Early warning system for late-onset sepsis in preterm infants based on maturity slope of microbiota and construction method thereof

Abstract

The application provides a preterm infant late-onset sepsis early warning system based on a flora maturity slope and a construction method thereof. The application can capture the signal that the flora maturity process is blocked before the clinical symptoms of late-onset sepsis appear by dynamically monitoring the change trend of the intestinal flora maturity slope K value of the preterm infant, the early warning time is earlier than that of traditional blood culture, sufficient time is provided for early clinical intervention, and the incidence rate and mortality rate of late-onset sepsis are effectively reduced.

Description

Technical Field

[0001] This application relates to biomedical informatics, microbiome, and clinical auxiliary diagnosis, specifically to an early warning system for late-onset sepsis in premature infants based on the slope of microbial maturity and its construction method. Background Technology

[0002] Late-onset sepsis (LOS) in premature infants refers to sepsis occurring within 72 hours of birth or later. It is one of the leading causes of morbidity, mortality, and long-term neurodevelopmental damage in premature infants in the neonatal intensive care unit (NICU). Because premature infants have immature immune systems, weak mucosal barriers, and often require invasive procedures (such as endotracheal intubation and intravenous catheterization) and broad-spectrum antibiotic treatment, their incidence of LOS is significantly higher than that of full-term newborns, seriously threatening their lives and health.

[0003] Currently, the clinical diagnosis of late-onset sepsis in premature infants mainly relies on traditional indicators such as blood culture, complete blood count, C-reactive protein (CRP), and procalcitonin (PCT). However, blood culture has drawbacks such as a long testing cycle (usually 24-48 hours) and a high false negative rate, making it difficult to provide early warning and delaying the best treatment time.

[0004] Recent studies on the link between gut microbiota and the health of premature infants have shown that the development and maturation of the gut microbiota in premature infants is closely related to the body's immune function and metabolic status, and gut microbiota dysbiosis is one of the important inducing factors of late-onset sepsis. As the largest micro-ecological system in the body, the composition and structure of the gut microbiota exhibit a regular succession characteristic with the increasing age of premature infants; that is, the maturity of the microbiota gradually increases with the increase of corrected age. When premature infants develop late-onset sepsis, the normal maturation process of the gut microbiota is significantly inhibited, resulting in stagnant or even regressed microbiota maturation.

[0005] In existing technologies, some studies have attempted to assess infection risk by analyzing the composition of the gut microbiota in premature infants, but most of these studies are limited to microbiota abundance analysis at a single time point and fail to dynamically capture the changing trends in microbiota maturity, thus failing to provide early warning of the occurrence of late-onset sepsis. Summary of the Invention

[0006] The purpose of this application is to provide an early warning system and construction method for late-onset sepsis in premature infants based on the slope of gut microbiota maturity. By dynamically monitoring the changing trend of gut microbiota maturity in premature infants, the system can achieve early and accurate warning of late-onset sepsis, providing timely and reliable reference for clinical intervention and reducing the incidence and mortality of late-onset sepsis in premature infants.

[0007] To achieve the above objectives, this application provides the following technical solution: As a first aspect, this application relates to a method for constructing an early warning system for late-onset sepsis in preterm infants based on the slope of gut microbiota maturity, which includes the following steps: S101. Data acquisition and benchmark establishment to generate classification abundance files; S102, Screening for core characteristic bacterial groups; S103, Training the microbial community maturity model; S104. The rate of change of microbial community maturity with actual correction age is defined as maturity slope. S105. Determine the warning threshold and perform a logical judgment between the maturity slope and the warning threshold, so as to output a warning signal when the maturity slope is less than the warning threshold.

[0008] Further setup: In step S101, longitudinal fecal samples are collected from the preterm infant cohort, and raw data on the microbial community composition are obtained using 16S rRNA high-throughput sequencing technology.

[0009] Further steps include: quality control and deduplication of the raw data on the obtained microbial community composition, and comparison of the microbial dataset to generate a classification abundance file.

[0010] Further configuration: In S102, the contribution of each bacterial genus to age prediction is evaluated by cross-validation using a random forest regression algorithm, and clock bacteria within the target ranking range are selected. The clock bacteria are used as the core feature bacteria, and noise bacteria that colonize instantaneously are removed.

