Method and device for judging development maturity of intestinal flora of infant

Abstract

The invention provides an infant intestinal flora development maturity discrimination method and device. The method comprises the following steps: acquiring representative bacteria data of an infant to be discriminated; inputting the representative bacteria data of the infant to be discriminated into a pre-constructed infant intestinal flora development maturity discrimination model to obtain an intestinal flora development state discrimination result; wherein the infant intestinal flora development maturity discrimination model is obtained by performing training by using a large number of infant intestinal flora data samples with labels based on a random forest model, selecting a training result of which the accuracy meets a screening condition, screening representative bacteria according to a data-model contribution degree evaluation value, and performing simplification. According to the method, simplification is carried out on the basis of a model established by using a random forest method, the obtained discrimination model can predict the development state of the intestinal flora by inputting data of representative bacteria of infants, the error rate is low, and the accuracy is high.

Description

Technical Field

[0001] The present invention relates to the technical field of intestinal flora prediction, and in particular to a method and device for judging the maturity of intestinal flora in infants and young children. Background Art

[0002] The intestinal microbiota refers to a group of microorganisms living in the human gastrointestinal tract. It plays an important role in resisting the invasion of external pathogenic microorganisms, intestinal nutrient absorption, immune regulation, etc., and has an important impact on the healthy development of individuals. The intestinal flora is not static, but is in a process of constant change, affected by factors such as diet, disease, and age. In the life course of a healthy individual, the infant period is a critical period for the establishment of the intestinal flora, especially for the newborn period, which is an important window period for the establishment of the intestinal flora. This period is also the period of initialization of the intestinal flora. The intestinal flora of infants is transmitted by the mother and is affected by factors such as the mode of delivery and feeding, showing different development trajectories. As children continue to grow, the intestinal flora gradually transitions from a low-diversity, low-abundance, immature, and unstable intestinal flora state to a high-diversity, high-abundance, and more stable intestinal flora state. This process lasts about three to five years. The intestinal development state of infants and young children gradually changes with age. When the intestine is immature or matures prematurely, it will have an impact on the health of infants and young children.

[0003] In the prior art, the research method for intestinal flora mainly relies on the use of sequencing technology to sequence the intestinal flora marker genes (V3V4 region of 16SrRNA), and further use bioinformatics technology to process the sequencing results. The data processing method analyzes and processes the data to achieve observation of the intestinal flora. However, the result output of the existing model has high data requirements for the input samples, and often requires all the intestinal flora data of an infant. The intestinal flora of infants and young children has individual differences. In practical applications, the model will be misjudged due to missing data. In addition, the existing database sample range is small, which will cause the trained model to only distinguish the development age of the intestinal flora of infants in a fixed age group, which will also cause model misjudgment.

[0004] In summary, the existing technology has the problems of high error rate and low accuracy. Summary of the invention

[0005] The present invention provides a method and device for determining the maturity of intestinal flora in infants and young children, so as to solve the defects of high error rate and low accuracy in the prior art, and realize the determination of the maturity of intestinal flora in infants and young children with low error rate and high accuracy.

[0006] The present invention provides a method for determining the maturity of intestinal flora in infants and young children, comprising the following steps:

[0007] Obtain representative bacteria data of infants and young children to be identified;

[0008] Inputting the representative bacteria data of infants to be identified into a pre-constructed infant intestinal flora development maturity identification model to obtain an intestinal flora development status identification result;

[0009] Among them, the infant intestinal flora development maturity discrimination model is based on the random forest model, which is trained using labeled infant intestinal flora data samples, and the training results whose accuracy meets the screening conditions are selected and simplified by screening representative bacteria according to the data-model contribution evaluation value.

[0010] According to a method for distinguishing the maturity of intestinal flora in infants and young children provided by the present invention, based on a random forest model, a labeled infant intestinal flora data sample is used for training, and a training result whose accuracy meets the screening condition is selected, and representative bacteria are selected according to the data-model contribution evaluation value to simplify the infant intestinal flora development maturity discrimination model, which specifically includes:

[0011] S1: Obtain a large number of labeled infant intestinal flora data samples to obtain a training data set; wherein the labels include species annotations and age labels;

[0012] S2: Divide the training data set into a test set and a validation set according to a preset ratio;

[0013] S3: construct a random seed according to the first number of randomly generated random numbers, and set basic parameters for constructing a random forest model;

[0014] S4: constructing a random forest model according to the basic parameters using the random seed, the test set and the validation set;

[0015] S5: Performing accuracy detection on the first number of first random forest models obtained, and selecting the first random forest model whose accuracy meets the screening condition as the training result;

[0016] S6: For the current training result, the importance of the species annotations in the training process of the training result is ranked using the data-model contribution evaluation value to obtain a ranking result;

[0017] S7: Select the first target number of the species annotations in the sorting result as representative bacteria, screen the training data set according to the species annotations, delete the infant intestinal flora data samples that do not have the species annotations, and update the training data set; execute steps S2-S4 based on the updated training data set to obtain a first number of second random forest models; wherein the target number is a random value within a preset range;

[0018] S8: Update the target number, and repeat steps S6-S8 until a second random forest model of all the training results is obtained;

[0019] S9: Perform a simplified effect evaluation on the second random forest model according to the model prediction error rate, and select the second random forest model with the lowest model prediction error rate as the model for distinguishing the developmental maturity of the infant intestinal flora.

