WO2020251051A1 - Prediction method, prediction device and prediction program - Google Patents

Abstract

[Problem] To provide a prediction method whereby an intestinal environment-improving effect or a feces condition-improving effect by the intake of Bifidobacterium longum can be predicted, and a prediction device and a prediction program. [Solution] According to the present embodiment, an intestinal environment-improving effect or a feces condition-improving effect on a prediction subject by the intake of Bifidobacterium longum is predicted on the basis of the relative abundance ratio in a feces specimen, which is collected from the subject, of a bacterium present in the feces specimen and/or the content of a metabolite in the feces specimen in a preset amount of the feces specimen.

Description

Prediction method, prediction device and prediction program

The present invention relates to a prediction method, a prediction device and a prediction program.

Research on the intestinal flora based on omics analysis has been actively conducted in recent years. In particular, it has been clarified that indigenous bacteria in the gastrointestinal tract are closely linked to the diet and their composition fluctuates with reproducibility (Non-Patent Document 1). The effects of diet on the digestive tract via these indigenous bacteria are both negative and positive.

Examples of negative effects include the occurrence of impaired glucose tolerance due to changes in intestinal bacteria caused by excessive intake of artificial sweeteners (Non-Patent Document 2) and inflammatory effects due to excessive intake of emulsifiers (Non-Patent Document 2). Reference 3) and the like can be mentioned.

As an example of the positive effect, it is possible to improve the intestinal bacterial flora by taking some microorganisms, and it has been used as a probiotic for a long time (Non-Patent Document 4). For example, Bifidobacterium is a typical probiotic that is frequently used because of its effective intestinal regulation and harmlessness to the human body (Non-Patent Document 5). In addition, Bifidobacterium is used for both treatment and prevention of diseases, and in recent years, the molecular mechanism has been elucidated (Non-Patent Documents 6 to 8).

Here, as an effective test of the conventional bifidobacteria, improvement of constipation state due to intestinal regulation effect has been evaluated particularly (Non-Patent Document 9). Bacterial analysis of the stool of constipated patients based on the culture method has also confirmed that bifidobacteria increase and decrease, which is a microorganism of particular interest (Non-Patent Documents 10 and 11). However, a comprehensive analysis of the mechanism of action of Bifidobacterium on constipation has not been advanced so far.

In addition, it has been reported that the efficacy of the administered drug appears strongly in some subjects, and the mechanism is being elucidated with various drugs. For substances that act on the digestive tract, there is a movement to seek this cause from the intestinal flora, especially the second meal effect (the idea that the first meal affects the next). The mechanism has been actively studied (Non-Patent Document 12). The relationship between changes in blood glucose levels with respect to the barley diet and the genus Prevotella in the intestinal flora has been suggested in clinical trials through metagenomic analysis, and the effect of improving glucose tolerance using mice has been experimentally verified. (Non-Patent Document 13). Furthermore, a metabolome analysis of blood was also performed, and it was reported that the insulin response to the diet was changed by the intestinal flora (Non-Patent Document 14).

The analysis of such a mechanism is also advanced for human intestinal bacteria, and there are individual differences in the increase in blood glucose level between barley diet and wheat diet, and the tendency is accurate from the omics information of intestinal bacteria. It was suggested that it could be predicted (Non-Patent Documents 15 and 16). Analysis of individual differences targeting intestinal bacteria was also performed for anticancer agents, and the effects of immune checkpoint inhibitors targeting PD-1 / PD-L1 were shown in Bifidobacterium longum, Collincella aerofaciens, and Enterococcus faecium. Alternatively, Akkermansia mucinifila was associated, and it was suggested that Bacteroides was associated with the inhibitory effect on CTLA-4 (Non-Patent Documents 17 to 19).

By clarifying the pattern of intestinal environmental dynamics for external intervention using omics data, the elucidation of the mechanism of action can be advanced with higher accuracy and speed. In fact, in the case of the HIV therapeutic drug, it was experimentally clarified that the effect of the drug was reduced by focusing on some anaerobic bacteria existing in the bacterial flora in the vagina (Non-Patent Document 20).

In this way, if the effects caused by drugs and functional foods can be predicted from the state of the intestinal environment, precise prescriptions can be made in each case. That is, through the realization of companion diagnostics based on the intestinal environment, medical costs can be reduced and personalized healthcare can be achieved. For example, as described above, if it is possible to predict the effect of improving the intestinal environment or the effect of improving stool condition by ingesting Bifidobacterium (for example, Bifidobacterium longum), which is known to have an intestinal regulating effect, Although the precise prescription and the personalized healthcare can be performed, such a prediction has not been possible in the past.

The present invention has been made in view of the above, and provides a prediction method, a prediction device, and a prediction program capable of predicting the effect of improving the intestinal environment or the effect of improving stool condition by ingesting Bifidobacterium longum. The purpose is.

In order to solve the above-mentioned problems and achieve the object, the prediction method according to the present invention belongs to the genus Serimonas, a bacterium belonging to the genus Cristensenera, a bacterium belonging to the genus Cristensenella, in a stool sample collected from a prediction target. Bacteria, bacteria belonging to the genus Parabacteroides, bacteria belonging to the genus Elyspyrotricus, bacteria belonging to the genus Clostridium XIII AD3011, bacteria belonging to the genus Lacnospira UCG-001, bacteria belonging to the genus Lacnospira UCG-003, bacteria belonging to the genus Faecalitalea , Bacteria belonging to the genus Aristipes, bacteria belonging to the genus Anaerostipes, bacteria belonging to the genus Luminococcus 1 and bacteria belonging to the genus Luminococcus 2 relative abundance ratios of at least one bacterium in the stool sample and the stool. About the prediction target based on at least one of the contents in the stool sample of a predetermined amount of at least one metabolite of N, N-dimethylglycine, butyric acid, asparagine and 3-hydroxybutyric acid in the sample. It is characterized by including a predictive step of making a prediction about the effect of improving the intestinal environment or the effect of improving the stool condition by ingesting Bifidobacterium longum.

Further, in the prediction method according to the present invention, in the prediction step, based on the relative abundance ratio and the content, for the prediction target, the effect of improving the intestinal environment by ingesting the bifidobacteria longum or the stool. It is characterized by making predictions about the effect of improving the condition.

Further, the prediction method according to the present invention is characterized in that, in the prediction step, it is predicted whether or not there is the improvement effect.

Further, the prediction method according to the present invention is characterized in that the improvement of the stool condition is the improvement of bowel movement.

Further, the prediction method according to the present invention is characterized in that the improvement of the bowel movement is an increase in the frequency of bowel movements.

Further, the prediction method according to the present invention is characterized in that the prediction step is executed in the control unit of the information processing device including the control unit.

Further, the prediction device according to the present invention is a prediction device including a control unit, and the control unit is a bacterium belonging to the genus Cristensenera, a bacterium belonging to the genus Celimonas, in a stool sample collected from a prediction target. Bacteria belonging to the genus, bacteria belonging to the genus Parabacteroides, bacteria belonging to the genus Elyspirotrics, bacteria belonging to the genus Clostridium XIII AD3011, bacteria belonging to the genus Lacnospira UCG-001, bacteria belonging to the genus Lacnospira UCG-003, genus Faecalitalea Relative abundance of at least one of the bacteria belonging to the genus Aristipes, the bacterium belonging to the genus Anaerostipes, the bacterium belonging to the genus Luminococcus 1 and the bacterium belonging to the genus Luminococcus 2 in the stool sample And based on at least one of the contents in the stool sample of a predetermined amount of at least one of the metabolites of N, N-dimethylglycine, butyric acid, asparagine and 3-hydroxybutyric acid in the stool sample. It is characterized in that the prediction target is provided with a prediction means for predicting the effect of improving the intestinal environment or the effect of improving the stool condition by ingestion of Bifidobacterium longum.

Further, the prediction program according to the present invention is a prediction program to be executed in an information processing apparatus provided with a control unit, and is a bacterium in a stool sample collected from a prediction target to be executed by the control unit. -Rectal, bacteria belonging to the genus Kristensenera, bacteria belonging to the genus Serimonas, bacteria belonging to the genus Parabacteroides, bacteria belonging to the genus Elyspirotrics, bacteria belonging to the genus Clostridium XIII AD3011, bacteria belonging to the genus Lakunospyra UCG-001, lacnospira At least one of the bacteria belonging to the genus UCG-003, the bacterium belonging to the genus Faecalitalea, the bacterium belonging to the genus Aristipes, the bacterium belonging to the genus Anaerotipes, the bacterium belonging to the genus Luminococcus 1 and the bacterium belonging to the genus Luminococcus Relative abundance ratio of one bacterium in the stool sample and a predetermined amount of at least one metabolite of N, N-dimethylglycine, butyric acid, asparagine and 3-hydroxybutyric acid in the stool sample in the stool sample. Based on at least one of the contents, the prediction target is characterized by including a prediction step for predicting the effect of improving the intestinal environment or the effect of improving the stool condition by ingestion of Bifidobacterium longum. To do.

According to the present invention, it is possible to predict the effect of improving the intestinal environment or the effect of improving the stool condition by ingesting Bifidobacterium longum.