[0011] Further configuration: In S103, the selected clock microbial community is used to establish a nonlinear regression model, which is used to output the microbial community maturity corresponding to the microbial community abundance of a specific individual.

[0012] Further setting: In S104, the maturity slope is the degree of change in the maturity of the microbial community with the actual correction age.

[0013] Further setting: In S105, a warning threshold is determined by retrospectively analyzing data of children with sepsis. When the maturity slope is less than the warning threshold, it is determined that the premature infant has a high risk of late-onset sepsis.

[0014] As a second aspect, this application relates to an early warning system for late-onset sepsis in preterm infants based on the slope of microbial community maturity, constructed using the above-described construction method, which includes a sample collection and processing module, a data input module, a kinetic processing module, and an early warning decision module. The sample collection and processing module is used to generate microbial classification and abundance files; The data input module is used to receive the microbial classification and abundance file generated by the sample collection and processing module, and transmit the relevant microbial classification and abundance file to the kinetic processing module; The dynamic processing module is equipped with a maturity regression engine and a slope calculation logic unit. The maturity regression engine is used to calculate the microbial maturity MA value of preterm infants in real time. The slope calculation logic unit is used to perform time series difference operation on the microbial maturity MA value to calculate the maturity slope K value and generate the microbial maturity slope K value curve. The early warning decision module includes a logic comparator, which is used to compare and judge the maturity slope K value generated in real time by the dynamics processing module with the early warning threshold stored in the database.

[0015] Further features include a visual terminal interface, which is electrically connected to the early warning decision module. This interface provides a succession diagram of the MA / CA curve for each premature infant and issues an early warning alert when the K-value curve deviates from its normal trajectory to the early warning threshold.

[0016] Further setting: When the maturity slope continues to decrease, the risk of late-onset sepsis is determined in premature infants.

[0017] Compared with existing technologies, the solution in this application has the following advantages: 1. The early warning system for late-onset sepsis in premature infants based on the slope of gut microbiota maturation proposed in this application can capture signals of obstructed microbiota maturation process before the clinical symptoms of late-onset sepsis appear by dynamically monitoring the changing trend of the slope K value of gut microbiota maturation in premature infants. The early warning time is earlier than that of traditional blood culture, providing sufficient time for early clinical intervention and effectively reducing the incidence and mortality of severe late-onset sepsis.

[0018] 2. In the construction method of the early warning system for late-onset sepsis in premature infants based on the slope of microbial maturity in this application, the system is based on longitudinal data of a large number of premature infant cohorts. The system uses a random forest regression algorithm to screen core characteristic microbial communities and determines the early warning threshold through multiple rounds of model training and validation optimization. The construction process is scientific and rigorous, ensuring the stability and reliability of the early warning system and enabling it to meet the early warning needs of different premature infant groups.

[0019] 3. The preterm infant late-onset sepsis early warning system based on microbial maturity slope in this application adopts a non-invasive fecal sample collection method, avoiding invasive operation damage to preterm infants; the system has a high degree of automation in each module, can quickly complete sample processing, data calculation and early warning judgment, and the visual terminal interface is intuitive and easy to understand, which is convenient for medical staff to operate and interpret, and is suitable for widespread application in neonatal intensive care units.

[0020] Additional aspects and advantages of this application will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of this application. Attached Figure Description

[0021] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the construction method of the early warning system for late-onset sepsis in premature infants based on the slope of microbial maturity in this application. Figure 2 This is a succession trajectory diagram of gut microbiota maturation in premature infants for the purposes of this application; Figure 3 This is a schematic diagram illustrating the model construction and verification process of the early warning system in this application; Figure 4 This is a module architecture diagram of the early warning system for late-onset sepsis in premature infants based on the slope of microbial maturity in this application. Figure 5 ROC curve for performance verification of the early warning system in this application; Figure 6 This is a graph showing the contribution weight analysis of key characteristic bacteria in the prediction model of the early warning system of this application; Figure 7 This is a graph showing the predictive effect of the early warning system in this application on the risk of late-onset sepsis. Figure 8 This is a multivariate regression analysis plot showing the maturity slope as an independent risk factor in this application. Detailed Implementation

[0022] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0023] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0024] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "connection" can refer to a direct connection or an indirect connection via intermediate components (elements). The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the following description.