[0020] According to a method for determining the maturity of intestinal flora in infants and young children provided by the present invention, a large number of labeled intestinal flora data samples of infants and young children are obtained, which specifically includes:

[0021] Collect a large number of infant and young children fecal samples according to multiple preset sampling age nodes;

[0022] Extracting the infant intestinal flora sequencing data of the infant feces sample;

[0023] Splicing the infant intestinal flora sequencing data, and extracting characteristic sequences according to the splicing results;

[0024] Comparing the characteristic sequence with a preset database, filtering chimeras, and obtaining an initial infant intestinal flora data sample;

[0025] Aggregating the initial infant intestinal flora data samples based on similarity to obtain a set of initial infant intestinal flora data samples belonging to multiple classification units;

[0026] Annotating the initial infant intestinal flora data sample set corresponding to the classification unit according to the preset database to obtain a process infant intestinal flora data sample with the species annotation label;

[0027] The process infant intestinal flora data samples are merged and grouped according to the preset age groups based on the sampling age nodes, and the process infant intestinal flora data samples corresponding to the age groups are marked with age groups to obtain infant intestinal flora data samples with both the species annotation label and the age label.

[0028] According to a method for determining the maturity of intestinal flora in infants and young children provided by the present invention, the accuracy of the first random forest models obtained is tested, specifically comprising:

[0029] The ratio of the number of correct predictions to the total number of predictions is calculated, and the accuracy of the first random forest model generated under different random seeds is tested based on the ratio.

[0030] According to a method for determining the developmental maturity of intestinal flora in infants and young children provided by the present invention, the first number of random numbers includes a second number of three-digit random numbers and the remaining four-digit random numbers.

[0031] According to a method for judging the developmental maturity of intestinal flora in infants and young children provided by the present invention, the sampling age nodes include the seventh day, the fifteenth day, the first month, the second month, the third month, the fourth month, the fifth month, the sixth month, the eighteenth month, the twenty-fourth month, the thirtieth month, the thirty-sixth month, the fourth year, the fifth year and the sixth year.

[0032] The present invention also provides a device for determining the maturity of intestinal flora in infants and young children, comprising the following modules:

[0033] An input unit, used to obtain representative bacteria data of infants to be identified;

[0034] A discrimination unit, used for inputting the representative bacteria data of infants to be discriminated into a pre-constructed discrimination model for the developmental maturity of infant intestinal flora to obtain a discrimination result of the developmental state of intestinal flora;

[0035] The training unit is used to train based on a random forest model using a large number of labeled infant intestinal flora data samples, select training results whose accuracy meets the screening conditions, and select representative bacteria according to the data-model contribution evaluation value to simplify and obtain the infant intestinal flora development maturity discrimination model.

[0036] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements any of the above-described methods for determining the maturity of the intestinal flora of infants and young children.

[0037] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for determining the maturity of the intestinal flora of infants.

[0038] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for determining the maturity of the intestinal flora of infants.

[0039] The method and device for distinguishing the maturity of intestinal flora of infants provided by the present invention obtain the data of representative bacteria of infants to be distinguished; input the data of representative bacteria of infants to be distinguished into a pre-constructed discrimination model of intestinal flora development maturity of infants to obtain the discrimination result of intestinal flora development status; wherein the discrimination model of intestinal flora development maturity of infants is based on a random forest model, and is trained using a large number of labeled infant intestinal flora data samples, and the training results whose accuracy meets the screening conditions are selected to simplify the representative bacteria according to the data-model contribution evaluation value. The present invention is simplified on the basis of the model established by the random forest method, and the obtained discrimination model can predict the development status of intestinal flora by inputting the data of representative bacteria of infants, with low error rate and high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 It is a schematic diagram of the process of the method for determining the maturity of the intestinal flora of infants and young children provided by the present invention.

[0042] Figure 2 This is a simplified effect evaluation table of the second random forest model of an embodiment of the method for discriminating the developmental maturity of the infant intestinal flora provided by the present invention.

[0043] Figure 3 It is a schematic diagram of the structure of the device for determining the maturity of intestinal flora in infants and young children provided by the present invention.

[0044] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Combine the following Figure 1-Figure 2 The present invention describes the method for determining the maturity of the intestinal flora of infants and young children. Figure 1: is a schematic diagram of the process of the method for determining the maturity of the intestinal flora of infants and young children provided by the present invention, such as Figure 1 As shown, the method includes:

[0047] Step 110: Obtain representative bacteria data of the infants to be identified.

[0048] It should be noted that what is obtained is the representative bacteria data of the intestinal flora of the infants to be identified. The representative bacteria refers to the representative bacteria model selected in the process of building the infant intestinal flora development maturity discrimination model, that is, in the training process, the first target number of species annotations in the sorting results are selected as the corresponding bacterial species of the representative bacteria. The representative bacteria data is the absolute abundance data of the representative bacteria.

[0049] The present invention has lower requirements for input data, and only the absolute abundance of representative bacteria corresponding to the model needs to be provided to achieve the measurement of the sample.

[0050] The present invention does not limit the collection method of the representative bacteria data of infants to be identified. In actual implementation, the collection method can be selected according to actual conditions.

