A skin aging evaluation method and system based on macro-genome function-species coupling characteristics

Abstract

The present application relates to a kind of skin aging evaluation method and system based on macrogenomic function-species coupling characteristics.The method comprises: obtaining the skin microbial macrogenomic sequencing data of different aging state population, constructs species abundance matrix, functional pathway abundance matrix and function-species hierarchical annotation matrix;Filtering species and functional pathways related to skin aging;Calculate the relative contribution of each species to the corresponding functional pathway, and construct a function-species coupling index combined with the species abundance, input the feature, establish a skin aging evaluation model using machine learning method, and evaluate skin aging.The present application further integrates function source information based on traditional species or function abundance analysis, constructs a function-species coupling feature system, describes the skin microecological aging state from the structural level, significantly improves the accuracy, stability and individualized quantification ability of evaluation, and is suitable for skin health evaluation, aging monitoring and related health management fields.

Description

Technical Field

[0001] This invention belongs to the field of biomedical technology and relates to a method and system for assessing skin aging based on metagenomic function-species coupling characteristics. Background Technology

[0002] Skin aging is a complex physiological and pathological process induced by the long-term combined effects of endogenous genetic factors and exogenous environmental factors (such as ultraviolet radiation, air pollution, and oxidative stress). Clinical signs typically include deepening wrinkles due to collagen degradation in the dermis, sagging caused by elastic fiber degeneration, abnormal melanin deposition, and impaired epidermal barrier function.

[0003] Currently, the main methods for assessing the degree of skin aging include: (1) Clinical visual assessment methods: such as the Glogau photographic grading method or the Griffith scale, which classifies the degree of skin wrinkles, the distribution of pigmentation, and the overall photoaging performance. These methods are simple to operate, but are greatly affected by the subjective experience of the assessor. (2) Non-invasive physical parameter detection methods: physiological indicators are detected by a cutometer and a tewameter, but changes in these indicators often lag behind microscopic changes at the molecular level. (3) Biochemical detection methods: skin tissue is collected for histological staining or cytokine analysis. Although the results are accurate, they are invasive and difficult to be widely used in daily health monitoring.

[0004] Existing technologies mainly focus on aging characteristics at the phenotypic or tissue structure levels, making it difficult to capture the early evolution of skin condition from "quantitative change" to "qualitative change" at the molecular and metabolic network levels.

[0005] In recent years, microbiome analysis has made significant progress in the field of dermatology. In particular, the application of metagenomic sequencing technology has made it possible to analyze the skin microbiota at the genetic and metabolic levels. For example, CN119763674A discloses a method, device, electronic device, and computer-readable storage medium for predicting individual skin aging tendency. The relative abundance of the four genera of microorganisms is input into an individual skin aging tendency prediction model to obtain the age range of the individual to be tested. The individual skin aging tendency prediction model is constructed by training an initial model for predicting individual skin aging tendency using the known physiological age range of the target individual group and the relative abundance of the four genera of microorganisms in the skin swab as training samples.

[0006] However, existing microbiome-based methods still suffer from drawbacks such as ambiguous functional attribution, lack of structural features, and lack of continuous risk expression mechanisms. Therefore, there is an urgent need to develop a more sensitive, predictive, and quantifiable molecular-level assessment system to achieve early identification and dynamic monitoring of skin aging. Summary of the Invention

[0007] To address the shortcomings of existing technologies and practical needs, this invention provides a skin aging assessment method and system based on metagenomic function-species coupling characteristics. By constructing a function-species coupling index, the system integrates functional pathway hierarchical information and species abundance information to characterize the proportion of functional sources of the microbiome and its steady-state changes at the structural level. This achieves a transformation path from high-dimensional metagenomic data to continuous individual aging scores, overcoming the deficiencies of existing technologies in functional attribution, structural modeling, and risk quantification, and improving the accuracy and interpretability of skin aging.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a skin aging assessment method based on metagenomic function-species coupling characteristics, the method comprising the following steps: (1) Construct a feature matrix of function-species coupling index. Based on skin metagenomic data, integrate relative abundance of species, abundance of functional pathways and species hierarchical contribution information of functional pathways to construct coupling index at the level of function-species pairing units. (2) Construct a skin aging assessment model. Based on the functional-species coupling index feature matrix and machine learning algorithm constructed in step (1), systematic feature screening, data partitioning, category weight adjustment and hyperparameter optimization are carried out to establish a supervised learning model. (3) Score the individual’s skin aging status, obtain the skin metagenomic data of the individual to be tested, and calculate the function-species coupling index. Use the skin aging assessment model constructed in step (2) to calculate, standardize the probability values ​​output by the model, and construct a continuous aging scoring system of 0-100 points.

[0009] This invention develops a function-species coupling analysis method based on skin metagenomic data. By constructing a function-species coupling index feature matrix, it structurally integrates species composition information and functional source information to characterize the skin microecological aging characteristics from the perspective of metabolic homeostasis. On this basis, machine learning algorithms are used for feature selection and parameter optimization modeling to construct a skin aging assessment model with strong generalization ability and classification accuracy. Furthermore, the aging probability value output by the model is converted into a continuous scoring system of 0-100 points to achieve an intuitive quantitative expression of the individual's skin aging status.

[0010] Optionally, step (1) specifically includes the following steps: (1-1) Acquisition and preprocessing of skin metagenomic data: Skin microbial samples were collected from subjects of different ages and metagenomic high-throughput sequencing was performed to obtain raw FASTQ format sequencing data and preprocess it. (1-2) Species and functional structure annotation and data format standardization: Based on the data obtained in step (1-1), a two-layer quantitative analysis of the species composition and functional pathways of the skin microbiome was performed. (1-3) Screening of differential features: Based on the data obtained in step (1-2), samples of different age groups are divided into young group and old group. Differential analysis of species and functional pathways between the two groups is performed to screen out species and functional pathways with significant differences in different aging groups and construct differential feature set. (1-4) Calculate the function-species coupling index. Based on the differential features obtained in step (1-3), integrate the functional stratification information and species abundance information obtained in step (1-2) to construct the function-species coupling index, which is used to characterize the microbial metabolic homeostasis structure. (1-5) Construct the function-species coupling index feature matrix. Based on the coupling index values ​​of each function-species pairing unit calculated in step (1-4), construct the function-species coupling index matrix. The matrix is ​​arranged with function-species pairing units as rows and samples as columns. Each matrix element represents the coupling index value of the corresponding function-species pairing unit in the sample.