[0025] It should be noted that the concepts of "first" and "second" mentioned in this invention are only used to distinguish between devices, modules or units, and are not intended to limit these devices, modules or units to necessarily be different devices, modules or units, nor are they intended to limit the order or interdependence of the functions performed by these devices, modules or units.

[0026] Please see Figures 1 to 8 This application discloses an early warning system for late-onset sepsis in premature infants based on the slope of gut microbiota maturity and its construction method. By constructing a predictive model of gut microbiota maturity (MA), the complex gut microbiota data is transformed into a single quantitative indicator. By calculating the rate of change of this indicator with actual age (i.e., the maturity slope K), the system aims to dynamically monitor the intestinal development trajectory of premature infants. This enables early identification of high-risk individuals whose gut microbiota development has been arrested due to antibiotic exposure, thereby achieving early warning and precise intervention for sepsis.

[0027] It is important to note that the maturity slope K is the rate of change of the gut microbiota maturity MA over actual corrected age. It directly reflects whether antibiotic exposure has disrupted the structure of the beneficial core microbiota, with Enterococcus faecalis at its core, leading to stunted intestinal development in the host (premature infant). Enterococcus faecalis, as a beneficial core microbiota in the gut of premature infants, participates in the construction of the intestinal barrier function, nutrient metabolism, and immune regulation. It is a core component in maintaining gut microbiota homeostasis and normal development. When antibiotic exposure disrupts its dominant colonization, the succession process of the gut microbiota is significantly blocked, manifested as a slow increase or even a decrease in the gut microbiota maturity MA value, leading to a decrease in the maturity slope K value. Therefore, the dynamic change of the maturity slope K value can serve as a core evaluation indicator of the degree of gut microbiota damage and recovery status after antibiotic exposure. It is also a key basis for predicting whether premature infants face the risk of sepsis due to abnormal gut microbiota development, thus facilitating timely adjustment of antibiotic treatment plans and the implementation of gut microbiota intervention measures to prevent further infection progression.

[0028] Specifically, please combine Figure 1 The method for constructing the early warning system for late-onset sepsis in preterm infants based on the slope of microbial maturity includes the following steps: S101. Data Acquisition and Baseline Establishment. Longitudinal fecal samples were collected from the preterm infant cohort, meaning fecal samples were collected periodically from the same preterm infant every 2-3 days to ensure that changes in microbial maturity could be captured. Metagenomic DNA was extracted from the fecal samples using standard metagenomic DNA processing methods, and the DNA samples were sequenced using 16S rRNA high-throughput sequencing technology to obtain raw data on microbial composition.

[0029] To improve the accuracy and precision of the data, the raw data on the above-mentioned microbial community composition needs to be preprocessed. This involves quality control and deduplication of the obtained raw sequencing sequences to remove invalid data. The valid sequences are then compared with microbial databases to generate a microbial classification and abundance file containing information on the abundance of various genera, thus establishing a basic database of the gut microbiota of premature infants. Furthermore, the microbial databases used in this embodiment include the Greengenes database or the SILVA database.

[0030] S102. Screening of core feature microbial communities. A random forest regression algorithm is used to analyze each genera in the microbial taxonomy abundance file. Cross-validation is used to evaluate the contribution of each genera to the prediction of premature infant age. Genera with low contribution and no significant impact on the prediction results are removed, thus screening out the clock microbial communities of the top 20-30 genera in terms of contribution, such as Bifidobacterium and Enterobacter. The relative abundance of the screened clock microbial communities is used as the core feature microbial communities for model training.

[0031] At the same time, it is also necessary to remove noise bacteria that colonize momentarily to ensure that the core characteristic bacterial groups can accurately reflect the maturity status of the gut microbiota, providing a reliable basis for subsequent calculation of the microbiota maturity (MA) value.

[0032] S103. Training the gut microbiota maturity model. A nonlinear regression model is established using the core characteristic microbiota selected above. Specifically, the relative abundance of clock microbiota is used as the independent variable, and the actual corrected age of the preterm infant is used as the dependent variable. This establishes a quantitative correlation between gut microbiota composition and the actual developmental age of the preterm infant. The model can accurately deduce the gut microbiota maturity (MA) at a corresponding time by inputting the relative abundance of microbiota for a specific individual. This transforms complex, multidimensional microbiota data into a single, dynamically trackable quantitative indicator, adapting to real-time clinical monitoring needs.

