JP7527807B2 - Periodontal disease inspection method, periodontal disease inspection system, periodontal disease inspection kit, machine learning device, and data structure - Google Patents

Description

The present disclosure relates to a periodontal disease testing method, a periodontal disease testing system, a periodontal disease testing kit, a machine learning device, and a data structure.

As a method for testing for periodontal disease, various techniques have been considered for determining whether or not a patient has periodontal disease based on the state of specific bacteria in the oral cavity. For example, Non-Patent Document 1 describes the relationship between Fretibacterium and periodontal disease.

Journal of Endodontics, Volume 41, Issue 12, December 2015 pp. 1975-1984

There are several reports, including Non-Patent Document 1, regarding the relationship between the state of specific bacteria in the oral cavity and periodontal disease, but there is still room for further investigation as to what information in the oral cavity is appropriate for determining whether or not a patient has periodontal disease.
In addition, finding a variety of methods for determining whether or not a person has periodontal disease based on information from the oral cavity is thought to be beneficial in terms of improving periodontal disease testing techniques and increasing the options available to those undergoing testing.

In view of the above circumstances, the present invention aims to provide a new machine learning device and data structure for use in periodontal disease testing. The present invention also aims to provide a new periodontal disease testing method, periodontal disease testing system, and periodontal disease testing kit.

Specific means for achieving the above object include the following embodiments.
<1> A periodontal disease testing method, comprising a detection step of detecting at least one bacterium selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp., Peptococcus sp., Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus from a sample collected from the oral cavity.
<2> The periodontal disease testing method described in <1>, further comprising an evaluation step of detecting at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, and Dialister invisus from a sample in the detection step, and evaluating that the donor of the sample is highly likely to be suffering from periodontal disease if the bacteria is detected.
<3> The periodontal disease testing method described in <1>, further comprising an evaluation step of detecting at least one type of bacteria selected from the group consisting of Granulicatella elegans and Haemophilus paraphrohaemolyticus from a sample in the detection step, and evaluating that the donor of the sample is unlikely to be suffering from periodontal disease if the bacteria is detected.
<4> A periodontal disease testing method described in any one of <1> to <3>, wherein the detection is carried out by one type of gene amplification selected from the group consisting of metagenomic analysis, DNA microarray analysis, real-time PCR, and LAMP.
<5> A periodontal disease testing method described in any one of <1> to <4>, wherein the sample is one selected from the group consisting of saliva, mouthwash, plaque, tongue coating, and gingival crevicular fluid.
<6> A periodontal disease testing system comprising a detection unit that detects at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus from a sample collected from the oral cavity.
<7> A periodontal disease test kit comprising a detection device for detecting at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus from a sample collected from the oral cavity.
<8> A machine learning device that learns a function for determining a probability of developing periodontal disease or a state of periodontal disease,
a learning unit that learns the function for determining the judgment value based on a plurality of learning data consisting of: oral cavity information on at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus detected from a specimen collected from the oral cavity; and at least one judgment value selected from the group consisting of a probability of periodontal disease or a state of periodontal disease of a donor of the specimen;
The learning unit includes a reward calculation unit that calculates a reward for a result of determining the probability of developing periodontal disease or the state of periodontal disease using the function based on the learning data;
a function update unit that updates the function so that the reward calculated by the reward calculation unit becomes higher;
A convergence determination unit that repeats the calculation by the reward calculation unit and the update by the function update unit until a predetermined convergence condition is satisfied;
A machine learning device including:
<9> A data structure including oral cavity information on at least one bacterium selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus detected from a specimen collected from the oral cavity, and information identifying a donor of the specimen,
A data structure used in a process of calculating the probability of periodontal disease or condition of the sample provider based on the oral cavity information using a trained model that has been pre-trained to output a value representing the probability of periodontal disease or the periodontal disease condition of the sample provider.

The present invention provides a novel machine learning device and data structure for use in testing for periodontal disease. The present invention also provides a novel method for testing for periodontal disease, a method for diagnosing periodontal disease, a system for testing for periodontal disease, and a kit for testing for periodontal disease.

FIG. 1 is a schematic diagram showing an example of the configuration of a machine learning device 100. FIG. 2 is a conceptual diagram showing an example of a machine learning processing routine of the machine learning device 100.