[0011] Optionally, the preprocessing in step (1-1) includes: performing quality control and low-quality sequence filtering on the raw sequencing data, comparing the quality-controlled sequences with the human reference genome, removing host-derived sequences by screening for unaligned sequences, and obtaining clean metagenomic data containing only microbial origin information.

[0012] Optionally, the bilayer quantitative analysis described in step (1-2) includes: Microbial communities were annotated and their relative abundance was calculated to obtain the relative abundance values ​​of each species in the samples. Information at the genus level and below was screened, and the original complex taxonomic names were standardized to the "genus.species" format to construct a species abundance matrix with unified naming. Clean metagenomic data were functionally annotated to calculate the overall abundance of each functional pathway and the species-level abundance information of the functional pathways. Pathway names were standardized by removing illegal characters and adding a prefix to pathway names that start with numbers. At the same time, a mapping table between the original pathway names and the standardized names was established.

[0013] Optionally, the difference analysis in steps (1-3) includes setting an LDA effect size threshold greater than 2.0 and a statistical significance P-value less than 0.05, and extracting features that meet the threshold conditions.

[0014] Optionally, steps (1-4) specifically include: Based on the differential features obtained in steps (1-3), the composite annotation information of functional pathways and species obtained in steps (1-2) is decomposed into pairing units of functional pathway F and species S. For each function-species pairing unit, the following calculations are performed sequentially: (a) Calculate the pathway contribution The calculation formula is as follows: in, This represents the stratified abundance value of species S for functional pathway F; This represents the total abundance value of pathway F in the sample; It is a smoothing factor; (b) Calculate species abundance weights AbundWeight The calculation formula is as follows: in, This represents the stratified abundance value of species S for functional pathway F; This represents the total abundance of species S in the sample; It is a smoothing factor; (c) Couple pathway contribution with species abundance weights and perform log smoothing to construct a function-species coupling index. SFCI The calculation formula is as follows: .

[0015] Optionally, step (2) specifically includes: (2-1) Data loading and group label encoding: Read the feature matrix of the function-species coupling index constructed in step (1), and extract the function-species coupling feature X and the sample label vector y; encode the young group sample data as 0 and the old group sample data as 1 to construct a binary classification label structure; (2-2) High-dimensional feature screening and dimensionality reduction: perform low variance filtering to remove feature variables that have minimal fluctuations and lack discriminative power among samples; calculate the F-statistic and significance level between each feature and the classification label, sort the features according to the size of the F-statistic value, and screen the core features that rank highly in relation to aging status. (2-3) Data partitioning and category weight adjustment: A random stratified sampling strategy is adopted to divide the core feature data into training set and test set; (2-4) Hyperparameter optimization based on grid search: A binary classification model is constructed using machine learning algorithms, and the hyperparameters of the model are systematically optimized using the grid search method; (2-5) Model training, performance evaluation and key feature identification: After hyperparameter optimization, the machine learning algorithm is trained using the training set to build a skin aging assessment model; the test set data is input into the trained model for independent verification.

[0016] Optionally, the variation threshold set for the low variance filtering in step (2-2) is 0.01.

[0017] Optionally, the machine learning algorithm described in steps (2-4) includes the XGBoost algorithm.

[0018] Optionally, steps (2-4) specifically include: Construct a multidimensional parameter search space that covers key parameters that affect model complexity and regularization strength, including at least one of the following: number of trees, maximum depth, learning rate, sample sampling ratio, or regularization parameter; Subsequently, a 5-fold cross-validation was set up, using the area under the receiver operating characteristic (ROC) curve as the model performance evaluation index. The algorithm traversed all combinations in the parameter network, performed cross-validation under each parameter condition to calculate the average ROC curve area, and selected the parameter combination with the highest ROC curve area as the optimal model parameters.

[0019] Optionally, after constructing the skin aging assessment model, the average gain contribution value of each feature is obtained, all features are sorted by gain value, the top few function-species coupling units with the highest contribution are automatically selected, and their corresponding specific species and related metabolic pathways are analyzed to reveal the core microbial functional nodes driving skin aging at the molecular and metabolic network level.

[0020] Optionally, step (3) specifically includes the following steps: (3-1) Model prediction and aging probability calculation: Input the functional-species coupling index data of the individual to be tested into the model constructed in step (2) for prediction; select the predicted probability that the sample belongs to the elderly group. P old As an indicator of the probability of individual skin aging risk; (3-2) Construct a continuous aging score and standardized mapping, mapping the aging probability value output by the model to a standardized aging score of 0-100 points, the specific formula is as follows: , P old The predicted probability that the sample belongs to the elderly group; An AgingScore close to 0 indicates that the skin microecological structure is closer to the stable structure of youth, while an AgingScore close to 100 indicates that the skin microecological structure is closer to the aging state.

[0021] Secondly, the present invention provides a skin aging assessment system based on metagenomic function-species coupling features, the system comprising a function-species coupling index feature matrix construction module, a skin aging assessment model construction module, and an individual skin aging status scoring module; The function-species coupling index feature matrix construction module is used to perform the following: based on skin metagenomic data, integrate relative abundance of species, abundance of functional pathways, and species hierarchical contribution information of functional pathways to construct a coupling index at the function-species pairing unit level; The skin aging assessment model construction module is used to perform the following processes: based on the functional-species coupling index feature matrix constructed by the functional-species coupling index feature matrix construction module and machine learning algorithms, systematic feature selection, data partitioning, category weight adjustment and hyperparameter optimization are carried out to establish a supervised learning model; The individual skin aging status scoring module is used to perform the following: acquiring the skin metagenomic data of the individual to be tested, calculating the function-species coupling index, using the skin aging assessment model constructed by the skin aging assessment model construction module to perform calculations, standardizing and mapping the probability values ​​output by the model, and constructing a continuous aging scoring system of 0-100 points.

[0022] Optionally, the system is used to perform the steps in the skin aging assessment method based on metagenomic function-species coupling characteristics described in the first aspect.