[0051] Step 120: inputting the representative bacteria data of infants to be identified into a pre-constructed model for identifying the maturity of the intestinal flora of infants to obtain the identification result of the development status of the intestinal flora.

[0052] In step 120, the evaluation is performed by inputting the data of representative bacteria of infants and young children into the infant intestinal flora development maturity discrimination model, so as to predict the development status of intestinal flora. It should be emphasized that the output result of the model is the judgment of the intestinal development status of the input sample. Based on the result, the difference between the development status of intestinal flora and the actual age status can be observed, so as to improve the intestinal development status in time and reduce the adverse effects caused by unhealthy intestines.

[0053] It should be noted that the infant intestinal flora development maturity discrimination model referred to in the present invention is constructed based on the random forest model on the basis of machine learning.

[0054] It is understandable that machine learning is a technology that simulates human autonomous learning of data through computer programs, thereby continuously obtaining and optimizing the program itself. According to the desired results, it can be divided into classification tasks and regression tasks. Regression tasks mainly deal with discrete value type data. By taking part of the training data as the independent variable and the expected observation target as the dependent variable, the program automatically learns and derives the relationship between the dependent variable and the independent variable; classification tasks are mainly for discrete value type data, and the data is divided into different categories by reducing the dimension and summarizing the features of the training data. In the field of intestinal flora research, machine learning usually performs classification tasks and clusters the intestinal flora. Generally, according to the flora sequencing data and basic clinical grouping information, different data processing methods and machine learning algorithms are selected to study the intestinal flora sequencing information under different groups, so as to realize the association between intestinal flora and clinical characteristics. In the machine learning model established with intestinal flora data as training samples, the commonly used algorithms are support vector machine, K nearest neighbor, random forest and artificial neural network.

[0055] In addition, the random forest model is an integrated learning algorithm based on decision trees. The main principle is to randomly extract samples from the original training data set using the autonomous sampling method, produce multiple training subsets, establish a decision tree based on each subset, loop the sampling steps, generate a specified number of decision trees, and vote based on the classification results of the majority of decision trees to achieve the final classification judgment of the sample. The random forest model provides importance assessment and model influence weights for the characteristic variables of the data sample. When processing multi-dimensional data, it assigns different importance assessments to the data, thereby selecting different features of the training data, further ignoring unimportant features, and further optimizing the model. In addition, the random forest algorithm has the advantages of fast data processing, high accuracy, ability to suppress data overfitting problems, and ability to process large amounts of data.

[0056] Furthermore, in the process of constructing a model for discriminating the developmental maturity of the infant intestinal flora, the model is based on a random forest model, which is trained using a large number of labeled infant intestinal flora data samples. The training results whose accuracy meets the screening conditions are selected and simplified by screening representative bacteria based on the data-model contribution evaluation value.

[0057] The following is a further description of the training steps of the infant intestinal flora development maturity discrimination model. In some embodiments, a large number of labeled infant intestinal flora data samples are used for training, and the training results whose accuracy meets the screening conditions are selected to select representative bacteria according to the data-model contribution evaluation value to simplify the infant intestinal flora development maturity discrimination model, specifically including:

[0058] S1: Obtain a large number of labeled infant intestinal flora data samples to obtain a training data set; wherein the labels include species annotations and age labels.

[0059] In step S1, a data sample of the intestinal flora of infants and young children with species annotation labels and age labels is obtained. The present invention will provide a method for constructing a data sample of the intestinal flora of infants and young children with labels in subsequent embodiments.

[0060] S2: Divide the training data set into a test set and a validation set according to a preset ratio.

[0061] In step S2, the labeled data is input into the random forest model program, and the data is divided into a test set and a validation set according to a preset ratio. In a specific embodiment, 95% of the randomly selected data set is the test set, and the remaining 5% is the validation set.

[0062] S3: construct a random seed according to the first number of randomly generated random numbers, and set basic parameters for constructing a random forest model.

[0063] In step S3, a first number of random numbers are randomly generated, and a random seed is constructed according to the generated random numbers. The embodiment of the present invention does not limit the method of generating random numbers. In some embodiments, the random number function RANDBETWEEN in Excel is used to randomly generate random numbers.

[0064] In addition, the embodiment of the present invention does not limit the first number. In a specific embodiment, the first number is set to three hundred.

[0065] Further, the generated random numbers are further described below. In some embodiments, the first number of random numbers includes a second number of three-digit random numbers and the remaining four-digit random numbers.

[0066] Specifically, this embodiment further limits the number of digits of the random numbers. The first number of random numbers includes a second number of three-digit random numbers, and the rest are four-digit random numbers.

[0067] In a specific embodiment, three hundred random numbers are randomly generated, one hundred of which are three-digit random numbers and the remaining two hundred random numbers are four-digit random numbers.

[0068] Furthermore, after obtaining the random number, a first number of different random seeds are set according to the generated random number using the method of seed=random value.

[0069] Afterwards, the basic parameters for constructing the random forest model are set. For example, in a specific embodiment, in Rstudio software, the randomfrost package is used to construct the random forest model, and in the random forest program, the number of decision trees designed for each random forest model is 500.