Hereinafter, an embodiment of the present invention will be described in detail. However, the present invention is not limited to the following embodiment.
In the present disclosure, the term "step" includes not only a step that is independent of other steps, but also a step that cannot be clearly distinguished from other steps as long as the purpose of the step is achieved.

<Machine learning device>
The machine learning device of the present embodiment is a machine learning device that learns a function for determining the probability of developing periodontal disease or the state of periodontal disease,
a learning unit that learns the function for determining the judgment value based on a plurality of learning data consisting of: oral cavity information on at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus detected from a specimen collected from the oral cavity; and at least one judgment value selected from the group consisting of a probability of periodontal disease or a state of periodontal disease of a donor of the specimen;
The learning unit includes a reward calculation unit that calculates a reward for a result of determining the probability of developing periodontal disease or the state of periodontal disease using the function based on the learning data;
a function update unit that updates the function so that the reward calculated by the reward calculation unit becomes higher;
and a convergence determination unit that causes the reward calculation unit to repeat the calculation and the function update unit to repeat the update until a predetermined convergence condition is satisfied.

The above machine learning device can be used to determine whether or not the person providing the sample is suffering from periodontal disease, whether or not periodontal disease has been cured, whether or not periodontal disease will be cured, whether or not periodontal disease will progress, etc.

In one embodiment, saliva is used as a sample, but is not limited to this. Examples include mouthwash, plaque, tongue coating, and gingival crevicular fluid. In particular, by using saliva as a sample, it is possible to obtain information about the donor's oral cavity in a simple manner without damaging the gums, etc., to collect the sample.

At least one judgment value selected from the group consisting of the probability of developing periodontal disease or the state of periodontal disease includes, but is not limited to, a judgment value based on the presence or absence or ratio of a specific bacterium in the bacterial flora composition in the oral cavity.

The learning data may include other information as necessary. Examples of such information include the age, sex, smoking habits, eating habits, oral care status, dental treatment history, chronic or past illnesses, presence or absence of cavities, teeth alignment, amount of saliva, oral hygiene, habits (teeth grinding, mouth breathing, etc.), and stress state of the specimen donor.

The machine learning device of this embodiment may further include a judgment unit that determines the judgment value from the input intraoral information using the function learned by the learning unit.

Figure 1 is a schematic diagram showing an example of the configuration of a machine learning device 100. The machine learning device 100 shown in Figure 1 can be configured with a computer including a CPU, a RAM, and a ROM that stores programs and various data for executing the machine learning processing routine described below. Functionally, this machine learning device 100 includes an input unit 10, a calculation unit 20, and an output unit 90, as shown in Figure 1.

The input unit 10 receives a plurality of learning data consisting of: oral cavity information on at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus detected from a sample; and at least one judgment value selected from the group consisting of the probability of periodontal disease or the state of periodontal disease of the donor of the sample. The input unit 10 also receives oral cavity information detected from the sample to be judged.

The calculation unit 20 includes a learning data storage unit 30, a learning unit 40, a trained model storage unit 50, and a judgment unit 60.

The learning data storage unit 30 stores a plurality of learning data received by the input unit 10 .
The learning unit 40 learns a function for determining a judgment value based on a plurality of learning data. Here, the function is a learned model, which will be described later.
Specifically, the learning unit 40 includes a reward calculation unit 42 , a function update unit 44 , and a convergence determination unit 46 .
The remuneration calculation unit 42 calculates remuneration for the result of determining the probability of developing periodontal disease or the state of periodontal disease using a function based on each of the plurality of learning data.
The function update unit 44 updates the function based on the reward calculated by the reward calculation unit 42 for each of the multiple learning data so as to increase the reward.
The convergence determination unit 46 causes the reward calculation unit 42 to repeat the calculation and the function update unit 44 to repeat the update until a predetermined convergence condition is satisfied.
The trained model storage unit 50 stores the functions trained by the training unit 40 .
The judgment unit 60 determines at least one judgment value selected from the group consisting of the probability of the sample provider suffering from periodontal disease or the state of periodontal disease from the intraoral information detected from the sample input as the judgment target, using the function learned by the learning unit 40.