FIG. 1 is a principle configuration diagram showing the basic principle of the present embodiment. FIG. 2 is a diagram showing an example of the configuration of the prediction device 100. FIG. 3 is a diagram showing a cohort design of the first embodiment. FIG. 4 is a diagram showing details of the breakdown of 20 persons selected for analysis in Example 1. FIG. 5 is a diagram showing a test design of Example 1. FIG. 6 is a bar graph showing the number of defecations per day of 20 subjects (MO01 to MO06, MO08 to MO13, MO15 to MO19, and MO22 to MO24) in Example 1. FIG. 7 is a table showing the results of model comparison based on the information criterion (WAIC) in Example 1. FIG. 8 is a table showing the results of calculating the posterior mean value and the Bayesian confidence interval from the posterior distribution of the intake effect using the model No. 5 in which the best WAIC was obtained in Example 1. FIG. 9 is a graph showing the results of confirming the effect of ingesting the test meal in Example 1 from the probability density function of the Weibull distribution using the posterior mean value and the histogram of the actual defecation time. FIG. 10 is a diagram showing a heat map of the relative abundance ratio of intestinal bacteria of each subject, top 50 genus, at each time point of stool sample collection in Example 1. FIG. 11 is a diagram showing the relative abundance ratio of intestinal bacteria for each subject and for each time point in Example 1. FIG. 12 is a diagram showing the relative abundance ratio of intestinal bacteria for each subject and for each time point in Example 1. FIG. 13 is a diagram showing a plot (using the same color for each subject) by multidimensional scaling using the calculated beta diversity for intestinal flora composition in Example 1. FIG. 14 is a diagram showing a plot (using the same color for each time point) by multidimensional scaling using the calculated beta diversity for intestinal flora composition in Example 1. FIG. 15 is a table showing the results of comparing changes in the intestinal flora between the test diet intake group, the control diet intake group, and the normal group in Example 1 using the Wilcoxon-Mann-Whitney Test. .. FIG. 16 shows a boxplot showing the relative abundance ratio of Bifidobacterium longum in the whole subject at each time point in Example 1, and the relative abundance ratio of Bifidobacterium longum for each subject. It is a line graph which shows. FIG. 17 is a diagram showing a heat map of the stool content top 50 genus of metabolites of each subject at each time point of stool sample collection in Example 1. FIG. 18 is a diagram showing a plot (using the same color for each subject) by multidimensional scaling using the calculated beta diversity for metabolite composition in Example 1. FIG. 19 is a diagram showing a plot (using the same color for each time point) by multidimensional scaling using the calculated beta diversity for metabolite composition in Example 1. FIG. 20 is a scatter plot of the responder's Fold Change to the non-responder for enterobacteria and metabolites in Example 1. FIG. 21 is a graph showing the relative abundance ratio of each group for Ruminococcus 2 genus, Erysipelotrichaceae_UCG-003 and Eubacterium rectal in Example 1. FIG. 22 is a graph showing the stool content of each group for 3-hydroxybutyric acid, asparagine and N, N-dimethylglycine in Example 1. FIG. 23 is a flowchart showing the flow of responder prediction by the machine learning method in the second embodiment. FIG. 24 is a ROC curve showing the result when the responder is predicted by the machine learning method in the second embodiment. FIG. 25 is a table showing the results of the responder prediction by the machine learning method in Example 2. FIG. 26 is a graph showing the results of extracting the features that contribute to the responder prediction in Example 2.

The prediction method, the prediction device, and the embodiment of the prediction program will be described in detail below based on the drawings. The present invention is not limited to the present embodiment.

Outline of the Embodiment Here, the outline of the present embodiment will be described with reference to FIG. FIG. 1 is a principle configuration diagram showing the basic principle of the present embodiment.

First, the relative abundance ratio of bacteria in the stool sample collected from the prediction target (for example, an individual such as an animal or a human) in the stool sample and the content of a predetermined amount of metabolites in the stool sample in the stool sample. At least one of (= stool content) is acquired (step SA1: acquisition step in FIG. 1).

In step SA1, when the prediction target is a person, for example, a person having a bad or bad intestinal environment is preferable. The person having a bad or bad intestinal environment is, for example, a person with constipation. The constipated person is, for example, a person who defecates about 3 to 5 times a week. However, the prediction target is not limited to the examples in this paragraph, and may be, for example, a person whose intestinal environment is not bad or a person who is good.

In step SA1, the bacterium is, for example, a bacterium belonging to the genus Cellimonas, a bacterium belonging to the genus Tyzserella 3, a bacterium belonging to the genus Luminococcus 2, a bacterium belonging to the genus Peptoniphilus, a bacterium belonging to the genus uncultarect. Bacteria belonging to, Bacteria belonging to the genus Faecalitalea, Bacteria belonging to the genus Coprobacillus, Bacteria belonging to the genus uncurted_Lachnospiraceae, Bacteria belonging to the genus Eubacterium ([Eubacterium] lectaleGuup. , Bacteria belonging to the genus Crostridium XIII AD3011 (Family_XIII_AD3011_group), Bacteria belonging to the genus Erycipelotrichaceae_UCG-003, Bacteria belonging to the genus Lacnospira UCG-001 (Lachnospiraceae_UCG- Bacteria, Bacteria belonging to the genus Parasuterella, Bacteria belonging to the genus Alitipes, Bacteria belonging to the genus Parabacteroides, Bacteria belonging to the genus Luminococcus 1, Bacteria belonging to the genus Anaerostipes Bacteria belonging to and at least one bacterium in the Christensenella seae-T7_group (Christensenellaceae-T7_group).

In step SA1, the relative abundance ratio is calculated by, for example, amplifying the 16S rRNA gene region of DNA (Deoxyribo Nucleic acid) extracted from the stool sample by PCR (Polymerase Chain Reaction) and performing sequencing by a next-generation sequencer. can do. For example, as a result of the sequencing, when the total amount of DNA is 4, the amount of DNA of the genus Kristensenera is 3, and the amount of DNA of the genus Anaerostipes is 1, the relative abundance ratio of the genus Kristensenera is It is 0.75, and the relative abundance ratio of the genus Anaerostipes is 0.25.

In step SA1, the content can be calculated, for example, by performing a metabolome analysis on the stool sample using CE-TOFMS (capillary electrophoresis-time-of-flight mass spectrometer). For example, as a result of the metabolome analysis, when 3000 nmol of N, N-dimethylglycine is contained in 300 g of the stool sample, the stool content of N, N-dimethylglycine is 10 nmol / g.

The metabolite is, for example, at least one metabolite of N, N-dimethylglycine (N, N-Dimethylglycine), butyric acid (Butyric_acid), asparagine (Asn) and 3-hydroxybutyric acid (3-Hydroxybutyric_acid). ..

Next, based on at least one of the relative abundance ratio and the content obtained in step SA1, the effect of improving the intestinal environment or the effect of improving the stool condition by ingesting Bifidobacterium longum for the predicted target. By making a prediction about (step SA2 in FIG. 1: prediction step), a prediction result can be obtained.

In step SA2, the improvement means that, for example, when the prediction target ingests Bifidobacterium longum, the intestinal environment or stool condition of the prediction target changes to a better one. That is, the improvement may mean that the intestinal environment or stool condition of a person who has poor or poor intestinal environment or stool condition is improved, or the intestinal environment or stool condition is not bad or is not bad. It may be that the intestinal environment or stool condition of a good person is further improved.

In the prediction step in step SA2, the improvement effect may be quantified and predicted, or whether or not the improvement effect is present may be predicted. The latter prediction may be a multi-step prediction according to the degree of the improvement effect, or may be an alternative prediction of whether or not the improvement effect is present.

In step SA2, the intestinal environment is, for example, the relative abundance ratio of bacteria existing in the intestine and the content of the compound. It is known that certain bacteria and certain metabolites have harmful effects on the human body. Therefore, in step SA2, the prediction of the effect of improving the intestinal environment is the prediction of the behavior of a specific bacterium and a specific metabolite in the intestine to be predicted by ingesting Bifidobacterium longum. It may be.

In step SA2, the improvement of the stool condition is, for example, improvement of bowel movement, improvement of the condition of the stool itself, and the like. The improvement of bowel movement may be an increase in the frequency of bowel movements or an increase in the number of bowel movements. As a specific example of the prediction of the increase in stool frequency, when the increase in stool frequency is predicted by the multi-step prediction, the stool frequency increases remarkably when the prediction target ingests Bifidobacterium longum. There is a method of predicting whether the stool frequency will increase but not significantly, or whether the stool frequency will not increase. In addition, as another specific example of the prediction of the increase in stool frequency, when the increase in stool frequency is predicted by the alternative prediction, the stool is predicted by ingesting Bifidobacterium longum. One method is to predict whether the frequency will increase, whether it is significant or not, or whether the frequency of stools will not increase.

Here, in the field of studying the intestinal flora, it is known that the intestinal environment of constipated patients is improved if the stool condition is improved (for example, the frequency of stools is increased). (Cummings JH et al, PASSCLAIM --- gut health and immunity, European Journal of Constipation, 2004, 43, p118-173). Therefore, for example, when an increase in stool frequency is observed for a constipated patient, it may be predicted that the intestinal environment of the constipated patient has improved.

In step SA2, the prediction is preferably made based on the relative abundance ratio and the content. Thereby, for example, more accurate prediction can be performed.

The prediction step of step SA2 may be executed in the control unit of the information processing device including the control unit.

Hereinafter, the threshold value (cutoff value) used in the prediction will be described by taking the case of predicting the increase in the frequency of flights as an example.