[0070] The present invention adopts a fixed random seed to better replicate the model in practical applications, avoids the situation where the random forest model program establishes different random forest models due to different random seeds, and provides a more stable and accurate replicable random model.

[0071] S4: Using the random seed, the test set and the validation set, a random forest model is constructed according to the basic parameters.

[0072] In step S4, a random forest model is constructed according to the aforementioned settings.

[0073] S5: Performing accuracy detection on the first number of first random forest models obtained, and selecting the first random forest models whose accuracy meets the screening condition as the training result.

[0074] In step S5, the accuracy of the first random forest model is tested, and the present invention will further explain the accuracy test method in the subsequent embodiments. Afterwards, the first random forest model whose accuracy meets the screening condition is selected as the result of this training.

[0075] In a specific embodiment, the screening criteria include an accuracy greater than 90%.

[0076] S6: For the current training result, the importance of the species annotations in the training process of the training result is ranked using the data-model contribution evaluation value to obtain a ranking result.

[0077] In step S6, the training results obtained above are simplified, and the main step of simplification is to sort and select according to the different model contribution evaluation values ​​given to the data in the random forest model generation process. Specifically, the mean decrease accuracy value and the mean decrease gini value are used for sorting and selection.

[0078] It can be understood that the mean decrease accuracy value indicates the degree of decrease in the prediction accuracy of the random forest. The larger the value, the greater the importance of the variable. The mean decrease gini value is used to calculate the impact of each variable on the heterogeneity of the observed values ​​at each node of the classification tree, thereby comparing the importance of the variable. The larger the value, the greater the importance of the variable.

[0079] According to the mean decrease accuracy value and mean decrease gini value, the species annotations in the training process of the training results are sorted.

[0080] S7: Select the first target number of the species annotations in the sorting results as representative bacteria, screen the training data set according to the species annotations, delete the infant intestinal flora data samples that do not have the species annotations, and update the training data set; execute steps S2-S4 based on the updated training data set to obtain a first number of second random forest models; wherein the target number is a random value within a preset range.

[0081] In step S7, for the training results screened out in step S5, the species annotations with the highest evaluation values ​​are selected in turn, the training data set is screened according to the selected species annotations, the infant intestinal flora data samples without species annotations are deleted, and the training data set is updated. A random forest model is established again according to the updated training data set to obtain a second random forest model. The steps for establishing the second random forest model are the same as the steps for establishing the first random forest model, and the present invention will not be repeated here.

[0082] S8: Update the target number and repeat steps S6-S8 until a second random forest model of all the training results is obtained.

[0083] It should be noted that this embodiment does not limit the number of times the training results generate the second random forest model, that is, it does not mean that the first random forest model corresponding to each training result with an accuracy that meets the screening condition can only select the first target number of species annotations once to screen the training data set and reconstruct the random forest model based on the new training data set. For each training result, the second random forest model can be regenerated multiple times.

[0084] S9: Perform a simplified effect evaluation on the second random forest model according to the model prediction error rate, and select the second random forest model with the lowest model prediction error rate as the model for distinguishing the developmental maturity of the infant intestinal flora.

[0085] In step S9, the simplification effect of the model is evaluated using the error rate predicted by the model, the optimal model is evaluated according to the error rate table, and the second random forest model with the lowest error rate is selected as the infant intestinal flora development maturity discrimination model. It should be noted that the present invention does not limit the calculation method of the error rate predicted by the model.

[0086] The present invention optimizes the error rate by simplifying the model and reduces the difficulty of processing input data.

[0087] The following further describes a method for constructing labeled infant intestinal flora data samples. In some embodiments, obtaining a large number of labeled infant intestinal flora data samples specifically includes:

[0088] Step S11: Collect a large number of infant feces samples according to a plurality of preset sampling age nodes.

[0089] In step S11, firstly, infant feces samples are collected according to a plurality of preset sampling age nodes. It should be noted that the present invention does not limit the collection area of ​​infant feces samples. In order to further improve the accuracy of the final trained model, the collection area of ​​infant feces samples should cover the representative area of ​​the country where the model is applied as much as possible.

[0090] In addition, the sampling age nodes need to cover as many representative growth stages of infants as possible. In some embodiments, the sampling age nodes include the seventh day, the fifteenth day, the first month, the second month, the third month, the fourth month, the fifth month, the sixth month, the eighteenth month, the twenty-fourth month, the thirtieth month, the thirty-sixth month, the fourth year, the fifth year, and the sixth year.

[0091] The data set constructed in this way covers a wider range of infant and toddler ages and provides more reference value for actual testing.

[0092] Step S12: extracting the infant intestinal flora sequencing data of the infant feces sample.

[0093] In step S12, the collected fecal samples of infants and young children are subjected to bacterial flora sequencing data extraction. The embodiment of the present invention is described by taking the Illumina NovaSeq sequencing platform as an example, but this does not represent a limitation of the present invention, and other sequencing platforms may be used to extract data.

[0094] In a specific implementation, 16S amplicon sequencing technology is used to perform double-end sequencing on a library including infant fecal samples based on the Illumina NovaSeq sequencing platform to obtain sequencing data corresponding to the infant intestinal flora.

[0095] Step S13: splicing the infant intestinal flora sequencing data, and extracting characteristic sequences according to the splicing results.