The components of the machine learning device of this embodiment are not particularly limited and may be components of a known machine learning device.

<Pre-trained model>
The trained model of the present embodiment is, for example, a model in which the weights of the parameters of a neural network model are learned by a deep learning method, and is a model that is trained in advance to input data related to intraoral information detected from a sample and output a value representing the probability of having periodontal disease or the state of periodontal disease of the sample provider. The model is trained by, for example, learning the weights of the parameters so that when data related to actually acquired intraoral information is input, a value representing the probability of having periodontal disease or the state of periodontal disease that is the correct answer is output. The deep learning method may be any method such as GAN (Generative Adversarial Network) or LSTM (Long Short-Term Memory). When data related to intraoral information is input to the trained model trained in this way, a value representing the probability of having periodontal disease or the state of periodontal disease is output.

In one embodiment, saliva is used as a sample, but is not limited to this. Examples include mouthwash, plaque, tongue coating, and gingival crevicular fluid. In particular, by using saliva as a sample, it is possible to obtain information about the donor's oral cavity in a simple manner without damaging the gums, etc., to collect the sample.

Values expressing the probability of developing periodontal disease or the state of periodontal disease include, but are not limited to, values determined based on the presence or absence or proportion of specific bacteria in the composition of the oral bacterial flora.

The trained model may cause a computer to function to output a value representing the probability of having periodontal disease or the state of periodontal disease based on the intraoral information and other information, as necessary. Such information may include, for example, the age, sex, smoking habits, eating habits, oral care status, dental treatment history, chronic or past illnesses, the presence or absence of cavities, teeth alignment, amount of saliva, oral hygiene, habits (teeth grinding, mouth breathing, etc.), and stress state of the specimen donor.

The components of the trained model of this embodiment are not particularly limited and may be similar to the components of known trained models.

<Data structure>
The data structure of this embodiment is a data structure including oral cavity information related to at least one bacterium selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus detected from a specimen collected from the oral cavity, and information identifying a donor of the specimen,
This is a data structure used in the process of calculating the probability of periodontal disease or the state of periodontal disease of the sample provider based on the intraoral information using a trained model that has been pre-trained to output a value representing the probability of periodontal disease or the state of periodontal disease.

The above data structure can be used to determine whether the sample donor is suffering from periodontal disease, whether the periodontal disease has been cured, whether the periodontal disease will be cured, whether the periodontal disease will progress, etc.

In one embodiment, saliva is used as a sample, but is not limited to this. Examples include mouthwash, plaque, tongue coating, and gingival crevicular fluid. In particular, by using saliva as a sample, information on the donor's oral cavity can be obtained in a simple manner without damaging the gums, etc., to collect the sample.

Values expressing the probability of developing periodontal disease or the state of periodontal disease include, but are not limited to, values determined based on the presence or absence or proportion of specific bacteria in the composition of the oral bacterial flora.

The above data structure may be used, as necessary, in a process of calculating the probability of having or being in a state of periodontal disease using a trained model that has been trained in advance to output a value representing the probability of having or being in a state of periodontal disease based on the intraoral information and other information. Such information may include, for example, the age, sex, smoking habits, eating habits, oral care status, dental treatment history, chronic or past illnesses, the presence or absence of cavities, teeth alignment, amount of saliva, oral hygiene, habits (teeth grinding, mouth breathing, etc.), and stress state of the specimen donor.

The components of the trained model of this embodiment are not particularly limited and may be similar to the components of known trained models.

<Operation of machine learning device 100>
When the input unit 10 receives a plurality of pieces of learning data, the machine learning device 100 stores the plurality of pieces of learning data in the learning data storage unit 30. Then, the machine learning device 100 executes a machine learning processing routine shown in FIG.