[0023] Optionally, the function-species coupling index feature matrix construction module is used to perform the following steps: (1-1) Acquisition and preprocessing of skin metagenomic data: Skin microbiome samples were collected from subjects of different ages and metagenomic high-throughput sequencing was performed to obtain raw FASTQ format sequencing data, which was then preprocessed. (1-2) Species and functional structure annotation and data format standardization: Based on the data obtained in step (1-1), a two-layer quantitative analysis of the species composition and functional pathways of the skin microbiome was performed. (1-3) Screening of differential characteristics: Based on the data obtained in step (1-2), samples of different ages were divided into young and old groups, and differential analysis of species and functional pathways between the two groups was performed to screen out species with significant differences in different aging states. (1-4) Calculate the function-species coupling index. Based on the differential features obtained in step (1-3), integrate the functional stratification information and species abundance information obtained in step (1-2) to construct the function-species coupling index, which is used to characterize the microbial metabolic homeostasis structure. (1-5) Construct the function-species coupling index feature matrix. Based on the coupling index values ​​of each function-species pairing unit calculated in step (1-4), construct the function-species coupling index matrix. The matrix is ​​arranged with function-species pairing units as rows and samples as columns. Each matrix element represents the coupling index value of the corresponding function-species pairing unit in the sample.

[0024] Optionally, the skin aging assessment model construction module is used to perform the following steps: (2-1) Data loading and group label encoding, reading the function-species coupling index feature matrix constructed in step (1), and extracting the function-species coupling feature X and sample label vector y; encoding the young group sample data as 0 and the old group sample data as 1 to construct a binary classification label structure; (2-2) High-dimensional feature screening and dimensionality reduction, performing low variance filtering to remove feature variables that have minimal fluctuations between samples and lack discriminative ability; calculating the F statistic and its significance level between each feature and the classification label, and determining the significance level based on the magnitude of the F statistic value. (2-3) Data partitioning and category weight adjustment: a random stratified sampling strategy is adopted to divide the core feature data into training and test sets; (2-4) Hyperparameter optimization based on grid search: a binary classification model is constructed using machine learning algorithms, and the hyperparameters of the model are systematically optimized using the grid search method; (2-5) Model training, performance evaluation and key feature identification: after hyperparameter optimization, the machine learning algorithm is trained using the training set to construct a skin aging assessment model; the test set data is input into the trained model for independent verification.

[0025] Optionally, the skin aging assessment model construction module is used to perform the following steps: (3-1) Model prediction and aging probability calculation, inputting the function-species coupling index data of the individual to be tested into the model constructed in step (2) for prediction; selecting the predicted probability that the sample belongs to the elderly group. P old As an individual skin aging risk probability indicator; (3-2) Construct a continuous aging score and standardized mapping, mapping the aging probability value output by the model to a standardized aging score of 0-100 points, the specific formula is as follows: , P old This represents the predicted probability that the sample belongs to the elderly group.

[0026] Thirdly, the present invention provides an electronic device comprising one or more processors and a memory for storing executable instructions, characterized in that the one or more processors are configured to invoke the executable instructions stored in the memory to implement the function of the skin aging assessment system based on metagenomic function-species coupling characteristics described in the second aspect.

[0027] Fourthly, the present invention provides a computer-readable storage medium having stored thereon computer program instructions, characterized in that, when the computer program instructions are executed by a processor, they implement the functions of the skin aging assessment system based on metagenomic function-species coupling characteristics described in the second aspect.

[0028] Compared with the prior art, the present invention has at least the following beneficial effects: This invention proposes a skin aging assessment scheme based on metagenomic data. From the perspective of "function-species coupling structure", it achieves accurate quantitative assessment of skin aging status by constructing new structured indicators and combining them with machine learning algorithms.

[0029] (1) Propose a function-species coupling index to characterize the functional source structure of the microbial community. This invention proposes for the first time a Species-Function Coupling Index (SFCI) for describing the functional source structure of microorganisms. Based on metagenomic hierarchical functional annotation data, this index integrates functional pathway contribution and species functional abundance weights and performs logarithmic smoothing to construct a composite index that can simultaneously reflect the source of functional execution and functional intensity. The functional pathway contribution measures the relative contribution ratio of a species to a specific functional pathway, while the species functional abundance weight measures the activity of the function in the metabolic network of that species. This index can effectively solve the problem of "opaque functional source" in traditional pathway analysis, enabling metagenomic data to not only reflect changes in functional level but also reveal the key species driving these changes.

[0030] (2) Construct a function-species coupled feature matrix to realize the quantitative expression of microecological structure information. Based on the function-species coupling index, this invention further constructs a coupling feature matrix with "functional pathway-contributing species pairing units" as the basic structural unit. This matrix expands metagenomic data from the traditional two-dimensional structure (species or function) to a function-species structural feature space, thereby enabling a more complete characterization of the metabolic network structure of the skin micro-ecosystem. This structured feature matrix not only reflects the functional composition of the microbiome but also reveals the changing roles of different species in key metabolic pathways, thus providing more biologically meaningful high-dimensional input features for subsequent machine learning models.

[0031] (3) Construct a skin aging assessment model optimized by machine learning To address the challenges of high dimensionality and noise in metagenomic data, this invention establishes a skin aging assessment model through a multi-step feature screening and model optimization strategy. This includes using low-variance filtering and statistical testing methods to screen core features; constructing training and testing sets through random stratified sampling; addressing sample imbalance by incorporating class weighting mechanisms; and optimizing hyperparameters of models such as XGBoost using grid search combined with cross-validation. This method can automatically identify the most discriminative function-species coupling features for skin aging in a complex microbial feature space, thereby improving the model's predictive accuracy and generalization ability.

[0032] (4) Achieve continuous quantitative scoring of skin aging risk This invention further transforms the aging probability predicted by the model into a continuous aging score of 0-100, realizing the quantitative expression of skin aging risk. Compared with the traditional discrete classification method, this continuous scoring system can more meticulously reflect the different stages of an individual's aging process, providing a more intuitive evaluation indicator for skin health assessment, aging monitoring and personalized intervention.

[0033] (5) Revealing the key microbial functional nodes driving skin aging By analyzing the model feature importance (Gain value), this invention can automatically identify the functional-species pairing units that contribute the most to skin aging, thereby revealing the core microbial functional nodes driving skin aging at the molecular and metabolic levels, and providing a theoretical basis for further research on skin microecological regulation and anti-aging. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of a skin aging assessment method based on metagenomic function-species coupling characteristics.

[0035] Figure 2 ROC curve for a skin aging assessment model.

[0036] Figure 3 This image shows a comparison of skin aging scores between the younger and older groups.