[0033] After the model training is completed, validation tests are required to ensure that the prediction error of the microbial community maturity (MA) is within an acceptable error range.

[0034] S104. The rate of change of microbial community maturity with actual correction age is defined as the maturity slope K, and the formula for calculating the maturity slope K is as follows: , where MA nMA represents the maturity of the bacterial community in the nth test. n-1 CA represents the maturity of the bacterial community in the (n-1)th test. n CA represents the actual corrected age of the preterm infant at the Nth test. n-1 This represents the actual corrected age in the (n-1)th test. The maturity slope K value accurately reflects the dynamic trend of microbial community maturity. Furthermore, by concatenating multiple continuously calculated K values ​​according to the test sequence, a dynamically changing K value curve can be generated.

[0035] S105. Determine the warning threshold and perform a logical judgment between the maturity slope and the warning threshold, so as to output a warning signal when the maturity slope is less than the warning threshold.

[0036] The warning threshold in this application is determined by retrospectively analyzing data from children with sepsis. The retrospective K-value data of children diagnosed with late-onset sepsis are compared with the K-value data of healthy premature infants. Specifically, ROC curve analysis can be used to determine the warning threshold. ROC curve analysis is chosen because it can accurately distinguish the difference in K-values ​​between the case group and the control group, balancing the sensitivity and specificity of the warning.

[0037] This application sets the warning threshold at 0.3. When the real-time K value is ≥0.3 and the K value change curve of maturity slope is within the normal fluctuation range, it indicates that the intestinal flora maturation process is normal and there is no obvious stagnation of flora development. It can be determined that the preterm infants being tested have a low risk or no risk of sepsis. When the real-time K value is <0.3, it is determined that the preterm infants have a high risk of sepsis, indicating that antibiotic exposure may have destroyed the structure of beneficial core flora with Enterococcus faecalis as the core, resulting in stagnation of intestinal flora development. The risk of late-onset sepsis in preterm infants is significantly increased, triggering an emergency warning signal.

[0038] Furthermore, the maturity slope K-value of normally developing premature infants is usually stable between 0.8 and 1.2. Therefore, by comparing the maturity slope K-value curve of normally developing premature infants with that of the tested premature infants, it is possible to intuitively determine whether the maturation process of the gut microbiota in the tested premature infants has deviated from the normal trajectory. When the K-value curve of the tested premature infants consistently fluctuates within the normal range of 0.8-1.2, without any abnormal trends such as continuous decrease or flattening, it indicates that their gut microbiota maturation process is normal. If the K-value curve of the tested premature infants deviates from this normal range, or although it does not completely deviate, it shows a continuous downward trend or a trend towards flattening, or even gradually approaches or falls below the warning threshold of 0.3, signals of stagnant gut microbiota development can be identified in advance. This assists medical staff in conducting clinical assessments and interventions earlier, further improving the foresight and clinical applicability of the early warning system, and providing a more comprehensive reference for the precise management of the risk of late-onset sepsis in premature infants.

[0039] Please combine Figure 2 ,Figure 2 This illustrates the successive trajectory of intestinal maturity in preterm infants. Figure 2 This is a dual-axis line graph comparison, where the vertical axis represents the predicted microbial community maturity (MA, days), and the horizontal axis represents the actual corrected age (CA, days). Figure 2 The graph contains two main developmental trajectory curves. The solid line represents the developmental trajectory of normal preterm infants, where the gut microbiota maturity (MA) increases linearly and steadily with the actual corrected age (CA), and the maturity slope (K) approaches 1.0. The dashed line represents the high-risk trajectory for sepsis. The graph highlights the "slope deviation interval." 3-5 days before the onset of clinical sepsis symptoms, the dashed line first flattens out, meaning the maturity slope (K) significantly decreases and deviates from the baseline used for the developmental trajectory of normal preterm infants. This visually demonstrates the early characteristic of arrested gut microbiota maturation in high-risk preterm infants for sepsis, providing healthcare professionals with intuitive visual support for identifying early signs of sepsis and achieving early warning through curve comparison. Meanwhile, Figure 2 This demonstrates that the application achieves an earlier warning window by capturing the slope, a dynamic derivative indicator, rather than the static value of the maturity age (MA).