First, in step S100, for each of a plurality of learning data, a reward is calculated for a result of determining the probability of having periodontal disease or the state of periodontal disease using the current function based on the learning data. For example, the reward is calculated so that the closer the result of determining the probability of having periodontal disease or the state of periodontal disease using the current function is to a judgment value included in the learning data, the higher the reward is.
In step S102, the function is updated based on the reward calculated for each of the multiple learning data in step S100 above, so that the reward becomes higher.
In step S104, it is determined whether or not a predetermined convergence condition is satisfied, and if the convergence condition is not satisfied, the process returns to step S100. On the other hand, if the convergence condition is satisfied, the process proceeds to step S106.
In step S106, the final updated function is stored in the trained model storage unit 50, and the learning process routine is terminated.
Then, when the input unit 10 receives the intraoral information detected from the sample input as the subject of judgment, the judgment unit 60 of the machine learning device 100 determines at least one judgment value selected from the group consisting of the probability of the sample provider suffering from periodontal disease or the state of periodontal disease from the intraoral information detected from the sample input as the subject of judgment using the function learned by the learning unit 40, and outputs it by the output unit 90.

<Periodontal disease testing method>
The periodontal disease testing method of this embodiment includes a detection step of detecting at least one type of bacteria (hereinafter also referred to as the detection target) selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp., Peptococcus sp., Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus from a sample.

The periodontal disease testing method of this embodiment includes a step of detecting at least one type selected from specific bacterial species, thereby improving the accuracy of periodontal disease testing.

The above-mentioned testing method can be used to determine whether or not the sample donor is suffering from periodontal disease, whether or not periodontal disease has been cured, whether or not periodontal disease will be cured, whether or not periodontal disease will progress, etc.

(Detection step)
The detection step may be performed by any method that can detect the target substance contained in the sample. For example, metagenomic analysis (meta 16S analysis, etc.), metagenomic analysis, DNA microarray analysis, real-time PCR, and LAMP are included. In one embodiment, the detection step is performed by metagenomic analysis. The metagenomic analysis can detect the presence or absence and the presence rate of multiple types of bacteria in a single step, and is advantageous in terms of test efficiency compared to the gene amplification method, which is basically detecting one type of bacteria per step.

Specific information that can be detected includes bacterial phylogenetic tree and branching length (Unifrac distance); number of reads (sequence and read amount) obtained by next-generation sequencing; type and amount of interbacterial signaling (quorum sensing, autoinducer); type and amount of mutual nutritional symbiosis between bacteria; bacterial growth conditions (oxygen requirement, nutrition requirement, etc.); type and amount of antigens that affect immune function; type and amount of proteases that destroy biological structures; type and amount of polynucleotides involved in gene expression or gene control; type and amount of endotoxin and lipopolysaccharide that constitutes the outer membrane of bacterial cell walls; and features that represent bacterial functions. Each of these pieces of information may be combined with a predetermined weight, or features extracted by principal component analysis or combinations thereof may be used.

In one embodiment, saliva is used as the sample, but is not limited thereto. Examples include saliva, mouthwash, plaque, tongue coating, and gingival crevicular fluid. In particular, by using saliva as the sample, information on the donor's oral cavity can be obtained in a simple manner without damaging the gums, etc., to collect the sample.

From the viewpoint of improving the accuracy of the test, the detection target is at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus, preferably at least two types of bacteria, and more preferably at least three types of bacteria.

(Evaluation process)
The periodontal disease testing method of this embodiment may include, after the detection step, an evaluation step of evaluating the possibility that the sample provider has periodontal disease based on the results obtained in the detection step. The evaluation step may be performed by any method as long as it is possible to evaluate the possibility that the sample provider has periodontal disease based on the results obtained in the detection step. For example, a computer or a color reaction may be used.

The evaluation process may include an analysis process for calculating the total number of bacteria contained in the sample and the number of bacteria to be detected, and the evaluation may be performed based on the analysis results obtained in the analysis process.

In one embodiment, in the evaluation step, a score is calculated using the ratio of the number of sequences of each of the bacteria to be detected to the total number of sequences obtained in the detection step and the trained model, and if the score is equal to or greater than a predetermined value, it is evaluated that there is a high possibility that the sample donor is suffering from periodontal disease.

In the periodontal disease testing method of this embodiment, "the number of sequences of bacteria to be detected" means the number of sequences of one or more types of bacteria adopted as the detection target, which are bacteria contained in the sample and are included in the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp., Peptococcus sp., Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus.
Therefore, the number of sequences of bacteria that are included in the above group but were not used as detection targets is not included in the "number of sequences of bacteria to be detected."