[0096] In step S13, the obtained infant intestinal flora sequencing data is spliced ​​and feature sequences are extracted. The embodiment of the present invention is described using the qiime bioinformatics platform as an example, but this does not limit the present invention, and other bioinformatics platforms can be used to splice and extract data.

[0097] In a specific embodiment, the fastq files downloaded from the sequencing machine (i.e., infant intestinal flora sequencing data) are spliced ​​based on the qiime bioinformatics platform, and then tags (feature sequences) are intercepted and filtered.

[0098] Step S14: compare the characteristic sequence with a preset database, filter out chimeras, and obtain an initial infant intestinal flora data sample.

[0099] In step S14, the tags (characteristic sequences) obtained by quality control are compared with a preset database, chimeras are filtered out, and an initial infant intestinal flora data sample is obtained.

[0100] In a specific embodiment, the preset database is the Silva database.

[0101] Step S15: Aggregating the initial infant intestinal flora data samples based on similarity to obtain a set of initial infant intestinal flora data samples belonging to multiple classification units.

[0102] In step S15, the initial infant intestinal flora data sample is aggregated into OTUs (multiple taxonomic units). It is understandable that OTUs (Operational Taxonomic Units) are a feature in microbiome data analysis. OTU is a classification unit based on sequence similarity, which is used to represent a group of sequence variants under a specific threshold.

[0103] Step S16: annotating the initial infant intestinal flora data sample set corresponding to the classification unit according to the preset database to obtain a process infant intestinal flora data sample with the species annotation label.

[0104] In step S16, the preset database is used to perform species annotation and mark the species annotation label on the aggregated initial infant intestinal flora data sample to obtain a process infant intestinal flora data sample with the species annotation label. In a specific embodiment, the data results are presented in the form of absolute abundance, involving 1572 taxonomic units at the genus level. It should be explained that absolute abundance refers to the actual number of a particular organism or gene (such as bacteria or their gene copies) in a given sample. Unlike relative abundance, absolute abundance reflects the actual number of the organism or gene in the sample, rather than its proportion relative to other organisms or genes.

[0105] Step S17: The process infant intestinal flora data samples are merged and grouped according to the preset age groups based on the sampling age nodes, and the process infant intestinal flora data samples corresponding to the age groups are marked with age groups to obtain infant intestinal flora data samples with both the species annotation label and the age label.

[0106] In step S17, multiple sampling age nodes are divided into a certain number of age groups, and the process infant intestinal flora data samples are merged and grouped according to the corresponding sampling time. Afterwards, the process infant intestinal flora data samples of each age group are marked with age groups to obtain infant intestinal flora data samples with both the species annotation label and the age label.

[0107] The accuracy detection is further described below. In some embodiments, the accuracy detection is performed on the first number of first random forest models obtained, specifically including:

[0108] The ratio of the number of correct predictions to the total number of predictions is calculated, and the accuracy of the first random forest model generated under different random seeds is tested based on the ratio.

[0109] Specifically, an ensemble learning method is used to test the accuracy of the models generated under different random seeds according to the method of correct number of predictions / total number of predictions, and obtain the first number of random forest models with accuracy evaluation. It should be explained that ensemble learning refers to combining the prediction results of multiple models to improve the overall performance and stability of the model, which can improve the prediction accuracy and enhance the generalization performance of the model.

[0110] The present invention uses multiple random seeds and ensemble learning verification, which can effectively improve the accuracy of the model.

[0111] In order to provide a more detailed description of the method for distinguishing the maturity of intestinal flora in infants and young children provided by the present invention, a specific embodiment is given. In this embodiment, based on the maternal and infant nutrition cohort research project launched in 2014 across multiple representative regions and long-term tracking in my country, the project established a multi-point multi-center, prospective maternal and infant nutrition cohort across 7 provinces and cities, 8 dietary circles, and 15 points, with 3766 mothers and infants enrolled. Samples were collected at these time points on the seventh day, the fifteenth day, the first month, the second month, the third month, the fourth month, the fifth month, the sixth month, the eighteenth month, the twenty-fourth month, the thirtieth month, the thirty-sixth month, the fourth year, the fifth year, and the sixth year, and 28,354 biological samples such as breast milk, feces, and urine were collected. 1,180 infant fecal samples were selected as the subsequent design data set.

[0112] For the infant fecal samples obtained, the 16S amplicon sequencing technology was used to perform double-end sequencing on the library based on the Illumina NovaSeq sequencing platform to obtain the sequencing data fstaq file corresponding to the infant intestinal flora. The fastq file downloaded from the sequencing machine was spliced ​​on the qiime 2 platform, and then the tags were intercepted and filtered. The tags obtained by quality control were compared with the database (Silva database), chimeras were filtered, and then aggregated into OTUs, and the Silva database was used for species annotation.

[0113] Afterwards, they are combined and grouped by age and labeled to obtain infant intestinal flora data samples, and then the training data set is obtained.

[0114] The age groups include the following five groups: Group 1: covers all samples within the first month after birth; Group 2: covers all samples from the fourth to the sixth month after birth; Group 3: covers all samples from the eighteenth to the twenty-fourth month after birth; Group 4: covers all samples from the twenty-fifth to the thirty-sixth month after birth; Group 5: covers all samples from the fourth to the sixth year after birth.