[0037] Figure 4 The image shows the results of the top 20 features with the highest importance weight in the skin aging assessment model. Detailed Implementation

[0038] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments. However, the following examples are merely simplified examples of the present invention and do not represent or limit the scope of protection of the present invention. The scope of protection of the present invention is determined by the claims.

[0039] Where specific techniques or conditions are not specified in the examples, they shall be performed in accordance with the techniques or conditions described in the literature in this field, or in accordance with the product instructions. Reagents or instruments whose manufacturers are not specified are all conventional products that can be purchased from legitimate channels.

[0040] This invention proposes a skin aging assessment method based on metagenomic microbiome functional structure characteristics. By constructing a function-species coupling index, it systematically integrates functional pathway hierarchical information and species abundance information to characterize the proportion of functional sources in the microbiome and their steady-state changes at the structural level. The coupling index can not only quantify the relative contribution of a specific species to core functional pathways, but also reflect the activity level of functions within the species' internal metabolic network, thus forming a composite feature system that combines "source structure information" and "functional intensity information." Based on this, machine learning algorithms (such as XGBoost) are introduced for feature selection and parameter optimization modeling to further improve the model's stability, sensitivity, and generalization ability in different population samples.

[0041] Through the above technical solution, the present invention realizes the transformation path from metagenomic high-dimensional data to individual aging continuous scores, makes up for the shortcomings of existing technologies in functional attribution, structural modeling and risk quantification, and improves the accuracy and interpretability of skin aging.

[0042] Specifically, this study focuses on the "species composition-functional execution capability-functional execution homeostasis coupling structure" of the skin microbiome as the core analytical object. It constructs a function-species coupling index feature system and combines it with machine learning algorithms for parameter optimization modeling to achieve continuous quantitative assessment of individual skin aging status. The overall technical solution includes three interconnected technical modules (such as...). Figure 1 (As shown): Function-species coupling feature matrix construction module, skin aging assessment model construction module, and individual skin aging status scoring module.

[0043] (I) Function-Species Coupling Feature Matrix Construction Module The function-species coupling feature matrix construction module aims to construct a coupling feature matrix from skin metagenomic data that simultaneously reflects the species composition structure and functional execution homeostasis. Unlike traditional methods that rely solely on single-dimensional analysis based on species abundance or functional abundance, this module integrates relative species abundance, functional pathway abundance, and species-stratified contribution information of functional pathways to construct a coupling index at the function-species pairing unit level. This characterizes the microecological homeostasis from two dimensions: metabolic execution dominance and functional activity, providing a high-resolution structured feature space for subsequent aging state modeling. The specific steps are as follows: S1: Acquisition and Preprocessing of Skin Metagenomic Data Skin microbiome samples were collected from subjects of different ages and subjected to metagenomic high-throughput sequencing to obtain raw FASTQ format sequencing data. To ensure data quality and the accuracy of subsequent analysis, Fastp software was first used to perform quality control and low-quality sequence filtering on the raw sequencing data, removing low-quality bases and adapter contamination sequences to improve the overall reliability of the data. Subsequently, Bowtie2 software was used to align the quality-controlled sequences with the human reference genome hg38, removing host-derived sequences by screening for unaligned sequences, ultimately obtaining clean metagenomic data containing only microbial origin information. This step effectively reduced the interference of host background noise on the quantification of microbial function, laying the foundation for subsequent species and functional structure analysis.

[0044] S2: Species and Functional Structure Annotation and Data Format Standardization Based on the obtained host-free clean metagenomic data, a two-layer quantitative analysis was performed on the species composition and functional structure of the skin microbiome.

[0045] First, MetaPhlAn4 software was used to perform species classification annotation and relative abundance calculation on the microbial community to obtain the relative abundance values ​​of each species in the sample. The species abundance table generated by MetaPhlAn4 was read, and information at the genus level (g__) and below was filtered. The original complex taxonomic names were standardized to the "genus.species" format (e.g., g__Cutibacterium.s__Cutibacterium_acnes), thus constructing a uniformly named species abundance matrix.

[0046] Secondly, the HUMAnN3 tool was used to perform functional annotation on the clean metagenomic data, calculating the overall abundance (in CPM) of each functional pathway and the species-stratified abundance information of the functional pathways. After reading the functional pathway abundance table generated by HUMAnN3, the pathway names were normalized by removing illegal characters (such as colons, commas, spaces, etc.) and adding a prefix (such as f_) to pathway names starting with numbers to meet the naming requirements of subsequent machine learning algorithms. Simultaneously, a mapping table between the original pathway names and the normalized names was established to ensure that subsequent results could be traced back to the original functional definitions. Through this step, standardized species abundance matrices, functional pathway abundance matrices, and functional stratification structure data were obtained.

[0047] S3: Differential Feature Filtering To improve the discriminative power of the feature space and reduce the impact of redundant variables on model stability, samples from different age groups were divided into young and old groups. The LEfSe tool was used to analyze the differences in species and functional pathways between the two groups. By setting an LDA effect size threshold greater than 2.0 and a statistical significance p-value less than 0.05, species and functional pathways with significant differences in different aging states were screened. After reading the LEfSe analysis results, features meeting the threshold conditions were extracted and a differential feature set was formed. This step not only compressed the original high-dimensional feature space but also enhanced the biological relevance of the subsequent coupling index construction, improving the model's interpretability.

[0048] S4: Calculation of Function-Species Coupling Index (SFCI) By integrating functional stratification information with species abundance information, a function-species coupling index (SFCI) is constructed to characterize the structure of microbial metabolic homeostasis.

[0049] First, based on the functional species stratified table output by HUMANN3, the pandas package was used in a Python 3.8 environment to parse the stratified data, decomposing the composite annotation information of "functional pathway | contributing species" into paired units of functional pathway F and species S. For each function-species paired unit, the following calculations were performed sequentially: (1) Calculate the pathway contribution ( Contribution This indicator measures the relative contribution of a specific species to the overall functioning of a particular pathway. The calculation formula is as follows: in, This represents the stratified abundance value of species S for functional pathway F; This represents the total abundance value of pathway F in the sample; As a smoothing factor, it is used to avoid zero denominators and reduce the influence of low abundance noise. This indicator reflects the degree to which a species dominates the performance of a specific function.

[0050] (2) Calculate the species abundance weights ( AbundWeight This index measures the activity of a specific function in the metabolic network of a species, and its calculation formula is as follows: in, This represents the stratified abundance value of species S for functional pathway F; Represents the total abundance of species S in the sample (based on MetaPhlAn4 calculations); As a smoothing factor, it is used to avoid zero denominator and reduce the influence of low abundance noise. This indicator reflects the proportion of function in the metabolic activities within a species.