[0040] At the same time, it can also be combined with Figure 3 This diagram illustrates the construction method of the early warning method proposed in this application, clearly presenting the complete process from model training, testing to generalization validation: First, the dataset is split using clinical cohort data (such as CALM05_A), with 70% of the data used for training the gut microbiota maturity prediction model. The core clock microbiota are selected and model parameters are optimized using a random forest regression algorithm to construct the initial prediction model. The remaining 30% of the data is used for internal model testing to verify the model's prediction accuracy in the training cohort and adjust the model threshold in a timely manner to reduce prediction errors. Subsequently, multiple independent external cohorts are used to validate the model's generalization ability. External independent data are used to test the model's suitability for different preterm infant groups, ensuring the model's prediction stability and reliability in non-training cohorts. The illustrated process clearly establishes the steps for building the gut microbiota maturity prediction model proposed in this application, clarifies the core logic and data allocation principles of training, testing, and external validation, and forms the foundation for the entire early warning system implementation. It also further confirms the scientific nature and reproducibility of the early warning method construction process, providing process-level support for the clinical promotion and application of the early warning system.

[0041] Based on the above construction method, please combine Figure 4 The resulting early warning system for late-onset sepsis in premature infants based on gut microbiota maturity slope includes a sample collection and processing module, a data input module, a kinetic processing module, an early warning decision module, and a visualization terminal module. Through the collaborative work of these modules, dynamic monitoring of gut microbiota maturity and accurate early warning of sepsis risk in premature infants can be achieved.

[0042] The sample collection and processing module relies on the data collection and benchmark establishment steps in the construction method. It is used to collect fecal samples from premature infants, perform standardized processing, and generate microbial classification and grading files to provide a basis for subsequent analysis of microbial community maturity.

[0043] The data input module, acting as an intermediary for data transmission, transmits the microbial taxonomy and abundance files generated by the sample collection and processing module to the kinetics processing module. Equipped with a standard interface, the data input module supports not only connection to the data terminal of the sample collection and processing module but also direct wired or wireless connection to external sequencing equipment. Furthermore, the data input module incorporates a data format adaptation unit, automatically recognizing the raw data formats generated by the sample collection and processing module or other external sequencing equipment and performing preliminary format conversion to ensure the stability and accuracy of data transmission, preventing data loss or format corruption.

[0044] The dynamics processing module, as the core of the early warning system in this application, is mainly used to dynamically calculate relevant parameters of the maturity of the gut microbiota in preterm infants. It has a built-in maturity regression engine and slope calculation logic unit. The maturity regression engine is used to calculate the value of the microbiota maturity (MA) in real time. It is equipped with a random forest regression algorithm, which analyzes each genera in the microbial taxonomy abundance file. The contribution of each genera to the prediction of the age of preterm infants is evaluated through cross-validation. The top 20-30 clock microbiota are selected. The relative abundance of the clock microbiota is used as the independent variable and the actual corrected age of the preterm infant is used as the dependent variable. This establishes a quantitative correlation between the composition of the gut microbiota and the actual developmental age of the preterm infant. A nonlinear regression model is constructed, which enables the model to accurately deduce the microbiota maturity (MA) at the corresponding time by inputting the relative abundance of the microbiota at a certain time. This transforms complex multidimensional microbiota data into a single, dynamically trackable quantitative indicator, which is suitable for clinical real-time monitoring needs.

[0045] The slope calculation logic unit is used to perform time series difference operations on the microbial maturity MA output in real time by the maturity regression engine, thereby dynamically generating the microbial maturity slope K value and the K value change curve, so as to realize the dynamic capture of the maturation trend of the gut microbiota in premature infants.

[0046] Specifically, the maturity slope K is the rate at which the microbiota maturity changes with the actual corrected age of the preterm infant, and its calculation formula is as follows: , where MA n MA represents the maturity of the bacterial community in the nth test. n-1 CA represents the maturity of the bacterial community in the (n-1)th test. n CA represents the actual corrected age of the preterm infant at the Nth test. n-1The actual corrected age is the value from the (n-1)th test. The maturity slope K directly reflects the speed of the maturation process of the gut microbiota in preterm infants. By concatenating multiple sets of continuously calculated K values ​​according to the test time sequence, a dynamically changing K value curve is generated.

[0047] The early warning decision module, as the terminal decision core of the early warning system of this application, includes a logic comparator. Based on the early warning threshold and logic judgment steps in the construction method, it can compare the real-time K value generated by the dynamic processing module with the early warning threshold stored in the database, thereby realizing real-time judgment and early warning triggering of the risk of late-onset sepsis in premature infants.