[0115] After that, the labeled data is input into the random forest model program, and the validation set and the test set are divided first. At the same time, three hundred random numbers are randomly generated, one hundred of which are three-digit random numbers, and the remaining two hundred random numbers are four-digit random numbers. The seed = random value method is used, and according to the generated random numbers, three hundred different random seeds are set to be calculated. In addition, in the random forest program, 500 decision trees are designed and established for decision judgment. According to the above settings, 300 first random forest models are constructed.

[0116] Afterwards, the ensemble learning method was used to test the accuracy of the first random forest model generated under different random seeds according to the method of predicting the correct number / total number of predictions. Three hundred random forest models with accuracy evaluation were obtained. The random forest models with an accuracy greater than 90% were selected as the training results for the next step of model simplification. There were 17 models that met the conditions. The random number seed numbers and model accuracy are shown in Table 1.

[0117] Table 1. Training results that meet the screening criteria

[0118]

[0119]

[0120] The training results were simplified, and the species annotations were sorted according to the mean decrease accuracy value and the mean decrease gini value. The sorting of bacterial species corresponding to the first 24 species annotations is shown in Table 2.

[0121] Table 2. The first 24 species annotations were used as representative bacterial sorting results

[0122]

[0123]

[0124] For the 17 random forest models with an accuracy greater than 90%, different numbers (5-30) of OTUs with the highest evaluation values ​​were selected to build random forest models again (i.e., the second random model), and the error rate of model prediction was used as the evaluation model simplification effect. The error rates of simplified models built with different numbers of OTUs are shown in Figure 2 .

[0125] The optimal model was selected and evaluated based on the error rate table. Finally, a random forest model with a random number seed of 787 and 24 OTUs was obtained. The overall accuracy of the model was 93.2203%, the error rate was 16.7707%, and the error rate of the simplified model was 14.6298%, which is the discriminant model for the developmental maturity of the intestinal flora of infants and young children.

[0126] After obtaining the discriminant model for the maturity of intestinal flora in infants and young children, 10 samples from each group were randomly selected from the samples, for a total of 50 samples, and the relative abundance of 24 bacteria in the corresponding samples was extracted to form a test set. The test set was input into the random forest model, and the model test results were obtained after model judgment as shown in Table 3.

[0127] Table 3. Judgment results of the model for distinguishing the maturity of intestinal flora in infants and young children

[0128]

[0129] It can be seen that the model fitting effect is good, and the predictions for each group of samples are consistent with reality.

[0130] The infant intestinal flora development maturity discrimination model obtained in this embodiment has lower requirements for input data, and only the absolute abundance of 24 bacteria corresponding to the model is required to achieve the measurement of the sample.

[0131] The method for distinguishing the maturity of intestinal flora of infants provided by the present invention obtains the data of representative bacteria of infants to be distinguished; the data of representative bacteria of infants to be distinguished is input into a pre-constructed discrimination model of intestinal flora development maturity of infants to obtain the discrimination result of intestinal flora development status; wherein the discrimination model of intestinal flora development maturity of infants is based on a random forest model, and is trained using a large number of labeled infant intestinal flora data samples, and the training results whose accuracy meets the screening conditions are selected, and the representative bacteria are screened according to the data-model contribution evaluation value for simplification. The present invention is simplified on the basis of the model established by the random forest method, and the obtained discrimination model can predict the development status of intestinal flora by inputting the data of representative bacteria of infants, with low error rate and high accuracy.

[0132] The following is a description of the device for determining the maturity of the infant intestinal flora provided by the present invention. The device for determining the maturity of the infant intestinal flora described below and the method for determining the maturity of the infant intestinal flora described above can be referenced to each other. Figure 3 Schematic diagram of the structure of the device for determining the maturity of intestinal flora in infants and young children provided by the present invention. Figure 3 As shown, the device includes the following modules:

[0133] An input unit 310 is used to obtain representative bacteria data of infants to be identified;

[0134] The discrimination unit 320 is used to input the representative bacteria data of infants to be discriminated into a pre-constructed discrimination model for the development maturity of infant intestinal flora to obtain a discrimination result of the development status of intestinal flora;

[0135] The training unit 330 is used to train based on the random forest model using a large number of labeled infant intestinal flora data samples, select the training results whose accuracy meets the screening conditions, and select representative bacteria according to the data-model contribution evaluation value to simplify and obtain the infant intestinal flora development maturity discrimination model.

[0136] According to a device for distinguishing the maturity of intestinal flora of infants and young children provided by the present invention, based on a random forest model, a large number of labeled intestinal flora data samples of infants and young children are used for training, and training results whose accuracy meets the screening conditions are selected, and representative bacteria are selected according to the data-model contribution evaluation value to simplify the infant intestinal flora maturity distinguishing model, which specifically includes:

[0137] S1: Obtain a large number of labeled infant intestinal flora data samples to obtain a training data set; wherein the labels include species annotations and age labels;

[0138] S2: Divide the training data set into a test set and a validation set according to a preset ratio;

[0139] S3: construct a random seed according to the first number of randomly generated random numbers, and set basic parameters for constructing a random forest model;

[0140] S4: constructing a random forest model according to the basic parameters using the random seed, the test set and the validation set;