The cutoff value may be, for example, the relative abundance ratio or the stool content for discriminating between a group in which the stool frequency increases and a group in which the stool frequency does not increase due to ingestion of Bifidobacterium longum. The cutoff value can be obtained, for example, by performing ROC analysis on the relationship between sensitivity and false positive rate (1-specificity). The sensitivity is, for example, the ratio at which the prediction target, which is the group in which the true state is increased, is correctly predicted to be the group in which the true state is increased. The specificity is a rate at which the prediction target, which is a group whose true state does not increase, is correctly predicted to be a group which does not increase. In the ROC analysis, first, the values of the sensitivity and the false positive rate when the cutoff value is continuously changed are obtained. Then, the values of the obtained sensitivity and false positive rate are plotted on a graph in which the vertical axis (Y axis) is the sensitivity and the horizontal axis (X axis) is the false positive rate, and among the plotted points. Therefore, the combination of the sensitivity and the specificity is determined so that (1-sensitivity) ² + false positive rate ² is minimized. The cutoff value corresponding to the combination of the sensitivity and the specificity determined in this way can be set as the final cutoff value. The method of determining the combination of the sensitivity and the specificity is not limited to the above-mentioned method of determining the combination of (1-sensitivity) ² + false positive rate ² to the minimum, and for example, the sensitivity. It may be a method of determining a combination that maximizes the product of and the specificity, or a method of determining a combination that maximizes (sensitivity + specificity) ÷ 2. May be good.

The cutoff value is, for example, the relative abundance ratio for discriminating between a group in which the stool frequency is significantly increased, a group in which the stool frequency is not significantly increased, and a group in which the stool frequency is not increased due to ingestion of Bifidobacterium longum. It may be stool content. That is, there may be two cutoff values. In this case, as a method of setting the two cutoff values, for example, a group in which the cutoff value obtained by the method described in the previous paragraph is made stricter is compared with the group in which the cutoff value is remarkably increased and the group in which the cutoff value is not remarkably increased. The cut-off value for distinguishing between the groups and the group in which the cut-off value is relaxed, which is obtained by the method described in the previous paragraph, is defined as the group in which the cut-off value is not remarkable but increases and the group in which the cut-off value is not increased. It may be a method of setting the cutoff value to be discriminated.

When predicting the alternatives using only the relative abundance ratio of intestinal bacteria, an increase in stool frequency can be predicted as follows, for example. It is assumed that the cut-off value for the relative abundance ratio that separates the group in which the stool frequency increases and the group in which the stool frequency does not increase is X%. In this case, if the prediction target satisfies X%, the prediction target is predicted to belong to the group in which the flight frequency increases, and conversely, if the prediction target does not satisfy X%, the prediction target is It can be predicted that it belongs to the group in which the stool frequency does not increase.

When the multi-step prediction is made using only the relative abundance ratio of intestinal bacteria, an increase in stool frequency can be predicted as follows, for example. The cut-off value for the relative abundance ratio that separates the group in which the stool frequency increases significantly and the group in which the stool frequency does not increase but increases is X1%, and the stool frequency increases with the group in which the stool frequency does not increase significantly. It is assumed that the cutoff value for the relative abundance ratio that separates the non-group is X2%. In this case, if the prediction target satisfies X1%, the prediction target is predicted to belong to the group in which the stool frequency increases remarkably, and if the prediction target does not satisfy X1% but satisfies X2%. The prediction target is predicted to belong to the group in which the stool frequency is not remarkable but increases, and when the prediction target does not satisfy X2%, the prediction target is predicted to belong to the group in which the stool frequency does not increase. Can be done.

When predicting the alternatives using only the stool content of the metabolite, an increase in stool frequency can be predicted as follows, for example. It is assumed that the cut-off value for the stool content that separates the group in which the stool frequency increases and the group in which the stool frequency does not increase is Ynmol / g. In this case, if the prediction target satisfies Ynmol / g, the prediction target is predicted to belong to the group in which the stool frequency increases, and conversely, if the prediction target does not satisfy Ynmol / g, the prediction is made. The subject can be predicted to belong to the group in which the stool frequency does not increase.

When the multi-step prediction is made using only the stool content of the metabolite, an increase in stool frequency can be predicted as follows, for example. The cut-off value for stool content that separates the group with a marked increase in stool frequency from the group with a less pronounced stool frequency but an increase is Y1 nmol / g. It is assumed that the cut-off value for the stool content that separates the non-group is Y2 nmol / g. In this case, when the prediction target satisfies Y1 nmol / g, the prediction target is predicted to belong to the group in which the stool frequency is significantly increased, and the prediction target does not satisfy Y1 nmol / g but satisfies Y2 nmol / g. In the case, the prediction target is predicted to belong to the group in which the stool frequency is not remarkable but increases, and when the prediction target does not satisfy Y2 nmol / g, the prediction target belongs to the group in which the stool frequency does not increase. Can be predicted.

When predicting the alternative using both the relative abundance ratio of intestinal bacteria and the stool content of the metabolite, an increase in stool frequency can be predicted as follows, for example. When the prediction target satisfies X% and also satisfies Ynmol / g, the prediction target is predicted to belong to the group in which the stool frequency increases, and the prediction target does not satisfy X% and does not satisfy Ynmol / g. In that case, it can be predicted that the prediction target belongs to the group in which the stool frequency does not increase.

When the multi-step prediction is made using both the relative abundance ratio of intestinal bacteria and the stool content of the metabolite, an increase in stool frequency can be predicted as follows, for example. When the prediction target satisfies X1% and also satisfies Y1 nmol / g, it is predicted that the prediction target belongs to the group in which the stool frequency increases remarkably, and the prediction target does not satisfy X1% but satisfies X2%. If Y1 nmol / g is not satisfied but Y2 nmol / g is satisfied, the prediction target is predicted to belong to the group in which the stool frequency is not remarkable but increases, and the prediction target does not satisfy X2% and Y2 nmol / g. If the above conditions are not satisfied, it can be predicted that the prediction target belongs to the group in which the stool frequency does not increase.

[Structure of Embodiment]
Next, an example of the configuration of the prediction device 100 according to the present embodiment will be described with reference to FIG. FIG. 2 is a block diagram showing an example of the configuration of the prediction device 100.

The prediction device 100 is a commercially available desktop personal computer. The prediction device 100 is not limited to a stationary information processing device such as a desktop personal computer, but is portable information such as a commercially available notebook personal computer, a PDA (Personal Digital Assistants), a smartphone, and a tablet personal computer. It may be a processing device.

The prediction device 100 includes a control unit 102, a communication interface unit 104, a storage unit 106, and an input / output interface unit 108. Each part of the prediction device 100 is communicably connected via an arbitrary communication path.

The communication interface unit 104 connects the prediction device 100 to the network 300 so as to be communicable via a communication device such as a router and a wired or wireless communication line such as a dedicated line. The communication interface unit 104 has a function of communicating data with another device via a communication line. Here, the network 300 has a function of connecting the prediction device 100 and the server device 200 so as to be able to communicate with each other, and is, for example, the Internet or a LAN (Local Area Network).

An input device 112 and an output device 114 are connected to the input / output interface unit 108. As the output device 114, a speaker or a printer can be used in addition to a monitor (including a home television). As the input device 112, in addition to a keyboard, a mouse, and a microphone, a monitor that cooperates with the mouse to realize a pointing device function can be used. In the following, the output device 114 may be referred to as a monitor 114, and the input device 112 may be referred to as a keyboard 112 or a mouse 112.

Various databases, tables, files, etc. are stored in the storage unit 106. In the storage unit 106, a computer program for giving a command to a CPU (Central Processing Unit) in cooperation with an OS (Operating System) to perform various processes is recorded. As the storage unit 106, for example, a memory device such as a RAM (Random Access Memory) or a ROM (Read Only Memory), a fixed disk device such as a hard disk, a flexible disk, an optical disk, or the like can be used.

The storage unit 106 includes, for example, relative abundance ratio data 106a and content data 106b. The relative abundance data 106a and the content data 106b may be stored in the server device 200.

The relative abundance data 106a stores the relative abundance acquired in step SA1. The content data 106b stores the content obtained in step SA1.

The control unit 102 is a CPU or the like that comprehensively controls the prediction device 100. The control unit 102 has an internal memory for storing a control program such as an OS, a program that defines various processing procedures, required data, and the like, and performs various information processing based on these stored programs. Execute.

The control unit 102 functionally conceptually, for example, eubacterium rectal, a bacterium belonging to the genus Kristensenera, a bacterium belonging to the genus Serimonas, a bacterium belonging to the genus Parabacteroides, and Ellis in a stool sample collected from a prediction target. Bacteria belonging to the genus Pyrotricus, bacteria belonging to the genus Clostridium XIII AD3011, bacteria belonging to the genus Lacnospira UCG-001, bacteria belonging to the genus Lacnospira UCG-003, bacteria belonging to the genus Faecalitalea, bacteria belonging to the genus Aristipes, Anaerosty The relative abundance ratio of at least one bacterium belonging to the genus Pess, the bacterium belonging to the genus Luminococcus 1 and the bacterium belonging to the genus Luminococcus 2 in the stool sample, and N, N-dimethylglycine in the stool sample, Based on at least one of the contents of at least one metabolite of butyric acid, asparagine and 3-hydroxybutyric acid in the stool sample in a predetermined amount, for the predicted subject, the intestine by ingestion of Bifidobacterium longum It is provided with a prediction unit 102a as a prediction means for predicting the effect of improving the internal environment or the effect of improving the stool condition.

Other Embodiments In addition to the above-described embodiments, the present invention may be implemented in various different embodiments within the scope of the technical idea described in the claims.