[0051] (3) The pathway contribution is coupled with the species abundance weight and log smoothed to construct the function-species coupling index (SFCI), which is calculated as follows: Through the above calculations, SFCI integrates the species’ dominance in functional execution and the activity of function in the species’ metabolic network, characterizing the strength of microecological metabolic homeostasis from a dual structural dimension, and can more sensitively reflect the functional imbalance of the skin microecological structure during the aging process.

[0052] S5: Constructing a Functional-Species Coupling Index Matrix Based on the calculated SFCI values ​​of each function-species pairing unit, a function-species coupling index matrix is ​​constructed. This matrix is ​​structured with "function-species pairing units" as rows and samples as columns, where each matrix element represents the SFCI value of that function-species pairing unit in the corresponding sample. This matrix serves as the core input feature space for subsequent machine learning model training and predictive analysis, providing a data foundation for high-dimensional structured modeling of skin aging states.

[0053] (II) Skin Aging Assessment Model Construction Module The skin aging assessment model construction module aims to build a skin aging assessment model with high generalization ability and high classification accuracy based on the aforementioned high-dimensional function-species coupling index (SFCI) feature matrix. Through a systematic process of feature selection, data partitioning, class weight adjustment, and hyperparameter optimization, a stable, repeatable, and well-extrapolated supervised learning model is established to achieve accurate identification and quantitative assessment of skin aging status. Simultaneously, through model feature importance analysis, key microbial functional nodes driving skin aging are back-analyzed to execute the following steps: S1: Data Loading and Group Tag Encoding In the Python 3.8 environment, the Pandas library is used to read the constructed Function-Species Coupling Index (SFCI) matrix file and extract the feature variable matrix X (Function-Species Coupling Features) and the sample label vector y (Aging State Grouping Information).

[0054] To meet the numerical input requirements of supervised learning algorithms, the LabelEncoder function in the Scikit-learn library is called to re-encode the sample groups, encoding the young group as 0 and the old group as 1, thus constructing a standard binary classification label structure.

[0055] This step completes the conversion from biological group labels to machine learning-recognizable numerical labels, establishing a standardized data input structure for subsequent modeling processes.

[0056] S2: High-dimensional feature selection and dimensionality reduction Since the SFCI matrix belongs to a high-dimensional feature space, directly using it for model training may lead to the curse of dimensionality and overfitting risks. Therefore, this step performs systematic screening and dimensionality reduction on the features.

[0057] First, the `VarianceThreshold` function from the Scikit-learn library is called to perform low-variance filtering, setting the variance threshold to 0.01. This automatically removes feature variables that have minimal fluctuations among samples and lack discriminative power. This process effectively removes noisy variables and improves the feature signal-to-noise ratio.

[0058] Secondly, the SelectKBest method combined with the one-way ANOVA algorithm f_classif is used to calculate the F-statistic and its significance level between each feature and the classification label. Features are then ranked according to their F-values, and the top 100 core features most strongly correlated with aging status are selected. This method significantly reduces model complexity and improves computational efficiency and model stability while preserving key discriminative information.

[0059] Through the above dual screening strategy, the high-dimensional redundant space is compressed and transformed into an information-intensive core feature space.

[0060] S3: Data Partitioning and Category Weight Adjustment To objectively evaluate the model's generalization ability, a random stratified sampling strategy was adopted. The core feature data was divided into training and test sets in a 7:3 ratio using a function in Scikit-learn. This ensured that the proportion of young and aging samples in the training and test sets remained consistent, thereby avoiding the impact of class bias on model evaluation.

[0061] Given the potential for sample size imbalance between groups during biological specimen collection, the program automatically calculates the class weight coefficient (scale_pos_weight), which dynamically adjusts the loss function weights of the model during training by calculating the ratio of the total number of young samples to the total number of old samples in the training set, thereby improving the model's recognition ability.

[0062] S4: Hyperparameter Optimization Based on Grid Search This step uses the XGBoost algorithm to construct a binary classification model and then uses a grid search method to systematically optimize the model's hyperparameters to obtain the best generalization performance.

[0063] First, a multi-dimensional parameter search space is constructed, covering key parameters that affect model complexity and regularization strength, including but not limited to the number of trees (n_estimators), maximum depth (max_depth), learning rate (learning_rate), sample sampling ratio (subsample), and regularization parameters (reg_alpha, reg_lambda).

[0064] The algorithm then calls the GridSearchCV function from the Scikit-learn library to set up 5-Fold Cross-Validation, using the Area Under the Receiver Operating Characteristic (AUC) as the model performance evaluation metric. The algorithm iterates through all combinations in the parameter network, performs cross-validation under each parameter condition to calculate the average AUC, and automatically selects the parameter combination with the highest AUC as the optimal model parameters.

[0065] This optimization process ensures that the model has strong generalization ability during the training phase, avoiding subjective bias caused by reliance on experience-based parameter tuning.

[0066] S5: Model Training, Performance Evaluation, and Key Feature Recognition After obtaining the optimal hyperparameters, the XGBoost model is trained using the complete training set to construct a skin aging assessment model. After training, the average gain contribution value of each feature is obtained by calling the Booter interface in the XGBoost model. This value directly quantifies the importance of each function-species coupling feature in distinguishing aging states.

[0067] The program sorts all features by Gain value, automatically selects the top 20 function-species coupling units with the highest contribution, and analyzes their corresponding specific species and related metabolic pathways, revealing the core microbial functional nodes driving skin aging at the molecular and metabolic network level.

[0068] The test set data is then input into the trained model for independent validation. Predicted probability values ​​are calculated, and performance metrics, including AUC, accuracy, sensitivity, and specificity, are also calculated. Validation with the independent test set ensures the model has stable predictive capabilities and practical applicability.

[0069] (III) Individual Skin Aging Status Scoring Module The individual skin aging status scoring module applies a trained and parameter-optimized skin aging assessment model to individual samples, enabling a continuous and quantifiable expression of skin aging risk. Unlike the traditional binary "young / aged" discrete judgment method, this module standardizes and maps the probability values ​​output by the model to construct a continuous aging scoring system of 0-100 points. This allows the individual's skin microecological aging status to be presented in an intuitive numerical form, facilitating clinical interpretation, health management, and longitudinal dynamic monitoring.