[0141] S5: Performing accuracy detection on the first number of first random forest models obtained, and selecting the first random forest model whose accuracy meets the screening condition as the training result;

[0142] S6: For the current training result, the importance of the species annotations in the training process of the training result is ranked using the data-model contribution evaluation value to obtain a ranking result;

[0143] S7: Select the first target number of the species annotations in the sorting result as representative bacteria, screen the training data set according to the species annotations, delete the infant intestinal flora data samples that do not have the species annotations, and update the training data set; execute steps S2-S4 based on the updated training data set to obtain a first number of second random forest models; wherein the target number is a random value within a preset range;

[0144] S8: Update the target number, and repeat steps S6-S8 until a second random forest model of all the training results is obtained;

[0145] S9: Perform a simplified effect evaluation on the second random forest model according to the model prediction error rate, and select the second random forest model with the lowest model prediction error rate as the model for distinguishing the developmental maturity of the infant intestinal flora.

[0146] According to a device for determining the maturity of intestinal flora of infants and young children provided by the present invention, a large number of labeled intestinal flora data samples of infants and young children are obtained, which specifically includes:

[0147] Collect a large number of infant and young children fecal samples according to multiple preset sampling age nodes;

[0148] Extracting the infant intestinal flora sequencing data of the infant feces sample;

[0149] Splicing the infant intestinal flora sequencing data, and extracting characteristic sequences according to the splicing results;

[0150] Comparing the characteristic sequence with a preset database, filtering chimeras, and obtaining an initial infant intestinal flora data sample;

[0151] Aggregating the initial infant intestinal flora data samples based on similarity to obtain a set of initial infant intestinal flora data samples belonging to multiple classification units;

[0152] Annotating the initial infant intestinal flora data sample set corresponding to the classification unit according to the preset database to obtain a process infant intestinal flora data sample with the species annotation label;

[0153] The process infant intestinal flora data samples are merged and grouped according to the preset age groups based on the sampling age nodes, and the process infant intestinal flora data samples corresponding to the age groups are marked with age groups to obtain infant intestinal flora data samples with both the species annotation label and the age label.

[0154] According to a device for determining the maturity of intestinal flora in infants and young children provided by the present invention, the accuracy of the first random forest models obtained is tested, specifically including:

[0155] The ratio of the number of correct predictions to the total number of predictions is calculated, and the accuracy of the first random forest model generated under different random seeds is tested based on the ratio.

[0156] According to a device for determining the maturity of intestinal flora in infants and young children provided by the present invention, the first number of random numbers includes a second number of three-digit random numbers and the remaining four-digit random numbers.

[0157] According to a device for judging the maturity of intestinal flora development in infants and young children provided by the present invention, the sampling age nodes include the seventh day, the fifteenth day, the first month, the second month, the third month, the fourth month, the fifth month, the sixth month, the eighteenth month, the twenty-fourth month, the thirtieth month, the thirty-sixth month, the fourth year, the fifth year and the sixth year.

[0158] The device for distinguishing the maturity of intestinal flora of infants provided by the present invention obtains the data of representative bacteria of infants to be distinguished; inputs the data of representative bacteria of infants to be distinguished into a pre-constructed discrimination model for the maturity of intestinal flora of infants to obtain the discrimination result of the development state of intestinal flora; wherein the discrimination model for the maturity of intestinal flora of infants is based on a random forest model, and is trained using a large number of labeled samples of intestinal flora data of infants, and the training results whose accuracy meets the screening conditions are selected, and the representative bacteria are selected according to the data-model contribution evaluation value for simplification. The present invention is simplified on the basis of the model established by the random forest method, and the obtained discrimination model can predict the development state of intestinal flora by inputting the data of representative bacteria of infants, with low error rate and high accuracy.

[0159] Figure 4An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the method for distinguishing the maturity of the intestinal flora of infants and young children, the method comprising: obtaining representative bacteria data of infants and young children to be distinguished; inputting the representative bacteria data of infants and young children to be distinguished into a pre-constructed infant intestinal flora development maturity distinguishing model to obtain the intestinal flora development state distinguishing result; wherein the infant intestinal flora development maturity distinguishing model is based on a random forest model, and is trained using a large number of labeled infant intestinal flora data samples, and the training results whose accuracy meets the screening conditions are selected and simplified by screening representative bacteria according to the data-model contribution evaluation value.

[0160] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0161] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the method for distinguishing the developmental maturity of the infant intestinal flora provided by the above-mentioned methods, which includes: obtaining representative bacteria data of infants and young children to be distinguished; inputting the representative bacteria data of infants and young children to be distinguished into a pre-constructed infant intestinal flora development maturity distinction model to obtain a distinction result of the intestinal flora development status; wherein the infant intestinal flora development maturity distinction model is based on a random forest model, and is trained using a large number of labeled infant intestinal flora data samples, and the training results whose accuracy meets the screening conditions are selected and simplified by screening representative bacteria according to the data-model contribution evaluation value.