For example, of each of the processes described in the embodiments, all or part of the processes described as being automatically performed may be performed manually, or all of the processes described as being performed manually. Alternatively, a part thereof can be automatically performed by a known method.

In addition, the processing procedure, control procedure, specific name, information including parameters such as registration data and search conditions of each processing, screen examples, and database configuration shown in this specification and drawings are not specified unless otherwise specified. Can be changed arbitrarily.

Further, with respect to the prediction device 100, each component shown in the figure is a functional concept and does not necessarily have to be physically configured as shown in the figure.

For example, with respect to the processing functions included in the prediction device 100, particularly each processing function performed by the control unit, all or any part thereof may be realized by the CPU and a program interpreted and executed by the CPU. Also, it may be realized as hardware by wired logic. The program is recorded on a non-temporary computer-readable recording medium including a programmed instruction for causing the information processing apparatus to execute the processing described in the present embodiment, and the prediction apparatus 100 is required. Is read mechanically. That is, in a storage unit such as a ROM or an HDD (Hard Disk Drive), a computer program for giving instructions to the CPU in cooperation with the OS and performing various processes is recorded. This computer program is executed by being loaded into RAM, and constitutes a control unit in cooperation with a CPU.

Further, this computer program may be stored in an application program server connected to the prediction device 100 via an arbitrary network, and all or a part thereof can be downloaded as needed. ..

Further, the program for executing the process described in the present embodiment may be stored in a non-temporary computer-readable recording medium, or may be configured as a program product. Here, the "recording medium" includes a memory card, a USB (Universal Serial Bus) memory, an SD (Secure Digital) card, a flexible disk, a magneto-optical disk, a ROM, an EPROM (Erasable Program Read Only Memory), and an EPROM (registered). Trademarks) (Electrically Erasable and Programmable Read Only Memory), CD-ROM (Compact Disk Read Only Memory), MO (Magnet-Optical Red, Digital, Digital, Digital, Digital, Digital, Digital, Disk) It shall include any "portable physical medium".

A "program" is a data processing method described in any language or description method, regardless of the format such as source code or binary code. The "program" is not necessarily limited to a single program, but is distributed as a plurality of modules or libraries, or cooperates with a separate program represented by the OS to achieve its function. Including things. A well-known configuration and procedure can be used for a specific configuration and reading procedure for reading the recording medium in each device shown in the embodiment, an installation procedure after reading, and the like.

Various databases and the like stored in the storage unit are memory devices such as RAM and ROM, fixed disk devices such as hard disks, flexible disks, and storage means such as optical disks, and are used for various processes and website provision. Stores programs, tables, databases, files for web pages, etc.

Further, the prediction device 100 may be configured as an information processing device such as a known personal computer or workstation, or may be configured as the information processing device to which an arbitrary peripheral device is connected. Further, the prediction device 100 may be realized by mounting software (including a program or data) that realizes the processing described in the present embodiment on the device.

Furthermore, the specific form of distribution / integration of the device is not limited to that shown in the figure, and all or part of the device is functionally or physically in any unit according to various additions or functional loads. It can be distributed and integrated. That is, the above-described embodiments may be arbitrarily combined and implemented, or the embodiments may be selectively implemented.

Hereinafter, examples of the present invention will be described, but the present invention is not limited to the following Examples 1 and 2. In Examples 1 and 2 below, the statistical analysis was performed by the method described in the next paragraph.

All statistical analyzes were performed using Python. Shannon Diversity Index was used for alpha diversity, and Spearman Correlation Cooperative was used for beta diversity. Beta diversity was used for multidimensional scaling (MDS) calculations. The Wilcoxon-Mann-Whitney Test was used for the two-group comparison test, and the Jonckheere-Terpstra test was used for the trend test between the groups. The False Discovery Rate (FDR) and Bonferoni methods were used for the multiple test correction, and the BH method was used for the FDR. In the gut flora analysis, lineage composition data at Genus, Species, and OTU levels were used. Of these, OTU level lineage composition data was used only for the calculation of alpha diversity. Genus level phylogenetic composition data were used for other statistical analyzes. Species level phylogenetic composition data was used only for responder prediction. In the metabolite analysis, data from Reactive area and Content were used. The Reactive area data was used only for the calculation of alpha diversity. Content data was used for other statistical analyzes.

[Example 1] Confirmation of relative abundance ratio and stool content In Example 1, as described in (7) below, lumino was used in subjects who had a large effect of improving the intestinal environment by ingesting Bifidobacterium longum. It was confirmed that the relative abundance ratios of the genus Coccus 2 and the genus Serimonas were large, and conversely, the relative abundance ratios of Eubacterium and Rectal and the stool content of N, N-dimethylglycine and the like were small. Hereinafter, the processing performed in the first embodiment will be described in detail.

(1) Subject and Meal Information (1-1) Subject Information The cohort design of Example 1 will be described with reference to FIG. First, the cohort of Example 1 consisted of 50 Japanese. From the 50 participants who participated in the cohort, 24 (MO01 to MO24) who met the selection criteria and selection exclusion criteria considering the defecation status, age and gender ratio immediately before the study were selected for this study (the 24 concerned). Corresponds to Randomized (N = 24), which is the value obtained by subtracting Excluded (n = 26) from Assessed for elegance (n = 50) in FIG. 3). Participants who were prone to constipation were preferentially selected, especially to investigate the effect of the test diet on constipation. After the completion of this study, 20 subjects (MO01 to MO06, MO08 to MO13, MO15 to MO19, and MO22 to MO24) who did not violate the analysis exclusion criteria were selected for analysis (the 20 subjects are shown in FIG. 3). , Corresponds to the sum of Analyzed (n = 11) and Analyzed (n = 9)). The details of the breakdown of the 20 persons selected for analysis are shown in FIG. The details of the selection criteria, the selection exclusion criteria, and the analysis exclusion criteria are shown in (2-1), (2-2), and (2-3) below, respectively.

(1-2) The test diet information test meal, with Bifidobacterium longum (Bifidobacterium longum) BB536 strain per meal approximately 5.0 × 10 ⁹ cells encapsulated capsules. The weight per serving was 0.53 g. The capsule is spherical with a diameter of about 2.4 mm, and has two coatings: a coating (outer coating) that imparts acid-resistant pH-dependent disintegration ability and a protective layer (inner coating) that improves the barrier function against the passage of gastric acid and the like. It is composed of. B. Encapsulated in the capsule. It has been reported that the 2-hour survival rate of the longum BB536 strain in artificial gastric juice adjusted to pH 1.2 is about 90% (Kohno M, et al., Application of entry-based capsules coating). Bifidobacterium longum to functional foods. Nihon Yakurigaku Zassi Pharmacology Jpn 2016; 148: 310-4.). When the pH of the capsule that has passed through the stomach becomes neutral in the small intestine, the outermost layer, the acid-resistant pH-dependent disintegrating membrane, collapses. Subsequently, the surface-active action of bile acids, the digestive action of lipase, and the physical stimulation of intestinal motility dissolve or disintegrate the hardened fat layer, which is the intermediate layer, and release the bacterial powder inside. It has been reported that the released bacterial powder proliferates and acts actively by being rehydrated by the water in the intestine (Asada M et al., Seamless Capsule Organisms, Seibutu-kougaku kaishi, 2009). , 3, p123-128).

(1-3) Control food information As the control food, the same capsule containing only starch powder was used. The energy, protein, lipid, carbohydrate amount and sodium amount of the test meal and the control meal were adjusted to be equivalent.

(2) Supplementary information: Subject characteristics and exclusion criteria Implementation of Example 1 was requested to CPCC Co., Ltd. after obtaining the approval of the Ethics Committee. After fully explaining the contents of the test, the subjects consented in writing to participate in the test voluntarily. The details of the selection criteria, the selection exclusion criteria, and the analysis exclusion criteria are as follows.

(2-1) Selection Criteria Those who meet all of the following selection criteria 1 and 2 were selected as subjects.
1. 1. Those who are adults between the ages of 40 and 60 at the time of obtaining the consent. Those who defecate 3 to 5 times a week or 7 times or more a week

(2-2) Selection Exclusion Criteria In addition, among those selected in (2-1), those who meet any of the following selection exclusion criteria 1 to 10 were excluded from the subjects.
1. 1. Those who underwent laparotomy within 6 months after the start of the test 2. Those who took antibiotics for more than a week within 6 months after the start of the test. Those who are allergic to test foods 4. Those who plan to change their lifestyle significantly during the test period 5. Persons with a constitution that is prone to chronic diarrhea 6. Persons with a history of significant liver dysfunction, gastric dysfunction, cardiovascular disease, etc. 7. Persons suspected of having a chronic or acute infection 8. 9. Those who are pregnant, breastfeeding, or may be pregnant. Those who participated in other clinical trials within the past month 10. Those who are deemed inappropriate by other investigators

Twenty-four subjects (MO01 to MO24) selected in this way were observed for defecation dynamics for two weeks immediately before the start of the test, and confirmed that the number of defecations was 3 to 5 times per week or 7 times or more per week. Was done.

(2-3) Analysis Exclusion Criteria Among the 24 subjects (MO01 to MO24) who completed this study, those who met any of the following analysis exclusion criteria 1 to 6 were excluded from the analysis target.
1. 1. Those who are judged to be in conflict with the above selection exclusion criteria 2. Those with a test food inoculation rate of less than 80% 3. Those who have large fluctuations in the contents of meal records, or those who cannot confirm that there are no fluctuations from the records. Those who have large fluctuations in the contents of their life diary, or those who cannot confirm that there is no fluctuation from the records. 6. Those who have continuously or repeatedly ingested medicines, foods for specified health use, foods with functional claims, supplements or diet foods, etc. that may affect this study. Those who were judged by the investigator to be unsuitable for analysis for other reasons, and 20 excluded subjects (MO01 to MO06, MO08 to MO13, MO15 to MO19, and MO22 to MO24) were selected for analysis. It was.