[0070] S1: Model Prediction and Aging Probability Calculation After obtaining the optimal classifier model after training, the function-species coupling index (SFCI) feature data of the individual to be evaluated is input into the trained XGBoost model for prediction.

[0071] This module selects the predicted probability that the sample belongs to the elderly group. P old As an indicator of the probability of individual skin aging risk, this probability is essentially a confidence estimate of whether a sample falls into the "aging state" category based on the function-species coupled feature space, and its value ranges from 0 to 1.

[0072] By using a probability-based output method, the information loss caused by single-category labels can be avoided, enabling a continuous expression of an individual's aging state.

[0073] S2: Construction and Standardized Mapping of Continuous Aging Scores To improve the interpretability of the results, the aging probability values ​​output by the model are mapped to a standardized aging score (AgingScore) of 0-100. This score is constructed using a linear mapping method, transforming the probability space [0,1] into the score interval [0,100]. The specific formula is as follows: After conversion, an AgingScore close to 0 indicates that the skin microecological structure is closer to the youthful homeostatic structure, while an AgingScore close to 100 indicates that the skin microecological structure is closer to the aging state.

[0074] Example 1 To verify the applicability and discriminative ability of the skin aging assessment scheme proposed in this invention in a real population, skin microbial samples were collected from 81 healthy subjects, including 43 subjects in the young group (19-35 years old) and 38 subjects in the elderly group (55-75 years old). All subjects had no history of serious skin diseases and had not used antibiotics or topical antibacterial agents within one week prior to sampling.

[0075] Under uniform collection conditions, skin microbiota samples were taken from the cheek area near the earlobe of all subjects. This area is a relatively stable facial region in terms of sebum secretion and can well reflect the skin microecological status. After DNA extraction, the collected skin microbial samples underwent metagenomic sequencing to obtain the microbial community genome data for each sample.

[0076] Subsequently, the sequencing data were analyzed and processed according to the method proposed in this invention (the aforementioned function-species coupling feature matrix construction module, skin aging assessment model construction module, and individual skin aging status scoring module). First, the raw sequencing data underwent quality control and host-derived sequences were removed. Then, MetaPhlAn4 was used to calculate the relative abundance of species in the skin microbiome, and HUMAnN3 was used to calculate the abundance of microbial functional pathways and the function-species stratification contribution information. Based on this, function-species pairing units were constructed by parsing the functional stratification annotation data, and the function-species coupling index (SFCI) of each pairing unit was calculated according to the technical formula proposed in this invention, forming a high-dimensional functional structure feature matrix.

[0077] The constructed SFCI feature matrix is ​​input into the skin aging assessment model of this invention, and the corresponding skin aging score (AgingScore) is calculated for each subject. The ROC curve of the model is plotted as follows. Figure 2 As shown, the AUC value was 0.93, indicating that the model has strong recognition ability. Further statistical analysis showed that the mean skin aging score of the younger group was 32.6±10.4, while that of the older group was 71.8±12.7; the difference between the two groups was statistically significant (P<0.001). Figure 3 As shown, the method of the present invention can effectively distinguish the skin microecological aging status of people in different age groups.

[0078] Further analysis of the model's important features revealed that among the top-ranked function-species coupling features, multiple metabolic pathways were primarily contributed by skin symbiotic microorganisms, such as... Figure 4 As shown, for example, *Propionibacterium acnes* makes significant contributions to multiple metabolic pathways, including the L-lysine biosynthesis pathway (L-lysine biosynthesis VI, PWY-5097), the superpathway of purine nucleotide salvage (PWY-6609), the inosine-5′-phosphate biosynthesis pathway (PWY-6123), and the L-valine biosynthesis pathway (Valsyn-PWY).

[0079] In addition, some oral and skin-associated streptococci also contribute to the function of specific metabolic pathways. For example, *Streptococcus gordonii* is involved in the lactose and galactose degradation pathway (LACTOSECAT-PWY) and the gondoate biosynthesis pathway (PWY-7663); *Streptococcus vestibularis* is involved in the CDP-diacylglycerol biosynthesis pathway (PWY0-1319).

[0080] The aforementioned function-species coupling characteristics exhibited significantly different contribution structures between the young and old groups. Overall, the contribution proportions of amino acid synthesis, nucleotide metabolism, and lipid metabolism pathways dominated by dermatophytes such as Cutibacterium acnes were more stable in the young group samples, while in the old group, the contribution structures of these functional pathways were restructured, with some functions gradually being replaced by other microbial species.

[0081] These results indicate that skin aging is not only manifested in changes in microbial species abundance, but also in alterations in the structure of the sources of microbial function. The function-species coupling index (SFCI) proposed in this invention can effectively capture this structural change, thereby improving the accuracy and biological interpretability of skin aging assessment.

[0082] Therefore, the method of the present invention can be widely applied to fields such as skin aging research, skin health assessment, skin care product efficacy evaluation, and personalized skin health management.

[0083] Example 2 This embodiment further applies the skin aging assessment scheme.

[0084] Subject A, a 32-year-old female with no prior history of serious skin diseases, experienced an acute flare-up of atopic dermatitis during a period of high work stress and irregular sleep patterns. Clinical manifestations included significant itching, erythema, dryness, and desquamation of the cheeks, accompanied by worsening itching at night. A dermatologist diagnosed her and prescribed topical treatment with a low-potency corticosteroid ointment, and recommended improvements to her lifestyle and skincare routine.

[0085] Researchers collected skin surface microbial samples during the acute phase (T1) and the recovery phase (T2) of the disease. T1 sampling occurred when dermatitis symptoms were pronounced, while T2 sampling occurred after three weeks of continuous treatment when skin symptoms had largely subsided. All samples underwent metagenomic sequencing analysis according to the methods described in this invention.

[0086] First, the collected samples underwent sequencing data quality control and host removal processing. Then, MetaPhlAn4 was used to calculate the abundance of microbial species, and HUMAnN3 was used to calculate the abundance of functional pathways and function-species stratification information. Furthermore, the function-species coupling index (SFCI) matrix was constructed according to the present invention and input into the skin aging assessment model of the present invention for prediction, and the corresponding skin aging score (AgingScore) was calculated.

[0087] The analysis results showed that during the acute phase of the disease (T1), subject A's skin aging score was 74.21, and the model predicted that her skin microecological structure was significantly biased towards an aging-related pattern; during the treatment recovery phase (T2), her skin aging score decreased to 41.53, and the surface skin microecological structure tended to recover to a healthy homeostasis.

