline, tyrosine, tryptophan, cystine, and ornithine.

14. 14. The evaluation system according to claim 13, wherein the calculation unit uses, in the input information, the concentration values ​​of asparagine and aspartic acid as a single concentration value, and uses the glutamine / glutamic acid ratio as a single concentration value for the concentration values ​​of glutamine and glutamic acid.

15. 13. The evaluation system according to claim 11 or 12, wherein the subject is a female, the specific amino acids are ten types of amino acids, namely, valine, arginine, methionine, asparagine, aspartic acid, citrulline, glutamine, glutamic acid, tyrosine and isoleucine, and the calculation unit uses, in the input information, the concentration values ​​of asparagine and aspartic acid as a single concentration value, and uses the glutamine / glutamic acid ratio as a single concentration value for the concentration values ​​of glutamine and glutamic acid.

16. 13. The evaluation system according to claim 11 or 12, wherein the subject is a male, the specific amino acids are eleven types of amino acids, namely alanine, phenylalanine, glutamine, glutamic acid, histidine, tryptophan, ornithine, valine, proline, asparagine and aspartic acid, and the calculation unit uses, in the input information, the concentration values ​​of asparagine and aspartic acid as a single concentration value, and uses the glutamine / glutamic acid ratio as a single concentration value for the concentration values ​​of glutamine and glutamic acid.

17. 13. The evaluation system according to claim 11 or 12, and a prediction unit for predicting a risk of developing diabetic complications in a subject based on a predetermined evaluation index, using a value indicating an accumulation of past hyperglycemic states obtained by the evaluation system; and a disease risk prediction system comprising:

Images