(3) Study design and sample collection (3-1) Test design This study was conducted according to the schedule shown in FIG. 5 based on a randomized, double-blind crossover study. Specifically, first, 24 subjects who passed the selection criteria and the selection exclusion criteria were randomly divided into two groups (12 each). Then, the subjects in one group first orally ingested the test food (Test food in FIG. 5), sandwiched a washout period (Washout in FIG. 5), and subsequently the control food (in FIG. 5). , Placebo food) was orally ingested. Subjects in the other group ingested in the reverse order. As shown in FIG. 5, the intake period of the subject was 2 weeks for both the test meal and the control meal, and the washing period was 4 weeks.

Each subject ingested the test meal or the control meal once a day with water at any time. By the time they were ingested, the test meal or control meal was stored at room temperature by each subject. The subjects' activities during the study period were described in a daily questionnaire for each individual. The questionnaire included living conditions, diet and defecation status. The subjects were notified of the following 1 to 6.
1. 1. Prohibit the intake of beverages or foods high in lactic acid bacteria and bifidobacteria. 2. Prohibit the intake of beverages or foods high in dietary fiber and oligosaccharides. Prohibit the intake of supplements 4. Prohibit the intake of functional yogurt 5. Avoid ingestion of natto, and if ingested, describe in the above questionnaire. Avoid ingestion of health foods, and if ingested, describe in the above questionnaire

(3-2) Sample collection From each subject, 1 to 6 days before ingestion of the test meal or the control meal, 7 to 13 days after the start of ingestion, and 1 to 7 days after the end of ingestion. During the day, stool samples were collected. That is, in the case of the group in the order of Placebood → Testhood (upper group) in FIG. 5, as shown by P → T in FIG. 5, the six points P1, P2, P3, T1, T2 and T3 are arranged in order from the left. A stool sample was collected. On the other hand, in the case of the group in the order of Testhood → Placebood (lower group) in FIG. 5, as shown by T → P in FIG. 5, 6 points of T1, T2, T3, P1, P2 and P3 are in order from the left. A stool sample was collected at. The collected stool sample was collected at home by an individual subject using a stool collection sheet "Nagasale" (manufactured by Ozax) in the toilet bowl and a stool collection tube "Faces container 54 x 28 mm" (manufactured by Zalstat). Was done. Immediately after collection, it was stored in a domestic freezer, and the collected stool was collected by freezing transportation.

(4) Determination of defecation responder Based on the questionnaire information recorded for each subject over time, the subjects (responders) whose defecation activity was confirmed to be improved by the test meal were estimated. The effect of the test meal was judged mainly based on the shortening of the defecation time interval, but other factors (number of defecations, defecation probability, etc.) were also evaluated using a statistical model. The details of the judgment method based on the shortening of the defecation time interval will be explained in (4-1) below, and the details of the judgment method based on the number of defecations and the defecation probability will be explained in (4-2) below. The determination of the defecation responder will be described in (4-3) below.

(4-1) Details of judgment method based on shortening of defecation time interval A statistical model was constructed to confirm the shortening of defecation time. In this model, we focused on the fact that defecation activity tends to occur over time, and modeled the defecation time interval for each individual with a Weibull distribution. The Weibull distribution is a widely used distribution for investigating the effect of drugs on survival time and has shape and scale parameters. The effect of eating and drinking was estimated based on a proportional hazard model (Mudholkar GS et al., A Generalization of the Weibull Distribution with Application to the American StatisAvi. -1583). In this model, it is assumed that the effect of the covariate at time t occurs as a product to the hazard function in the reference state. Formula 1 is shown below.

Here, the hazard function is a function that expresses the probability that an event will occur after a minute time Δt has elapsed from a certain time t as a starting point. That is, in this test, it was assumed that the effect of covariates associated with defecation time, such as the effect of ingestion of the test meal, occurs as a product of the hazard function in the normal state, and the probability of occurrence of defecation activity changes. This assumption is equivalent to the fact that the hazard function does not change except for eating during the observation period. This is supported by the rule of thumb that the intestinal condition is stable in conditions where food intake is strictly restricted, such as in this cohort. In reality, the parameters and the effect of eating on the Weibull distribution in the model are unknown. Such numbers were estimated from the observed data. The defecation record obtained this time was not time data but data on the number of defecations per day. Therefore, the estimated time interval of each defecation activity was calculated by dividing 24 hours by the number of defecations. By incorporating this inaccuracy into the model, more accurate estimation is possible, but this time it is omitted for the sake of simplicity. It should be noted that one subject always had defecation once a day for the entire observation period of 85 days, so the value was always 24 hours / time. This value was inappropriate for model estimation and was clearly not a defecation responder, so it was excluded from the analysis. Furthermore, from the observed data, it was not clear whether there were individual differences in the parameters in the model or whether they were common to the whole, so the models were compared using the Widely applicable information criteria (WAIC) (Watanabe S., Cross-validation). Bayes Cross Validation and Widay Applicable Information Creation in Singular Learning Theory, Journal of Machine Learning Research, 2010, 11, p3571. WAIC was calculated using 2n times the generalization loss so that it can be compared with leading indicators such as AIC (Gelman A et al., Understanding predative information criterion criteria for Bayesian models, Station2 , 6, p997-1016). WAIC and AIC approximate the prediction error (generalization error) for unknown data of the model, and WAIC in particular can be used for singular models in which the posterior distribution of parameters cannot be approximated by a normal distribution. Is. The estimation of the parameters for the data was achieved by the Markov chain Monte Carlo method (MCMC method) by the NUTS algorithm using Python and Stan (Carpenter B et al., Stan: A Public Programming Language, Journal of Statistical Software17, Statistical Software 17 1, https: //doi.org/10.18637/jss.v076.i01). There is a report from a previous study on a generalized linear model of the Weibull distribution by the MCMC method (Dellaportas P et al., Bayesian Information for Generalized Liner and Proportional Hazards Models, Gibbs Via Gibbs3, Gibbs sampling, ). Miniconda was used to implement Python, and the version was 4.3.1 (Python version was 3.6.1). The version of PyStan, which is the interface of Stan, used 2.1.6. The number of repetition steps of MCMC was 3000, the number of chains was 8, and the first 1000 repetitions were excluded from the calculation as warmup. Convergence of MCMC was confirmed by the fact that the potential statistic reduction factor (PSRF or Rhat) was 1.1 or less in all the estimated values according to the previous study (Blocks SP et al., General Methods for). Monitoring Convergence of Iterative Simulations, Journal of Computational and Graphical Statistics: A Joint Publication of American Statistical Association, Institute of Mathematical Statistics, Interface Foundation of North America, 1998,7,4, p434-455). Rhat is calculated based on the variance of a plurality of MCMC chains and is used for the estimation convergence test. Initial values were used for other parameters.

(4-2) Details of the judgment method based on the number of defecations and the probability of defecation The dynamics of defecation were analyzed using the following model for the criteria of the number of defecations and the probability of defecation.

For the number of defecations, the rule-based method and the probability model-based method were used. The rule-based method assumed that responders and non-responders could be estimated based on the average number of bowel movements. The definition of responder was to meet both of the following two criteria: The first criterion is the average number of defecations during the test meal or control meal intake period: 1) 7 days immediately before the test meal intake period, 2) the last week of the pre-intake pre-observation period, and 3) pre-observation. Compared with the average number of defecations during the period and the rest period, the average number of defecations during the test meal intake period was higher than any of the periods. The second criterion was that the increase in the average number of defecations of the test meal was larger than that of the control meal in the comparison in any of the periods 1 to 3. In the probabilistic model-based method, it was assumed that the number of defecations per day for each individual could be modeled by the Poisson distribution. The Poisson distribution is a distribution used when modeling data for the number of times, and has only one ratio parameter. The effect of eating was estimated assuming that the ratio parameter changed with intake. For the above two methods, the number of questionnaire-based defecations obtained from each subject was used for estimation.

For defecation probability, only the probability model-based method was used. In this method, it was assumed that the occurrence of daily defecation for each individual could be modeled by the Bernoulli distribution. The Bernoulli distribution is a distribution used to express the probability of occurrence of alternative options such as the front and back of a coin, and has only one probability parameter. The effect of eating was estimated by assuming that the stochastic parameter changes with intake through the logistic function. For this method, the data on the number of defecations was converted into data on the presence or absence of defecation on a daily basis, such as 1 if there was defecation and 0 if there was no defecation.

(4-3) Defecation responder determination Then, the fluctuation tendency of the number of defecations corresponding to the intake situation was investigated. The result is shown in FIG. FIG. 6 is a bar graph showing the number of defecations per day of 20 subjects (MO01 to MO06, MO08 to MO13, MO15 to MO19, and MO22 to MO24). In FIG. 6, the vertical axis shows the number of defecations per day, and the horizontal axis shows 20 subjects (MO01 to MO06, MO08 to MO13, MO15 to MO19 and MO22 to MO24) and the overall average (all). .. Further, in FIG. 6, the purple bar graph indicates the test food intake period (Test), the green bar graph indicates the control food intake period (Placebo), and the red bar graph indicates the test food intake period and the control food. The normal period (None), which is a period other than the intake period, is shown, and the blue bar graph shows the normal period (None) and the control diet intake period (Placebo). Further, in FIG. 6, the test meal non-ingestion period (None; normal period, Placebo; control food intake period, None ∪ Placebo; normal period or control food intake period) vs. the test meal intake period (Test), Wilcoxon- A Mann-Whitney Test (one-sided test) was performed.