[0048] The visualization terminal module features a visual interface that provides clinicians with a succession chart of the MA / CA (microbial community / catheter) for each premature infant, i.e., a dynamic curve showing the change in microbial maturity relative to the actual corrected age, as well as a maturity leakage K-value curve. This allows medical staff to visually observe the changing trends in microbial maturity. Furthermore, when the K-value curve deviates from the normal trajectory to a certain threshold range (in this embodiment, the K-value is less than 0.3), the visualization terminal interface can issue a high-risk warning signal through sound, light, or a window, prompting medical staff to take timely intervention measures.

[0049] In addition, medical staff can also observe the maturity slope K-value curve. If the K-value continues to decrease, that is, when the K-value curve shifts from normal to flattening, it can identify the signal that intestinal development has stopped, thereby capturing the early signs of sepsis and achieving early clinical warning. In particular, in this embodiment, when the maturity slope K is abnormal (such as consecutive plateaus in the test), the visualization terminal interface outputs a warning of "high risk of late-onset sepsis" through a red graphic.

[0050] In summary, the early warning system for late-onset sepsis in premature infants based on the slope of gut microbiota maturity proposed in this application can achieve dynamic monitoring of gut microbiota maturity in premature infants through the coordinated linkage of sample collection and processing module, data input module, dynamic processing module and early warning decision module. Furthermore, based on the deviation of the microbiota development trajectory, it can identify high-risk groups before the appearance of clinical symptoms, thus gaining valuable time for preventive intervention for forest farms.

[0051] This application utilizes line-to-line trajectory comparison, and by introducing a time weighting factor and a regression slope threshold, effectively eliminates the interference of individual differences in population size, transforming the static microbial community composition into a dynamic developmental rate indicator, making the assessment structure more accurate and reliable.

[0052] To verify the feasibility and clinical value of a late-onset sepsis early warning system for preterm infants based on gut microbiota maturity slope, the research team collected and analyzed longitudinal gut microbiota data from a large clinical cohort of preterm infants. The results showed that several days before the diagnosis of sepsis, the gut microbiota maturity slope K in these infants exhibited a significant stagnation or decline, consistent with the previous findings. Figure 2 The slope deviation range of the high-risk trajectory for sepsis in preterm infants showed consistent characteristics, while that in healthy preterm infants remained steadily increasing, conforming to the normal range of 0.8-1.2, and matching the MA / CA succession trajectory of normal preterm infants. The results confirmed that the maturity slope K can significantly distinguish between healthy preterm infants and children with sepsis, and the area under the curve (AUC) of the maturity slope K for predicting sepsis is significantly better than that of traditional inflammatory markers such as CRP. (See [reference needed]). Figure 5 , Figure 5 To visually demonstrate the diagnostic accuracy, sensitivity, and specificity of the late-onset sepsis (LOS) early warning method for preterm infants based on the slope of microbial maturity in this application, in a clinical validation set, the graph uses sensitivity (true positive rate, i.e., the proportion of infants who actually have sepsis but are correctly identified as high-risk by this early warning method) as the ordinate and specificity (false positive rate, i.e., the proportion of actually healthy preterm infants who are misidentified as high-risk by this early warning method) as the abscissa. Both the ordinate and abscissa range from 0 to 1. The closer the curve is to the upper left corner, the better the overall performance of the early warning method, and the lower the missed diagnosis rate and misdiagnosis rate. The graph clearly indicates that the area under the ROC curve (AUC value) corresponding to the early warning method of this application is ≥0.85. The closer the AUC value is to 1, the higher the diagnostic accuracy of the early warning method. The AUC value of this application is ≥0.85, confirming that the early warning method based on the slope of microbial maturity has high diagnostic reliability. Meanwhile, the figure shows the ROC curves (represented by dashed lines) for traditional infection indicators (such as C-reactive protein CRP and white blood cell count WBC). A direct comparison reveals that the ROC curve of the early warning method based on the slope of bacterial flora maturity in this application is significantly more inclined towards the upper left corner, and its AUC value is significantly higher than that of traditional infection indicators such as CRP and white blood cell count. This demonstrates that, compared to traditional infection indicators, the monitoring method based on the slope of bacterial flora maturity in this application not only has a higher area under the curve (AUC) in the ultra-early diagnosis of late-onset sepsis in premature infants, but also effectively reduces the missed diagnosis rate and misdiagnosis rate. This further confirms the clinical advantages and application value of this early warning model, providing solid performance data support for its promotion and application in the neonatal intensive care unit.