[0088] Further analysis of function-species coupling characteristics revealed that during the acute phase of the disease, the contribution structure of some microbial functional pathways related to inflammatory response, oxidative stress, and lipid metabolism disorders underwent significant changes, with some opportunistic pathogens increasing their contribution to related metabolic pathways. In the treatment and recovery phase, these functional structures gradually returned to the stable patterns commonly seen in healthy individuals.

[0089] The above results indicate that the skin aging assessment method based on metagenomic function-species coupling index proposed in this invention can not only identify the skin aging status of different age groups, but also sensitively capture the fluctuations in the microecological structure caused by changes in physiological states such as skin inflammation or barrier damage, thereby realizing dynamic monitoring and quantitative assessment of individual skin health status.

[0090] In summary, this invention constructs a Function-Species Coupling Index (SFCI) feature matrix to structurally integrate species composition information with functional origin information, thus characterizing the skin microecological aging features from the perspective of metabolic homeostasis. Based on this, machine learning algorithms are used for feature selection and parameter optimization modeling to construct a skin aging assessment model with strong generalization ability and classification accuracy. Furthermore, the aging probability values ​​output by the model are transformed into a continuous scoring system of 0-100 points, achieving an intuitive quantitative expression of individual skin aging status and providing technical support for personalized skincare intervention and research on aging mechanisms.

[0091] The applicant declares that the above description is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention fall within the protection and disclosure scope of the present invention.

Claims

1. A method for assessing skin aging based on metagenomic function-species coupling characteristics, characterized in that, The method includes the following steps: (1) Construct a feature matrix of function-species coupling index. Based on skin metagenomic data, integrate relative abundance of species, abundance of functional pathways and species hierarchical contribution information of functional pathways to construct coupling index at the level of function-species pairing units. (2) Construct a skin aging assessment model. Based on the functional-species coupling index feature matrix and machine learning algorithm constructed in step (1), systematic feature screening, data partitioning, category weight adjustment and hyperparameter optimization are carried out to establish a supervised learning model. (3) Score the individual’s skin aging status, obtain the skin metagenomic data of the individual to be tested, and calculate the function-species coupling index. Use the skin aging assessment model constructed in step (2) to calculate, standardize the probability values ​​output by the model, and construct a continuous aging scoring system of 0-100 points.

2. The skin aging assessment method based on metagenomic function-species coupling characteristics according to claim 1, characterized in that, Step (1) specifically includes the following steps: (1-1) Acquisition and preprocessing of skin metagenomic data: Skin microbial samples were collected from subjects of different ages and metagenomic high-throughput sequencing was performed to obtain raw FASTQ format sequencing data and preprocess it. (1-2) Species and functional structure annotation and data format standardization: Based on the data obtained in step (1-1), a two-layer quantitative analysis of the species composition and functional pathways of the skin microbiome was performed. (1-3) Screening of differential features: Based on the data obtained in step (1-2), samples of different age groups are divided into young group and old group. Differential analysis of species and functional pathways between the two groups is performed to screen out species and functional pathways with significant differences in different aging groups and construct differential feature set. (1-4) Calculate the function-species coupling index. Based on the differential features obtained in step (1-3), integrate the functional stratification information and species abundance information obtained in step (1-2) to construct the function-species coupling index, which is used to characterize the microbial metabolic homeostasis structure. (1-5) Construct the function-species coupling index feature matrix. Based on the coupling index values ​​of each function-species pairing unit calculated in step (1-4), construct the function-species coupling index matrix. The matrix is ​​arranged with function-species pairing units as rows and samples as columns. Each matrix element represents the coupling index value of the corresponding function-species pairing unit in the sample.

3. The skin aging assessment method based on metagenomic function-species coupling characteristics according to claim 2, characterized in that, The preprocessing described in step (1-1) includes: performing quality control and low-quality sequence filtering on the raw sequencing data, comparing the quality-controlled sequences with the human reference genome, removing host-derived sequences by screening for unaligned sequences, and obtaining clean metagenomic data containing only microbial origin information; Optionally, the bilayer quantitative analysis described in step (1-2) includes: Microbial communities were annotated and their relative abundance was calculated to obtain the relative abundance values ​​of each species in the samples. Information at the genus level and below was screened, and the original complex taxonomic names were standardized to the "genus.species" format to construct a species abundance matrix with unified naming. Clean metagenomic data were functionally annotated to calculate the overall abundance of each functional pathway and the species-stratified abundance information of the functional pathways. Pathway names were standardized by removing illegal characters and adding a prefix to pathway names that start with numbers. At the same time, a mapping table between the original pathway names and the standardized names was established. Optionally, the difference analysis in steps (1-3) includes setting an LDA effect size threshold greater than 2.0 and a statistical significance P-value less than 0.05, and extracting features that meet the threshold conditions; Optionally, steps (1-4) specifically include: Based on the differential features obtained in steps (1-3), the composite annotation information of functional pathways and species obtained in step (1-2) is decomposed into pairing units of functional pathway F and species S. For each function-species pairing unit, the following calculations are performed sequentially: (a) Calculate the pathway contribution The calculation formula is as follows: in, This represents the stratified abundance value of species S for functional pathway F; This represents the total abundance value of pathway F in the sample; It is a smoothing factor; (b) Calculate species abundance weights AbundWeight The calculation formula is as follows: in, This represents the stratified abundance value of species S for functional pathway F; This represents the total abundance of species S in the sample; It is a smoothing factor; (c) The pathway contribution is coupled with the species abundance weight and log smoothed to construct the function-species coupling index (SFCI), which is calculated as follows: 。 4. The skin aging assessment method based on metagenomic function-species coupling characteristics according to any one of claims 1-3, characterized in that, Step (2) specifically includes: (2-1) Data loading and group label encoding: Read the feature matrix of the function-species coupling index constructed in step (1), and extract the function-species coupling feature X and the sample label vector y; encode the young group sample data as 0 and the old group sample data as 1 to construct a binary classification label structure; (2-2) High-dimensional feature screening and dimensionality reduction: perform low variance filtering to remove feature variables that have minimal fluctuations and lack discriminative power among samples; calculate the F-statistic and significance level between each feature and the classification label, sort the features according to the size of the F-statistic value, and screen the core features that rank highly in relation to aging status. (2-3) Data partitioning and category weight adjustment: A random stratified sampling strategy is adopted to divide the core feature data into training set and test set; (2-4) Hyperparameter optimization based on grid search: A binary classification model is constructed using machine learning algorithms, and the hyperparameters of the model are systematically optimized using the grid search method; (2-5) Model training, performance evaluation and key feature identification: After hyperparameter optimization, the machine learning algorithm is trained using the training set to build a skin aging assessment model; the test set data is input into the trained model for independent verification.