As shown in FIG. 6, as a whole subject, the number of defecations was significantly increased in Test with respect to None and None ∪ Placebo (pvalue None: 0.0153, None & Placebo: 0.0259). In addition, as shown in FIG. 6, for each individual, the number of defecations in MO04, MO10, and MO11 increased significantly in Test compared to None. As shown in FIG. 6, for MO04 and MO10, the number of defecations was significantly increased in Test for None & Placebo, but for MO11, the number of defecations was also significantly increased in Placebo for None. Was increasing.

That is, in a word, the result shown in FIG. 6 was that the number of defecations was significantly increased when the test meal was ingested as a whole, but there was a difference in the intensity among individuals. We hypothesized that the cause of this is that there are individual differences in the effect of taking the test meal, and that there are subjects (responders) who have a particularly strong effect. In order to quantify the effect and test the hypothesis, we focused on the improvement of constipation, which is the main effect of the test meal, and estimated the intake effect for each individual from the measured defecation frequency. The mathematical model was formulated by interweaving the placebo effect and the effect of the test meal itself in order to estimate the intensity of the precise effect. By confirming these values, it is possible to confirm the effect statistically. Subsequently, the parameters were estimated by applying Bayesian statistically the defecation time interval calculated based on the number of defecations to the constructed model.

First, the validity of the existence of individual differences was evaluated by comparing models based on the information criterion. The results are shown in the table of FIG. Note that the information criterion using Widely Acceptable Information Criteria (WAIC) (Watanabe S., Asymptotic Equivalence of Bayes Cross Validation and Widely Applicable Information Criterion in Singular Learning Theory, Journal of Machine Learning Research, 2010,11, p3571- 3594). WAIC is an index showing the degree of good predictive power (generalization error) for unknown data, and it is considered that the lower the value, the higher the predictive power.

As shown by hatching in the table of FIG. 7, the smallest WAIC was shown in the No. 5 model assuming that the effect of eating and the effect of the test meal differed from person to person. This supported the hypothesis that the effects produced by eating and the effects produced by the test diet differed from person to person.

Subsequently, the posterior mean value and Bayesian confidence interval were calculated from the posterior distribution of the intake effect using model No. 5 in which the best WAIC was obtained. The results are shown in the table of FIG. FIG. 8 shows the ex post facto average and the 95% confidence interval. Further, in FIG. 8, the parameter of the Weibull distribution is on both sides, and the intake effect is on one side.

As shown by linear hatching in the table of FIG. 8, in the subjects of MO04, MO05, and MO10, 0, which means no effect within the 95% Bayesian confidence interval, is found in the predicted distribution of the effect of the test meal. It was not contained, suggesting that it acts particularly strongly. For this reason, the three subjects, MO04, MO05, and MO10, had a particularly strong effect of the test meal (the defecation time was strongly shortened by ingesting the test meal), and the defecation responder (SR). Was defined as. On the other hand, as shown by the dot-shaped hatching in the table of FIG. 8, even when the ex post facto average value estimated in other subjects was confirmed, 9 other subjects (MO01, MO02, etc.) had a value exceeding 0. Since MO08, M009, MO11, MO13, MO17, MO22, and MO24) were present, the effect of the test meal was exhibited (the defecation time was shortened by ingesting the test meal), and the defecation responder (WR) was present. ). Then, seven subjects other than these (MO06, MO12, MO15, M016, MO18, MO19 and MO23) were defined as non-responders (NR).

Finally, the effect of ingesting the test meal was confirmed from the probability density function of the Weibull distribution using the posterior mean value and the histogram of the actual defecation time. The result is shown in FIG. In FIG. 9, the left vertical axis shows the value of the probability density function, and the right vertical axis shows the number of histograms. Further, in FIG. 9, the blue bar graph corresponds to the normal time, the green bar graph corresponds to the intake of the control meal, and the orange bar graph corresponds to the intake of the test meal. Then, in FIG. 9, the posterior distribution was confirmed, and the subject defined as Strong Responder; SR was in orange letters, the subject defined as Week Responder; WR was in yellow letters, and the subject defined as Non-Responder; NR was in black. Indicated by the letters. Since MO03 always had a defecation time of 24 hours and was unsuitable for analysis, and MO07, MO014, MO020 and MO021 did not meet the criteria for analysis, they were excluded from analysis.

(5) Trends in the composition of the relative abundance ratio of intestinal bacteria Next, regarding the analysis subjects MO01 to MO06, MO08 to MO13, MO15 to MO19, and MO22 to MO24, from the stool samples collected in (3-2). DNA was extracted and the 16sRNA gene region of the extracted DNA was PCR amplified. From this, it was found that it is the individual difference of each subject that influences the tendency in the composition of the abundance ratio of intestinal bacteria.

Specifically, DNA extraction from stool is described in a paper (Murakami S et al., The Consumption of Bicarbonate-Rich Mineral Water Industries Glycemic Control, Alternative Medicine83 It was done using the technique. For the extracted DNA, 27Fmod and 338R (Kim SW et al., Robustness of Gut Microbiota of Health Adults pyrosequencing Probiotic Technology), which are universal primers for the V1-V2 region of the bacterial 16S rRNA gene, are received. Research: An International Journal for Rapid Publication of Reports on Genes and Genomes, 2013, 20, 3, p241-253) was used for amplification. Illumina MiSeq was used for sequencing the amplicon DNA, and the sequence was performed under the conditions of paired-end mode and 600 cycles. The obtained 16S rRNA gene sequence can be used in DRA of DDBJ (DRA accessory number: DRA006874). The obtained DNA sequence is vsearch version 1.9.3 (Option: --fastq_maxee 9.0 --- fastq_truncqual 7 --- fastq_maxdiffs 300 --- fastq_maxmergelen 330 --fastelgenergen (Fastq_maxmergelen. The F and R side leads were merged using open source tool for metagenomics, PeerJ, 2016, e2584). Subsequently, fragments with an average quality of <25 were removed. All fragments are Bowtie2 version 2.2.9 (Option: --no-hd-no-sq-no-unal-I 280-X 400 --fr --- no-discordant --- phred33-D15-R 10 -N 0-L 22-i S, 1,1.15-q) (Langmead B et al., Fast gapped-read alignment with Bowtie 2, Nature Methods, 2012, 9, 4, p357-359) , SILVA SSU NR database version128 (Quast C et al., The SILVA ribosomal RNA gene data process project: improbated data processing andb-base, edited and web-base10, Nucleic acid Only fragments mapped within 3% were adopted. From the remaining fragments (29138 ± 4257), 20,000 fragments were subsampled and used for analysis.

Then, the obtained PCR product was sequenced by a next-generation sequencer. From this sequence data, the relative abundance, alpha diversity and beta diversity at the genus and species level of the gut flora were calculated. The results are shown in FIGS. 10 to 13. FIG. 10 shows a heat map of the relative abundance ratio of intestinal bacteria of each subject, top50 genus, at each time point (P1, P2, P3, T1, T2 and T3) of stool sample collection. 11 and 12 show the relative abundance of gut flora for each subject and for each time point. In FIGS. 11 and 12, the length of the horizontal bar indicates the relative abundance ratio (ratio) of the intestinal bacteria, and the same color is used for the same type (genus) of the intestinal bacteria. 13 and 14 show plots by Multi Dimensional Scaling (MDS) using calculated beta diversity (Spearman Coloration Cofficient) for intestinal flora composition, and FIG. 13 shows subjects. The same color was used for each time point, and in FIG. 14, the same color was used for each time point.

As shown in FIGS. 10, 11, 12 and 14, no overall tendency between groups was observed in the subject's intake or said time points. In contrast, as shown in FIG. 13, dots of the same color (indicating the same subject) were grouped together and plotted, observing similar bacterial communities of the same individual. From the above, it was suggested that it is the individual difference of each subject that influences the tendency of the composition of the abundance ratio of the intestinal bacteria, rather than the intake of the test meal.

Next, to compare in detail the changes in the gut flora between the test diet intake group (T2 and T3), the control diet intake group (P2 and P3) and the normal group (T1 and P1), Wilcoxon- Mann-Whitney Test was used. The results are shown in the table of FIG.

As shown in the table of FIG. 15, some bacterial genera were changed in the test diet intake group as compared with the other groups (p-value <0.05 not collected), but False Discovery Rate (FDR) correction. It was judged that there was no significant difference. No significant variation was also detected for alpha diversity (pvalue None vs Test: 0.246, Placebo vs Test: 0.258).

In addition, the relative abundance ratio of Bifidobacterium longum in the entire subject at each time point is shown as a boxplot in FIG. Then, the relative abundance ratio of Bifidobacterium longum for each subject is shown as a line graph of FIG.

As shown in FIG. 16, both the boxplot and the line graph are between the values at the time points T1 to T3 at the time of ingesting the test meal and the values at the time points P1 to 3 at the time of ingesting the control meal. No significant difference was observed. That is, no significant change in the relative abundance ratio of Bifidobacterium longum was observed by ingestion of the test meal.