[0053] This application, through contribution analysis using a random forest model, identifies *Enterococcus faecium* as the most critical positively correlated bacterium promoting intestinal development in preterm infants. This conclusion can be verified through... Figure 6 Further evidence. Figure 6This paper presents the results of Shapley value analysis based on a random forest model. This analysis, with the core objective of predicting gut microbiota maturity, systematically quantifies and ranks the core bacterial genera that contribute the most to the prediction of gut microbiota maturity. The figure shows that *Enterococcus faecalis* is the most positively correlated characteristic bacterium with the highest weight among all genera, and its positive contribution to the regulation of gut microbiota maturity in preterm infants is significantly higher than that of other genera. *Klebsiella pneumoniae* is the second most positively correlated genera. The above Shapley value analysis results quantitatively clarify the core position of *Enterococcus faecalis* in the gut microbiota maturation process of preterm infants, providing crucial data support for subsequent mechanism verification experiments targeting this genus and establishing its scientific validity as a core target of the early warning system.

[0054] Furthermore, this application further validates the early warning indicator based on the slope of gut microbiota maturity and the practical application value of the system from a clinical prognostic perspective by constructing a Kaplan-Meier cumulative incidence curve. For detailed validation results, please refer to [link to relevant documentation]. Figure 7 . Figure 7 Plotting the cumulative incidence of late-onset sepsis in preterm infants on the vertical axis and the follow-up time after birth on the horizontal axis, the differences in the risk of late-onset sepsis among different gut microbiota development phenotype groups are clearly presented. Specifically, the test group (purple curve) identified as having delayed gut microbiota development based on gut microbiota maturity-related indicators by the early warning system of this application showed a significant increasing cumulative incidence of late-onset sepsis with advancing actual corrected age, and was significantly higher than the test group identified as having rapid gut microbiota development (pink curve) throughout the process. Log-Rank test showed a highly statistically significant difference in the risk between the two groups (P=0.004), and the risk of late-onset sepsis in the delayed development group was 3.12 times that in the rapid development group (HR=3.12).

[0055] The above results fully demonstrate that the microbial maturity-related indicators proposed in this application can effectively distinguish high-risk premature infants with late-onset sepsis, and have clear and significant clinical early warning efficacy, providing a reliable quantitative basis for early clinical identification of high-risk individuals and timely targeted intervention.

[0056] This application also conducted independent validation of risk factors through multivariate Cox regression analysis, further clarifying the clinical predictive value of the core indicators of this application. The relevant analysis results are presented in the form of a forest plot. Figure 8 The forest plot system incorporates the microbial maturity-related indicators of this application with various traditional clinical indicators (including common influencing factors of late-onset sepsis in premature infants such as history of antibiotic use, birth weight, and sex). It conducts a cross-sectional comparative analysis of the hazard ratio (HR) and statistical significance of each indicator. At the same time, it strictly corrects for the above-mentioned traditional clinical confounding factors in the model to eliminate interference and accurately assess the independent predictive power of the core indicators of this application.

[0057] Figure 8 The results clearly show that, after multifactorial correction, the delayed microbiota maturation index defined in this application remains a significant independent risk factor for late-onset sepsis in preterm infants, with a hazard ratio of 2.96 (P=0.033), demonstrating clear statistical significance. In contrast, the predictive statistical significance of some traditional clinical indicators decreased significantly or became insignificant after correction. This result fully confirms that the microbiota maturation-related indicators proposed in this application have clinical predictive value superior to or independent of existing conventional clinical indicators. They overcome the predictive limitations of traditional clinical indicators, providing a new and irreplaceable independent predictive dimension for the risk assessment of late-onset sepsis in preterm infants. Furthermore, it further confirms that traditional clinical assessment methods cannot replace the early warning indicators proposed in this application.

[0058] In summary, the early warning system for late-onset sepsis in preterm infants based on gut microbiota maturation slope proposed in this application can capture signals of impaired gut microbiota maturation before the clinical symptoms of late-onset sepsis appear by dynamically monitoring the changing trend of the gut microbiota maturation slope K value in preterm infants. The early warning time is earlier than that of traditional blood culture, providing sufficient time for early clinical intervention and effectively reducing the incidence and mortality of severe late-onset sepsis.