In the evaluation step, if necessary, the possibility that the sample donor is affected by periodontal disease may be evaluated by referring to other information in addition to the results obtained in the detection step.
Such information may include, for example, the sample provider's age, sex, smoking habits, eating habits, oral care status, dental treatment history, chronic or past illnesses, presence or absence of cavities, teeth alignment, amount of saliva, oral hygiene, habits (teeth grinding, mouth breathing, etc.), and stress level.

<Periodontal disease testing system>
The periodontal disease testing system of this embodiment is equipped with a detection unit that detects at least one type of bacteria (detection target) selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp., Peptococcus sp., Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus from a sample.

The above-mentioned testing system can be used to determine whether or not the person providing the sample has periodontal disease, whether or not periodontal disease has been cured, whether or not periodontal disease will be cured, and whether or not periodontal disease will progress.

(Detection unit)
The detection unit is not particularly limited as long as it can detect the detection target contained in the sample. For example, there are devices that perform metagenomic analysis (meta-16S analysis, etc.), immunochromatography, gene amplification (real-time PCR, LAMP, etc.), etc. In one embodiment, the detection step is performed by metagenomic analysis. The presence or absence and the presence rate of multiple types of bacteria can be detected in a single step, which is advantageous in terms of test efficiency compared to gene amplification methods that basically detect one type of bacteria per step. In addition, the detection accuracy of bacteria is superior to that of the immunochromatography method of qualitative evaluation.

Specific information that can be detected includes bacterial phylogenetic tree and branching length (Unifrac distance); number of reads (sequence and read amount) obtained by next-generation sequencing; type and amount of interbacterial signaling (quorum sensing, autoinducer); type and amount of mutual nutritional symbiosis between bacteria; bacterial growth conditions (oxygen requirement, nutrition requirement, etc.); type and amount of antigens that affect immune function; type and amount of proteases that destroy biological structures; type and amount of polynucleotides involved in gene expression or gene control; type and amount of endotoxin and lipopolysaccharide that constitutes the outer membrane of bacterial cell walls; and features that represent bacterial functions. Each of these pieces of information may be combined with a predetermined weight, or features extracted by principal component analysis or combinations thereof may be used.

The specific configuration of the detection unit is not particularly limited and can be selected according to the detection method to be adopted. For example, the detection unit may perform detection using a calculator such as a computer, or may perform detection through a manual technique such as a test strip.

In one embodiment, saliva is used as the sample, but is not limited to this. Examples include gingival crevicular fluid, dental plaque, biofilm, tongue coating, etc. In particular, by using saliva as the sample, it is possible to obtain information about the donor's oral cavity in a simple manner without damaging the gums, etc., to collect the sample.

From the viewpoint of improving the accuracy of the test, the detection target is at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus, preferably at least two types of bacteria, and more preferably at least three types of bacteria.

(Evaluation Department)
The periodontal disease testing system of this embodiment may include an evaluation unit that evaluates the possibility that the sample provider has periodontal disease based on the results obtained by the detection unit. The evaluation unit is not particularly limited as long as it can evaluate the possibility that the sample provider has periodontal disease based on the results obtained by the detection unit. For example, the evaluation unit may use a computer or a color reaction or the like.

The evaluation unit may include an analysis unit that calculates the total number of sequences obtained in the detection process and the number of sequences for each of the bacteria to be detected, and may perform evaluation based on the analysis results obtained by the analysis unit and the trained model.

In one embodiment, the evaluation unit calculates a score based on the ratio of the number of sequences of each of the bacteria to be detected to the total number of sequences obtained by the detection unit and the trained model, and if the score is equal to or greater than a predetermined value, it is evaluated that there is a high possibility that the sample donor is suffering from periodontal disease.

In the periodontal disease testing system of this embodiment, "the number of sequences of bacteria to be detected" means the number of sequences of one or more types of bacteria adopted as the detection target, which are bacteria contained in the sample and are included in the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp., Peptococcus sp., Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus.
Therefore, the number of bacteria that are included in the above group but were not adopted as detection targets is not included in the "number of sequences of bacteria to be detected."