(6) Trends in the composition of stool content of metabolites Next, regarding the subjects to be analyzed, MO01 to MO06, MO08 to MO13, MO15 to MO19 and MO22 to MO24, metabolites from the stool samples collected in (3-2). Was extracted, and the extracted metabolite was metabolome-analyzed by CE-TOSMS. From this, it was found that it is the individual difference of each subject that influences the tendency of the composition of the stool content of the metabolite.

Specifically, in order to extract metabolites from the stool sample, the stool sample was first freeze-dried using a freeze-dryer VD-800R (manufactured by TIETECH Co., Ltd.) for at least 24 hours. The freeze-dried stool sample was crushed with 3.0 mm zirconia beads using a multi-sample cell crusher Shakemaster Neo Ver1.0 (manufactured by Biomedical Science) at 1,500 rpm for 10 minutes. .. 500 μl of methanol containing internal standards (20 μM each of methyl sulfone and D-camphor-10-sulfonic acid (CSA)) was added to 10 mg of the crushed stool sample. Further, the stool sample was crushed with 0.1 mm zirconia / silica beads using the Shakemaster Neo at 1,500 rpm for 5 minutes. Subsequently, 200 μl of ultrapure water and 500 μl of chloroform were added, and the additive was centrifuged at 4,600 g for 15 minutes at 20 ° C. Further, in order to remove protein and lipid molecules, a 150 μl aqueous layer was transferred to a centrifugal filtration filter unit Ultrafree MC-PLHCC 250 / pk for Metabolome Analysis (manufactured by Human Metabolome Technologies). Then, the filtrate was centrifugally concentrated and dissolved in 50 μl of ultrapure water immediately before the CE-TOFMS analysis.

CE-TOFMS analysis was performed to calculate the alpha and beta diversity of each metabolite. The results are shown in FIGS. 17 to 19. FIG. 17 shows a heat map of the stool content top50 genus of metabolites of each subject at each time point (P1, P2, P3, T1, T2 and T3) of stool sample collection. 18 and 19 show plots by multidimensional scaling using calculated beta diversity for metabolite composition, FIG. 18 uses the same color for each subject, and FIG. 19 shows the stool. The same color was used for each sample collection time point.

As shown in FIGS. 17 and 19, no overall trend between groups was observed in the subject's intake or said time points. In contrast, as shown in FIG. 18, dots of the same color (indicating the same subject) were grouped close together and plotted, observing similar metabolite compositions of the same individual. From the above, it was suggested that it is the individual differences of each subject that influence the tendency of the composition of the stool content of the metabolites, rather than the intake of the test meal.

Next, to compare in detail the changes in metabolites between the test diet intake group (T2 and T3), the control diet intake group (P2 and P3), and the normal group (T1 and P1), Wilcoxon-Mann- Whiteney Test was used. As a result, there was no metabolite in the test diet intake group that was significantly different from the other groups (pvalue <0.05 not collected). No significant variation was also detected for alpha diversity (pvalue None vs Test: 0.973, Placebo vs Test: 0.456).

(7) Differences between responders and non-responders Comparing defecation responders (SR and WR) and defecation non-responders (NR), the relative abundance ratio changed significantly and the gut flora and stool content changed significantly. In order to measure the metabolites, the Fold Change (FC _S ) for SR NR and the Fold Change (FC _W ) for WR NR were calculated for each gut flora or metabolite using the following formula 2. ..

In Equation _{2, Average abundance f, strong responder} is SR MO04, MO05, and MO 10 3 people each time point of the relative abundance of Enterobacteriaceae in (P1, P2, P3, T1 , T2 and T3) Alternatively, it is a numerical value obtained by averaging the stool content of the biotransformer, and the Average abundance _{f, weak responder} is the WR at each of the nine time points of MO01, MO02, MO08, M009, MO11, MO13, MO17, MO22 and MO24. It is a numerical value obtained by averaging the relative abundance ratio of intestinal bacteria or the stool content of biotransforms, and the Average abundance _{f, non responder r} is the above-mentioned 7 persons having NRs MO06, MO12, MO15, M016, MO18, MO19 and MO23. It is a value obtained by averaging the relative abundance ratio of intestinal bacteria or the stool content of biotransformers at each time point.

In order to search for a feature amount in which the relative abundance ratio of intestinal bacteria or the stool content of biotransforms is larger as the effect of the test diet is stronger, intestinal bacteria and biotransforms in which FC _S > FC _W > 0 are searched. In addition, in order to search for a feature amount in which the relative abundance ratio of intestinal bacteria or the stool content of biotransforms is smaller as the effect of the test diet is stronger, intestinal bacteria and biotransforms in which FC _S <FC _W <0 are searched. did. The result is shown in FIG.

FIG. 20 is a scatter plot of the responder's Fold Change against the non-responder in gut flora and metabolites. In FIG. 20, the vertical axis shows the value obtained by taking the log 2 of SR's Folder (FC _S ) with respect to NR, and the horizontal axis shows the value of WR's Folder (FC _W ) taken with respect to NR. In FIG. 20, the warm-colored plot corresponds to the intestinal flora, the size of which represents the relative abundance. In FIG. 20, the genus Bacteria shown in red plots (eg, the genus Sellimonas, the genus Tyzzerella 3, the genus Luminococcus 2, the genus Peptoniphilus, the genus Uncultured_Acticateca , coprocessor Bacillus (Coprobacillus) genus, uncultured_Lachnospiraceae spp., Eubacterium, Rekutaru ([Eubacterium] rectale group), Family_XIII_UCG-001 genera, Clostridium XIII AD3011 (Family_XIII_AD3011_group) genus, Ellis Pirot Rikusu (Erysipelotrichaceae_UCG-003) belonging to the genus, Rakunosupira UCG- 001 (Lachnospiraceae_UCG-001), Lacnospira UCG-003 (Lachnospiraceae_UCG-003, etc.) and Parasuterella (genus Parasuterella, etc.) are NR, WR, and SR. Was a bacterial genus. In FIG. 20, cool-colored plots correspond to metabolites, the size of which represents the concentration. In FIG. 20, the metabolites shown in the blue plot (eg, 3-hydroxybutyric_acid, asparagine (Asn) and N, N-dimethylglycine (N, N-Dimethylglycine), etc.) are NR, WR. And the metabolites with pvalue <0.05 in the intergroup comparison test (Jonckheere-Tarpstra) between SRs. In FIG. 20, the light orange part of the background indicates FC _W <FC _S , that is, NR <WR <SR, and the light blue part of the background is FC _W > FC _S , that is, NR> WR. > Indicates SR.

As shown in FIG. 20, the genus Sellimonas, the genus Tyzzerella, the genus Ruminococcus 2 and the genus Peptoniphilus belonged to the light orange part of the background, so the relative abundance ratio was NR <. It was found that WR <SR, that is, the relative abundance ratio was large in the responder. Similarly, as shown in FIG. 20, 3-hydroxybutyric acid (3-Hydroxybutyric_acid) also belonged to the light orange part of the background, so that the relative abundance ratio was NR <WR <SR, that is, relative in the responder. It was found that the abundance ratio was large. On the other hand, as shown in FIG. 20, the genus incultured_Actinomycetheae, the genus Faecalitalea, the genus Coprobacillus, the genus uncurtured_Lachnospiraceae, the genus Uncultured_Lachnospiraceae, the genus Eubacteriumriacure (Eubacterium. , Crostridium XIII AD3011 (Family_XIII_AD3011_group) genus, Erycipelotrichaceae_UCG-003 genus, Lacnospira UCG-001 (Lachnospiraceae_UCG-001) genus Lachnospiraceae_UCG-001 Since it belonged to the light blue part, it was found that the relative abundance ratio was NR> WR> SR, that is, the relative abundance ratio was small in the responder. Similarly, as shown in FIG. 20, asparagine (Asn) and N, N-dimethylglycine (N, N-Dimethylglycine) belonged to the light blue part of the background, so that the relative abundance ratio was NR> WR> SR. That is, it was found that the relative abundance ratio was small in the responder.

Subsequently, for each of the genus Ruminococcus 2, the genus Elyspyrotricus UCG-003, and Eubacterium rectal, the average value (Reactive Abunance) of the relative abundance ratios at each of the time points was calculated for each SR, WR, and NR. did. The result is shown in FIG. In FIG. 21, the horizontal axis represents Reactive abundance, the white graph corresponds to NR, the orange graph corresponds to WR, and the red graph corresponds to SR. In addition, for 3-hydroxybutyric acid, asparagine and N, N-dimethylglycine, the average value (Conent Avenance [nmol / g]) of the stool content at each of the time points was calculated for each of SR, WR and NR. The result is shown in FIG. In FIG. 22, the horizontal axis represents Content Avenance, the white graph corresponds to NR, the orange graph corresponds to WR, and the red graph corresponds to SR. In FIGS. 21 and 22, * represents p-value <0.05 by the Willcoxon-Mann-Whitney test, and ** represents Bonferroni corrected qvalue <0.05.

As shown in FIG. 21, it was found that the relative abundance ratio of the two genus Ruminococcus was NR <WR <SR, that is, the relative abundance ratio was large in the responder. Similarly, as shown in FIG. 22, it was found that the stool content of 3-hydroxybutyric acid was NR <WR <SR, that is, the stool content was high in the responder. On the other hand, as shown in FIG. 21, for the genus Elyspyrotricus UCG-003 and Eubacterium rectal, the relative abundance ratio is NR> WR> SR, that is, in the responder. It was found that the relative abundance ratio was small. Similarly, as shown in FIG. 22, it was found that the stool content of asparagine and N, N-dimethylglycine was NR> WR> SR, that is, the stool content was low in the responder.