[0059] The early warning system proposed in this application is constructed based on longitudinal data from a large number of preterm infant cohorts. Core characteristic microbial communities are screened through random forest regression algorithm, and the early warning threshold is determined through multiple rounds of model training, validation and optimization. The construction process is scientific and rigorous, ensuring the stability and reliability of the early warning system and enabling it to meet the early warning needs of different preterm infant groups.

[0060] The early warning system proposed in this application adopts a non-invasive method for collecting fecal samples, avoiding damage to premature infants caused by invasive procedures. The system has a high degree of automation in each module, which can quickly complete sample processing, data calculation and early warning judgment. The visual terminal interface is intuitive and easy to understand, making it easy for medical staff to operate and interpret. It is suitable for widespread application in neonatal intensive care units.

[0061] The above description is only a partial embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for constructing an early warning system for late-onset sepsis in premature infants based on the slope of gut microbiota maturity, characterized in that, Includes the following steps: S101. Data acquisition and benchmark establishment to generate classification abundance files; S102, Screening for core characteristic bacterial groups; S103, Training the microbial community maturity model; S104. The rate of change of microbial community maturity with actual correction age is defined as maturity slope. S105. Determine the warning threshold and perform a logical judgment between the maturity slope and the warning threshold, so as to output a warning signal when the maturity slope is less than the warning threshold.

2. The construction method according to claim 1, characterized in that, In step S101, longitudinal fecal samples are collected from the preterm infant cohort, and raw data on the microbial community composition are obtained using 16S rRNA high-throughput sequencing technology.

3. The construction method according to claim 2, characterized in that, The raw data of the obtained microbial community composition were quality controlled and deduplicated, and the microbial dataset was compared to generate a classification abundance file.

4. The construction method according to claim 1, characterized in that, In step S102, the contribution of each bacterial genus to age prediction is evaluated by cross-validation using a random forest regression algorithm. Clock bacteria within the target ranking range are selected, and these clock bacteria are used as core feature bacteria. Noise bacteria that colonize momentarily are removed.

5. The construction method according to claim 4, characterized in that, In step S103, the selected clock microbiota are used to establish a nonlinear regression model, which is used to output the microbiota maturity corresponding to the microbiota abundance of a specific individual.

6. The construction method according to claim 1, characterized in that, In S104, the maturity slope is the degree of change in the maturity of the microbial community with the actual corrected age.

7. The construction method according to claim 1, characterized in that, In S105, a warning threshold is determined by retrospectively analyzing data of children with sepsis. When the maturity slope is less than the warning threshold, the premature infant is judged to have a high risk of late-onset sepsis.

8. A late-onset sepsis early warning system for preterm infants based on gut microbiota maturity slope, constructed using the construction method described in any one of claims 1-7, characterized in that, It includes a sample acquisition and processing module, a data input module, a kinetics processing module, and an early warning and decision-making module; The sample collection and processing module is used to generate microbial classification and abundance files; The data input module is used to receive the microbial classification and abundance file generated by the sample collection and processing module, and transmit the relevant microbial classification and abundance file to the kinetic processing module; The dynamic processing module is equipped with a maturity regression engine and a slope calculation logic unit. The maturity regression engine is used to calculate the microbial maturity MA value of preterm infants in real time. The slope calculation logic unit is used to perform time series difference operation on the microbial maturity MA value to calculate the maturity slope K value and generate the microbial maturity slope K value curve. The early warning decision module includes a logic comparator, which is used to compare and judge the maturity slope K value generated in real time by the dynamics processing module with the early warning threshold stored in the database.

9. The early warning system for late-onset sepsis in premature infants based on the slope of gut microbiota maturity, as described in claim 8, is characterized in that... It also includes a visual terminal interface, which is electrically connected to the early warning decision module. The visual terminal interface is used to provide the succession diagram of MA / CA for premature infants and to issue an early warning prompt when the K value curve deviates from the normal trajectory to the early warning threshold.

10. The early warning system for late-onset sepsis in preterm infants based on the slope of gut microbiota maturity according to claim 9, characterized in that, When the maturity slope continues to decrease, premature infants are considered to be at risk of late-onset sepsis.

Images