The periodontal disease testing system of this embodiment may include parts that perform other functions in addition to the detection unit and the evaluation unit and analysis unit that are included as necessary. The periodontal disease testing system of this embodiment may be in the form of a device in which the above-mentioned units are integrated, or may be a combination of devices each including each unit. The periodontal disease testing system of this embodiment may also include simpler forms (for example, the periodontal disease testing kit described below).

<Periodontal disease test kit>
The periodontal disease test kit of this embodiment is equipped with a detection device that detects at least one type of bacteria (detection target) selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp., Peptococcus sp., Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus from a sample.

The detection device is not particularly limited as long as it can detect the target substance contained in the sample. For example, it may be one described as the detection unit included in the periodontal disease testing system described above. In addition, it may be equipped with a device that performs evaluation, analysis, or other functions as necessary.
When the periodontal disease test kit includes these devices, these devices may be integrated with the detection device (including cases where the detection device also performs the functions of these devices) or may be a combination of separate devices.

The above embodiment will be described in detail below with reference to examples, but the above embodiment is not limited to these examples.

Example 1
The severity of periodontal disease for each subject was classified as healthy (no periodontitis or gingivitis), mild (PD: stage I, CAL: 1-2mm), moderate (PD: stage II, CAL: 3-4mm), or severe (PD: stage III-IV, CAL: ≧5mm), and each was assigned a number: 0, 1, 2, or 3. The species and amount of bacteria in saliva were measured from the subjects, and the species with the greatest impact on the severity were analyzed using statistical machine learning techniques to find bacteria with the greatest impact on periodontal disease. PD is pocket depth, and CAL is clinical attachment level. Clinical attachment level refers to the distance from the boundary between the tooth enamel and cementum to the bottom of the periodontal pocket.

(Detection Method)
Bacterial genomic DNA was extracted from the saliva of each subject, and the V3-V4 region (F: 5'-CCTACGGGNGGCWGCAG-3', R: 5'-GACTACHVGGGTATCTAATCC-3') or V4 region (F: 5'-GTGCCAGCMGCCGCGGTAA-3', R: 5'-GGACTACHVGGGTWTCTAAT-3') as the region of 16SrDNA was amplified by PCR (Thermo Fisher, SimpliAmp) using primers (for V3-V4 or for V4) containing region-specific sequences. An adapter sequence and an index sequence are added to the amplified PCR fragment to construct a library. The library is then input into a next-generation sequencer MiSeq (Illumina), and the base sequence (Read) is obtained.

Then, the obtained sequences are compared with bacterial genome databases (public database 16SMicrobial, NCBI, etc.) using a computer via BLAST, and the bacterial flora composition is revealed from the number of sequences (reads) (meta-16S analysis). Next, the expression ratios are calculated for each of the 310 types of bacteria.

(Review of detection results)
Next, using each of the five models, namely Random Forest, Extremely Randomized Trees, XGBoost (Random Forest + Gradient Boosting), LightGBM, and Regularized Greedy Forest, the importance of each of the 310 types of bacteria relative to the severity was calculated based on the expression level ratio and severity of the bacteria, and the bacteria were arranged in descending order of absolute importance, as shown in Table 1 below.

From Table 1 above, we found that bacteria with a higher absolute importance to periodontal disease than P. gingivalis, which is said to have the highest periodontal disease pathogenicity, excluding bacteria known to affect periodontal disease, include Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus.

In addition, the codes for the importance of periodontal disease show that, when Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, and Dialister invisus are detected, it is assessed that the subject is highly likely to have periodontal disease.

In addition, the codes for the importance of Granulicatella elegans and Haemophilus paraphrohaemolyticus to periodontal disease indicate that when these bacteria are detected, the possibility that the subject is suffering from periodontal disease is assessed to be low.

These results demonstrate that a periodontal disease testing method that includes a step of detecting at least one type of bacteria selected from the group consisting of Filifactor alocis, Veillonella atypica, Catonella sp, Peptococcus sp, Dialister invisus, Abiotrophia defectiva, Lachnoanaerobaculum umeaense, Prevotella pallens, Prevotella histicola, Granulicatella elegans, and Haemophilus paraphrohaemolyticus from a sample is useful for evaluating the possibility that a sample donor is suffering from periodontal disease.