5. The skin aging assessment method based on metagenomic function-species coupling characteristics according to claim 4, characterized in that, The variation threshold set for the low variance filtering in step (2-2) is 0.01; Optionally, the machine learning algorithm described in steps (2-4) includes the XGBoost algorithm; Optionally, steps (2-4) specifically include: Construct a multidimensional parameter search space that covers key parameters that affect model complexity and regularization strength, including at least one of the following: number of trees, maximum depth, learning rate, sample sampling ratio, or regularization parameter; Subsequently, a 5-fold cross-validation was set up, using the area under the receiver operating characteristic (ROC) curve as the model performance evaluation index. The algorithm traversed all combinations in the parameter network, performed cross-validation under each parameter condition to calculate the average ROC curve area, and selected the parameter combination with the highest ROC curve area as the optimal model parameters.

6. The skin aging assessment method based on metagenomic function-species coupling characteristics according to any one of claims 1-5, characterized in that, Step (3) specifically includes the following steps: (3-1) Model prediction and aging probability calculation: Input the functional-species coupling index data of the individual to be tested into the model constructed in step (2) for prediction; select the predicted probability that the sample belongs to the elderly group. P old As an indicator of the probability of individual skin aging risk; (3-2) Construct a continuous aging score and standardized mapping, mapping the aging probability value output by the model to a standardized aging score of 0-100 points, the specific formula is as follows: , P old The predicted probability that the sample belongs to the elderly group; An AgingScore close to 0 indicates that the skin microecological structure is closer to the stable structure of youth, while an AgingScore close to 100 indicates that the skin microecological structure is closer to the aging state.

7. A skin aging assessment system based on metagenomic function-species coupling characteristics, characterized in that, The system includes a function-species coupling index feature matrix construction module, a skin aging assessment model construction module, and an individual skin aging status scoring module; The function-species coupling index feature matrix construction module is used to perform the following: based on skin metagenomic data, integrate relative abundance of species, abundance of functional pathways, and species hierarchical contribution information of functional pathways to construct a coupling index at the function-species pairing unit level; The skin aging assessment model construction module is used to perform the following processes: based on the functional-species coupling index feature matrix constructed by the functional-species coupling index feature matrix construction module and machine learning algorithms, systematic feature selection, data partitioning, category weight adjustment and hyperparameter optimization are carried out to establish a supervised learning model; The individual skin aging status scoring module is used to perform the following: acquiring the skin metagenomic data of the individual to be tested, calculating the function-species coupling index, using the skin aging assessment model constructed by the skin aging assessment model construction module to perform calculations, standardizing and mapping the probability values ​​output by the model, and constructing a continuous aging scoring system of 0-100 points.

8. The skin aging assessment system based on metagenomic function-species coupling characteristics according to claim 7, characterized in that, The functional-species coupling index feature matrix construction module is used to perform the following steps: (1-1) Acquisition and preprocessing of skin metagenomic data: Skin microbiome samples were collected from subjects of different ages and metagenomic high-throughput sequencing was performed to obtain raw FASTQ format sequencing data, which was then preprocessed. (1-2) Species and functional structure annotation and data format standardization: Based on the data obtained in step (1-1), a two-layer quantitative analysis of the species composition and functional pathways of the skin microbiome was performed. (1-3) Screening of differential characteristics: Based on the data obtained in step (1-2), samples of different ages were divided into young and old groups, and differential analysis of species and functional pathways between the two groups was performed to screen out species with significant differences in different aging states. (1-4) Calculate the function-species coupling index. Based on the differential features obtained in step (1-3), integrate the functional stratification information and species abundance information obtained in step (1-2) to construct the function-species coupling index, which is used to characterize the microbial metabolic homeostasis structure. (1-5) Construct the function-species coupling index feature matrix. Based on the coupling index values ​​of each function-species pairing unit calculated in step (1-4), construct the function-species coupling index matrix. The matrix is ​​arranged with function-species pairing units as rows and samples as columns. Each matrix element represents the coupling index value of the corresponding function-species pairing unit in the sample. Optionally, the skin aging assessment model construction module is used to perform the following steps: (2-1) Data loading and group label encoding, reading the function-species coupling index feature matrix constructed in step (1), and extracting the function-species coupling feature X and sample label vector y; encoding the young group sample data as 0 and the old group sample data as 1 to construct a binary classification label structure; (2-2) High-dimensional feature screening and dimensionality reduction, performing low variance filtering to remove feature variables that have minimal fluctuations between samples and lack discriminative ability; calculating the F statistic and its significance level between each feature and the classification label, and determining the significance level based on the magnitude of the F statistic value. (2-3) Data partitioning and category weight adjustment: A random stratified sampling strategy was adopted to divide the core feature data into training and test sets; (2-4) Hyperparameter optimization based on grid search: A binary classification model was constructed using a machine learning algorithm, and the hyperparameters of the model were systematically optimized using a grid search method; (2-5) Model training, performance evaluation, and key feature identification: After hyperparameter optimization, the machine learning algorithm was trained using the training set to construct a skin aging assessment model; the test set data was input into the trained model for independent verification; Optionally, the skin aging assessment model construction module is used to perform the following steps: (3-1) Model prediction and aging probability calculation, inputting the function-species coupling index data of the individual to be tested into the model constructed in step (2) for prediction; selecting the predicted probability that the sample belongs to the elderly group. P old As an individual skin aging risk probability indicator; (3-2) Construct a continuous aging score and standardized mapping, mapping the aging probability value output by the model to a standardized aging score of 0-100. AgingScore The specific formula is as follows: , P old This represents the predicted probability that the sample belongs to the elderly group.

9. An electronic device comprising one or more processors and a memory for storing executable instructions, characterized in that, The one or more processors are configured to invoke executable instructions stored in the memory to implement the functions of the skin aging assessment system based on metagenomic function-species coupling features as described in claim 7 or 8.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the functions of the skin aging assessment system based on metagenomic function-species coupling characteristics as described in claim 7 or 8.

Images