[0162] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the method for discriminating the developmental maturity of the infant intestinal flora provided by the above-mentioned methods, the method comprising: obtaining representative bacteria data of infants and young children to be discriminated; inputting the representative bacteria data of infants and young children to be discriminated into a pre-constructed model for discriminating the developmental maturity of the infant intestinal flora to obtain a result of discriminating the developmental state of the intestinal flora; wherein the model for discriminating the developmental maturity of the infant intestinal flora is based on a random forest model, and is trained using a large number of labeled infant intestinal flora data samples, and the training results whose accuracy meets the screening conditions are selected and simplified by screening representative bacteria according to the data-model contribution evaluation value.

[0163] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0164] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining the maturity of intestinal flora in infants and young children, characterized in that: include: Obtain representative bacteria data of infants and young children to be identified; Inputting the representative bacteria data of infants to be identified into a pre-constructed infant intestinal flora development maturity identification model to obtain an intestinal flora development status identification result; Among them, the infant intestinal flora development maturity discrimination model is based on the random forest model, which is trained using labeled infant intestinal flora data samples, and the training results whose accuracy meets the screening conditions are selected and simplified by screening representative bacteria according to the data-model contribution evaluation value.

2. The method for determining the maturity of intestinal flora in infants and young children according to claim 1, characterized in that: Based on the random forest model, the labeled infant intestinal flora data samples are used for training, and the training results whose accuracy meets the screening conditions are selected. The representative bacteria are selected according to the data-model contribution evaluation value to simplify the infant intestinal flora development maturity discrimination model, which specifically includes: S1: Obtain a large number of labeled infant intestinal flora data samples to obtain a training data set; wherein the labels include species annotations and age labels; S2: Divide the training data set into a test set and a validation set according to a preset ratio; S3: construct a random seed according to the first number of randomly generated random numbers, and set basic parameters for constructing a random forest model; S4: constructing a random forest model according to the basic parameters using the random seed, the test set and the validation set; S5: Performing accuracy detection on the first number of first random forest models obtained, and selecting the first random forest model whose accuracy meets the screening condition as the training result; S6: For the current training result, the importance of the species annotations in the training process of the training result is ranked using the data-model contribution evaluation value to obtain a ranking result; S7: Select the first target number of the species annotations in the sorting result as representative bacteria, screen the training data set according to the species annotations, delete the infant intestinal flora data samples that do not have the species annotations, and update the training data set; execute steps S2-S4 based on the updated training data set to obtain a first number of second random forest models; wherein the target number is a random value within a preset range; S8: Update the target number, and repeat steps S6-S8 until a second random forest model of all the training results is obtained; S9: Perform a simplified effect evaluation on the second random forest model according to the model prediction error rate, and select the second random forest model with the lowest model prediction error rate as the model for distinguishing the developmental maturity of the infant intestinal flora.

3. The method for determining the maturity of intestinal flora in infants and young children according to claim 2, characterized in that: Obtain a large number of labeled infant intestinal flora data samples, including: Collect a large number of infant and young children fecal samples according to multiple preset sampling age nodes; Extracting the infant intestinal flora sequencing data of the infant feces sample; Splicing the infant intestinal flora sequencing data, and extracting characteristic sequences according to the splicing results; Comparing the characteristic sequence with a preset database, filtering chimeras, and obtaining an initial infant intestinal flora data sample; Aggregating the initial infant intestinal flora data samples based on similarity to obtain a set of initial infant intestinal flora data samples belonging to multiple classification units; Annotating the initial infant intestinal flora data sample set corresponding to the classification unit according to the preset database to obtain a process infant intestinal flora data sample with the species annotation label; The process infant intestinal flora data samples are merged and grouped according to the preset age groups based on the sampling age nodes, and the process infant intestinal flora data samples corresponding to the age groups are marked with age groups to obtain infant intestinal flora data samples with both the species annotation label and the age label.

4. The method for determining the maturity of intestinal flora in infants and young children according to claim 2, characterized in that: The first random forest models obtained are tested for accuracy, specifically including: The ratio of the number of correct predictions to the total number of predictions is calculated, and the accuracy of the first random forest model generated under different random seeds is tested based on the ratio.

5. The method for determining the maturity of intestinal flora in infants according to claim 2, characterized in that: The first number of random numbers includes a second number of three-digit random numbers and the remaining four-digit random numbers.

6. The method for determining the maturity of intestinal flora in infants and young children according to claim 2, characterized in that: The sampling age nodes include the seventh day, the fifteenth day, the first month, the second month, the third month, the fourth month, the fifth month, the sixth month, the eighteenth month, the twenty-fourth month, the thirtieth month, the thirty-sixth month, the fourth year, the fifth year and the sixth year.

7. A device for determining the maturity of intestinal flora in infants and young children, characterized in that: include: An input unit, used to obtain representative bacteria data of infants to be identified; A discrimination unit, used for inputting the representative bacteria data of infants to be discriminated into a pre-constructed discrimination model for the developmental maturity of infant intestinal flora to obtain a discrimination result of the developmental state of intestinal flora; The training unit is used to train based on a random forest model using a large number of labeled infant intestinal flora data samples, select training results whose accuracy meets the screening conditions, and select representative bacteria according to the data-model contribution evaluation value to simplify and obtain the infant intestinal flora development maturity discrimination model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for determining the maturity of the intestinal flora of infants and young children as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for determining the maturity of the intestinal flora of infants and young children as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for determining the maturity of the intestinal flora of infants and young children as claimed in any one of claims 1 to 6 is implemented.

Images