REFERENCE SIGNS LIST 10 Input unit 20 Calculation unit 30 Learning data storage unit 40 Learning unit 42 Reward calculation unit 44 Function update unit 46 Convergence determination unit 50 Model storage unit 60 Judgment unit 90 Output unit 100 Machine learning device

Claims (6)

A detection step of detecting at least one bacterium selected from the group consisting of Catonella sp., Peptococcus sp ., and Haemophilus paraphrohaemolyticus from a human-derived specimen collected from the oral cavity ;
In the detection step, at least one bacterium selected from the group consisting of Catonella sp. and Peptococcus sp. is detected from the sample, and if the bacterium is detected, it is evaluated that the donor of the sample is highly likely to be suffering from periodontal disease;
and an evaluation step of evaluating that the donor of the sample is unlikely to be affected by periodontal disease if Haemophilus paraphrohaemolyticus is detected from the sample in the detection step. 2. The periodontal disease testing method according to claim 1 , wherein the sample is one selected from the group consisting of saliva, mouthwash, plaque, tongue coating, and gingival crevicular fluid. A detection unit for detecting at least one bacterium selected from the group consisting of Catonella sp., Peptococcus sp ., and Haemophilus paraphrohaemolyticus from a human-derived specimen collected from the oral cavity ;
detecting at least one bacterium selected from the group consisting of Catonella sp. and Peptococcus sp. from the specimen in the detection unit, and evaluating that the donor of the specimen is highly likely to be suffering from periodontal disease when the bacterium is detected;
and an evaluation unit that evaluates that the donor of the sample is unlikely to be affected by periodontal disease if Haemophilus paraphrohaemolyticus is detected in the sample by the detection unit. A detection device for detecting at least one bacterium selected from the group consisting of Catonella sp., Peptococcus sp ., and Haemophilus paraphrohaemolyticus from a human-derived specimen collected from the oral cavity ,
detecting at least one bacterium selected from the group consisting of Catonella sp. and Peptococcus sp. from a specimen using the detection device, and evaluating that the donor of the specimen is highly likely to be suffering from periodontal disease if the bacterium is detected;
A periodontal disease testing kit which, when Haemophilus paraphrohaemolyticus is detected from a sample by the detection device, evaluates that the possibility that the donor of the sample is suffering from periodontal disease is low . A machine learning device that learns a function for determining a probability of developing periodontal disease or a state of periodontal disease,
a learning unit that learns the function for determining the judgment value based on a plurality of learning data consisting of a detection result of at least one type of bacteria selected from the group consisting of Catonella sp., Peptococcus sp ., and Haemophilus paraphrohaemolyticus from a human-derived specimen collected from the oral cavity, and at least one judgment value selected from the group consisting of a probability of periodontal disease or a state of periodontal disease of the donor of the specimen;
The learning unit includes a reward calculation unit that calculates a reward for a result of determining the probability of developing periodontal disease or the state of periodontal disease using the function based on the learning data;
a function update unit that updates the function so that the reward calculated by the reward calculation unit becomes higher;
A convergence determination unit that repeats the calculation by the reward calculation unit and the update by the function update unit until a predetermined convergence condition is satisfied;
Including,
The function is
determining the judgment value indicating a high possibility that the donor of the sample is suffering from periodontal disease when at least one species of bacteria selected from the group consisting of Catonella sp. and Peptococcus sp. is detected from the sample;
A machine learning device that determines the judgment value indicating that the donor of the sample is unlikely to be suffering from periodontal disease when Haemophilus paraphrohaemolyticus is detected from the sample. A data structure including a detection result of at least one bacterium selected from the group consisting of Catonella sp., Peptococcus sp ., and Haemophilus paraphrohaemolyticus from a human-derived specimen collected from the oral cavity, and information for identifying a donor of the specimen,
A data structure used in a process of calculating a probability regarding the occurrence or state of periodontal disease of the sample provider using a trained model that has been trained in advance to output a value representing the probability of occurrence or state of periodontal disease of the sample provider based on the detection results, the process calculating the probability indicating that the sample provider is highly likely to have periodontal disease when at least one type of bacteria selected from the group consisting of Catonella sp and Peptococcus sp is detected from the sample, and calculating the probability indicating that the sample provider is low likely to have periodontal disease when Haemophilus paraphrohaemolyticus is detected from the sample.