[Example 2] Extraction of feature amount by machine learning In Example 2, from the intestinal multiomics data immediately before ingestion of the test meal, lasso regression and logistic regression are performed to determine whether the defecation responder increases the number of defecations by ingesting the test meal. Predicted by a combined machine learning method. In addition, by extracting the feature amounts that contribute to the prediction, the relative abundance ratio of the genus Ruminococcus 1, the genus Anaerotipes, the genus Ruminococcus 2 and the Kristensenella R-7 group and the stool content of butyric acid are large. It was confirmed that it contributed to the prediction of the responder, and conversely, the smaller the relative abundance ratio of the genus Eubacterium, Ruminococcus, and the genus Ruminococcus and the genus Parasatella, the more it contributed to the prediction of the responder. confirmed. Hereinafter, the processing performed in the second embodiment will be described in detail.

(1) Prediction of responders by machine learning method Specifically, 19 subjects (MO01 to MO02, MO04 to MO06, MO08 to MO13, MO15 to MO19, and MO22 to MO24) excluding MO03 immediately before ingesting a test meal (T1). ), The relative abundance ratio of the bacterial species or the genus of bacteria and the stool content of metabolites were used, and only those having the relative abundance ratio and the stool content of each of these above a certain threshold were used as feature amounts. A grid search was performed on the threshold value to search for the one with the maximum prediction accuracy (step SB1 in FIG. 23). Next, the range of possible values may differ between the relative abundance ratio of gut flora and the stool content of metabolites, the feature amount with a large range of possible values may have a large effect on prediction, and the intestine. Considering the problem that the magnitudes of contributions of the gut flora and metabolites cannot be directly compared, each feature was standardized using z score (step SB2 in FIG. 23).

Subsequently, 19 subjects were divided into two groups of responders and non-responders (that is, SR vs WR, NR or SR, WR vs NR) (step SB3 in FIG. 23), and one person from each group. It was divided into test data including each and training data of the remaining 18 persons (steps SB4 and SB5 in FIG. 23). Subsequently, using the training data, Lasso regression was performed on the average value of the effect of the test meal for each individual estimated by the statistical model, and only the features that greatly contributed to the effect of the test meal were extracted (). Step SB6 in FIG. 23). The parameters of the Lasso regression were grid-searched to find the one with the maximum prediction accuracy. Subsequently, the training data was learned by the logistic regression algorithm using the extracted features (step SB7 in FIG. 23), and the test data was predicted (step SB8 in FIG. 23). Initial values were used for the parameters of logistic regression. This was executed for all combinations of all responders and all non-responders (60-78 ways), and was used as Cross Validation.

The prediction results of the test data are shown in FIGS. 24 and 25. Among the extracted feature quantities of large contributions, bacterial species (Genus), bacterial species (Species), metabolites (Metabolite), bacterial genera (Genus) & metabolites (Metabolite), and bacterial species (Species) & metabolism. FIG. 24 is a receiver operating characteristic (ROC) curve when the prediction of “SR and WR” is performed for each of the five parameters of the substance (Metabolite), and the table is FIG. 25. In addition, in FIGS. 24 and 25, AUROC (AUC), Accuracy and F-mere are evaluation indexes of prediction accuracy.

Although not shown, SR was predictable with the highest accuracy (AUC = 0.875) when both bacterial species (Species) and metabolites (Metabolite) were used in combination. Further, as shown in FIGS. 24 and 25, the SR and WR (that is, the entire defecation responder) have the highest accuracy (AUC) when both the bacterial genus (Genus) and the metabolite (Metabolite) are used in combination. = 0.857) was predictable.

(2) Extraction of features that contribute to the prediction of responders Next, in Cross Validation, the relative abundance ratio of the genus Bacteria immediately before ingestion of the test meal, which is the condition with the highest prediction accuracy, is 0.01 or more, and metabolites. Lasso regression (alpha = 0.1) was performed on the responder score using the standardized features of z score with a stool content of 1000 or more. As a result, the regression coefficients of each bacterial genus and metabolite were calculated, and the features contributing to the responder judgment (prediction) were extracted. The result is shown in FIG.

FIG. 26 shows the regression coefficient of Lasso regression for the responder score in the parameter with the highest accuracy in Cross validation using the data of the bacterial genus and metabolites immediately before ingestion of the test meal. The size of the yellow circle adjacent to the bacterial genus name indicates the average value of the relative abundance. The size of light blue adjacent to the name of the metabolite indicates the average value of stool content.

As shown in FIG. 26, the higher the stool content of butyric acid, the higher the relative abundance ratio of Ruminococcus 1 genus, Anaerotipes genus, Kristensenella R-7 group and Ruminococcus 2 genus. The larger it was, the more it contributed to the responder's prediction. On the other hand, for Eubacterium rectal, Aristipes, uncurtured_Lachnospiraceae and Parabacteroides, the smaller the relative abundance ratio, the more contributed to the prediction of the responder.

As described above, the present invention can be widely implemented in many industrial fields (food, pharmaceuticals, medical treatment, etc.), and is extremely in the bioinformatics field, which predicts the effect of improving the intestinal environment. It is useful.

100 Prediction device 102 Control unit 102a Prediction unit 104 Communication interface unit 106 Storage unit 106a Relative presence ratio data 106b Content data 108 Input / output interface unit 112 Input device 114 Output device 200 Server 300 Network

Claims (9)

1. In stool samples collected from the prediction target, eubacterium rectal, bacteria belonging to the genus Kristensenera, bacteria belonging to the genus Serimonas, bacteria belonging to the genus Parabacteroides, bacteria belonging to the genus Elyspyrotricus, to the genus Clostridium XIII AD3011 Bacteria belonging to, Bacteria belonging to Lacnospira UCG-001, Bacteria belonging to Lacnospira UCG-003, Bacteria belonging to Faecalitalea, Bacteria belonging to Aristipes, Bacteria belonging to Anaerostipes, Bacteria belonging to Luminococcus 1 Relative abundance of at least one of the bacteria and bacteria belonging to the genus Luminococcus 2 in the stool sample and at least of N, N-dimethylglycine, butyric acid, asparagine and 3-hydroxybutyric acid in the stool sample. Based on at least one of the contents of a predetermined amount of one metabolite in the stool sample, regarding the prediction target, the effect of improving the intestinal environment or the effect of improving the stool condition by ingesting Bifidobacterium longum. Include a prediction step to make a prediction of
   A prediction method characterized by.
2. In the prediction step, based on the relative abundance ratio and the content, the prediction target is predicted for the effect of improving the intestinal environment or the effect of improving the stool condition by ingesting the bifidobacteria longum. ,
   The prediction method according to claim 1.
3. In the prediction step, predicting whether or not there is the improvement effect,
   The prediction method according to claim 1 or 2.
4. The improvement of the bowel movement is the improvement of bowel movement.
   The prediction method according to any one of claims 1 to 3, characterized in that.
5. The improvement in bowel movement is an increase in bowel movement frequency.
   4. The prediction method according to claim 4.
6. The prediction step is executed in the control unit of the information processing device including the control unit.
   The prediction method according to any one of claims 1 to 5, characterized in that.
7. It is a prediction device equipped with a control unit.
   The control unit
   In stool samples collected from the prediction target, eubacterium rectal, bacteria belonging to the genus Kristensenera, bacteria belonging to the genus Serimonas, bacteria belonging to the genus Parabacteroides, bacteria belonging to the genus Elyspyrotricus, to the genus Clostridium XIII AD3011 Bacteria belonging to, Bacteria belonging to Lacnospira UCG-001, Bacteria belonging to Lacnospira UCG-003, Bacteria belonging to Faecalitalea, Bacteria belonging to Aristipes, Bacteria belonging to Anaerotipes, Belonging to Luminococcus Relative abundance of at least one of the bacteria and bacteria belonging to the genus Luminococcus 2 in the stool sample and at least of N, N-dimethylglycine, butyric acid, asparagine and 3-hydroxybutyric acid in the stool sample. Based on at least one of the contents of a predetermined amount of one metabolite in the stool sample, regarding the prediction target, the effect of improving the intestinal environment or the effect of improving the stool condition by ingesting Bifidobacterium longum. To have a predictive means to make predictions
   A prediction device characterized by.
8. It is a prediction program to be executed in an information processing device equipped with a control unit.
   To make the control unit execute
   In stool samples collected from the prediction target, eubacterium rectal, bacteria belonging to the genus Kristensenera, bacteria belonging to the genus Serimonas, bacteria belonging to the genus Parabacteroides, bacteria belonging to the genus Elyspyrotricus, to the genus Clostridium XIII AD3011 Bacteria belonging to, Bacteria belonging to Lacnospira UCG-001, Bacteria belonging to Lacnospira UCG-003, Bacteria belonging to Faecalitalea, Bacteria belonging to Aristipes, Bacteria belonging to Anaerostipes, Bacteria belonging to Luminococcus 1 Relative abundance of at least one of the bacteria and bacteria belonging to the genus Luminococcus 2 in the stool sample and at least of N, N-dimethylglycine, butyric acid, asparagine and 3-hydroxybutyric acid in the stool sample. Based on at least one of the contents of a predetermined amount of one metabolite in the stool sample, regarding the prediction target, the effect of improving the intestinal environment or the effect of improving the stool condition by ingesting Bifidobacterium longum. Include a prediction step to make a prediction of
   A forecasting program featuring.
9. The prediction program according to claim 8 is stored.
   A recording medium characterized by.

Images