An assistant identification system for hand-foot-mouth disease by high-throughput sequencing of fecal flora

Abstract

The application relates to the technical field of microorganism detection, and discloses a hand-foot-mouth disease auxiliary identification system for high-throughput sequencing of fecal flora, which comprises a sample pretreatment module, a sequencing collection module, a flora feature mining module, a flora quantification module and an intelligent identification module. In the process of high-throughput sequencing auxiliary identification of the hand-foot-mouth disease fecal flora, multi-dimensional mining of flora features in different infection stages can be realized through non-reference sequencing and flora interaction network analysis, so that specific identification of the to-be-detected sample can be completed under the condition of no external database dependence, and flora disorder degree data in different disease course stages can be collected; then, the collected information data is subjected to hierarchical analysis, so that the infection stage and the severe risk are judged.

Description

Technical Field

[0001] This invention relates to the field of microbial detection technology, and more specifically discloses a hand-foot-mouth disease-assisted identification system using high-throughput sequencing of fecal microbiota. Background Technology

[0002] Hand, foot, and mouth disease (HFMD) is a common infectious disease in children caused by enteroviruses. Typical symptoms include fever and rashes or blisters on the hands, feet, and mouth. A small number of children may rapidly develop severe illnesses such as encephalitis and myocarditis, or even die. Traditional viral nucleic acid testing has limitations such as missed detection during the window period, inability to distinguish the stages of the disease, and difficulty in predicting the risk of severe illness. However, research has shown that HFMD virus infection will cause significant dysbiosis of the gut microbiota before clinical symptoms appear. Therefore, high-throughput sequencing of the gut microbiota in fecal samples is necessary.

[0003] In the prior art, document CN119530459A discloses a "composition for a six-component fluorescent detection of hand-foot-mouth disease pathogens," which includes primers and probes that amplify and identify enterovirus 71, enterovirus EV, Coxsackievirus A16, Coxsackievirus A6, Coxsackievirus A10, and Coxsackievirus A4, respectively. The composition provided in this application is suitable for the nucleic acid detection of enteroviruses extracted from samples such as pharyngeal swabs, anal swabs, vesicular exudate, feces, cerebrospinal fluid, and viral isolates, exhibiting high sensitivity and specificity.

[0004] The document CN104073569A discloses "Molecular markers for diagnosing severe hand-foot-mouth disease and their detection methods and kits," including miR-671-5p, miR-16-5p, and miR-150-3p. The biomarkers for diagnosing severe hand-foot-mouth disease involved in this invention possess specificity, good stability, and high efficiency, exhibiting good diagnostic characteristics. Furthermore, the detection kits for the biomarkers for diagnosing severe hand-foot-mouth disease involved in this invention are low-cost and easy to promote on a large scale.

[0005] While existing technologies can achieve pathogen typing and auxiliary diagnosis of severe hand-foot-mouth disease by detecting viral nucleic acids or specific miRNA markers, and have advantages such as high sensitivity and convenient operation, they only focus on the detection of viruses or specific molecular targets, and cannot cover the pre-pathological features of latent period dysbiosis. The false negative rate during the window period is high. At the same time, the microbiome-related detection relies on public databases, lacks specific microbiome annotation, has poor specificity, and only performs static abundance analysis, which is difficult to distinguish from common gastrointestinal diseases. The false positive rate is high, and it cannot stratify the course of the disease or predict severe cases. The utilization rate of sequencing data is low, which limits its clinical application value. Summary of the Invention

[0006] The main technical problem solved by this invention is to provide a hand-foot-mouth disease-assisted identification system based on high-throughput sequencing of fecal microbiota, which can solve the problems mentioned in the background art.

[0007] To solve the above-mentioned technical problems, according to one aspect of the present invention, more specifically, a fecal microbiota high-throughput sequencing auxiliary identification system for hand-foot-mouth disease, comprising: a sample preprocessing module, a sequencing acquisition module, a microbiota feature mining module, a microbiota quantification module, and an intelligent identification module; The sample preprocessing module performs stratified removal of contaminating bacteria, unbiased amplification of the entire microbial community, and initial screening and noise reduction of sequencing on fecal samples to prepare pure microbial nucleic acid samples. The sequencing acquisition module performs high-throughput sequencing of the entire microbial community without relying on external microbial databases, clustering and grouping based on the sequence characteristics themselves, and screening of effective microbial community sequences to generate the original microbial community sequencing dataset. The microbial feature mining module constructs a baseline map of healthy gut microbiota, screens hand-foot-mouth disease-specific disordered microbial clusters, and extracts exclusive pathological feature tags. The microbiome quantification module constructs a bidirectional interaction network of gut microbiota, calculates a quantitative score of microbiome disorder, and calibrates the specific identification threshold for hand-foot-mouth disease. The intelligent identification module completes the four-level disease progression determination of hand-foot-mouth disease from the incubation period to the recovery period, intelligent early warning of the risk of severe complications, and standardized auxiliary identification result output.

[0008] Furthermore, the sample preprocessing module includes: an impurity removal module, a microbial community amplification module, and a sequencing noise reduction module; Impurity removal module: Removes food residue, exfoliated cells from the host intestinal tract, and free viral nucleic acid fragments from fecal samples layer by layer through three-stage gradient centrifugation; Microbial community amplification module: Employs whole-genome random primer amplification method to non-discriminatory amplify nucleic acids of all microorganisms in the sample; Sequencing noise reduction module: Based on the signal-to-noise ratio and fragment length consistency of the sequencing fragments, invalid sequencing fragments are removed to complete data purification.

[0009] Furthermore, the sequencing acquisition module includes: a full-gene sequencing module, a sequencing clustering module, and an effective sequence screening module; Full-scale gene sequencing module: Performs unbiased high-throughput sequencing on the amplified bacterial community nucleic acids to obtain the complete bacterial community gene sequence; Sequencing clustering module: Based on sequence similarity, gene fragment conservation, and base arrangement characteristics, it autonomously clusters and groups sequencing sequences. Effective sequence screening module: Removes low-quality, short fragments and non-microbial sequences, and retains complete and effective bacterial community sequences.

[0010] Furthermore, the microbial community feature mining module includes: a healthy microbial community modeling module, a disordered microbial community screening module, and a disease feature extraction module; Healthy gut microbiota modeling module: Constructs a baseline structural map of healthy gut microbiota based on clustering data from infection-free healthy samples; Disordered microbiota screening module: Compares the microbiota structure of the sample to be tested with the healthy baseline map to screen for hand-foot-mouth disease-specific disordered microbiota clusters; Disease feature extraction module: Extracts hand-foot-mouth disease-specific feature tags based on microbial community structure shift, core beneficial microbial community attenuation, and abnormal proliferation of pathogenic microbial communities.

[0011] Furthermore, the microbial community quantification module includes: a microbial community relationship construction module, a disorder numerical calculation module, and a judgment threshold calibration module; Microbiome Relationship Construction Module: Based on clustered microbiome groups, calculate the abundance linkage, survival dependence, and competitive inhibition relationships among microbiome groups to construct a bidirectional interaction network of gut microbiota; Disorder numerical calculation module: quantifies the degree of imbalance in microbial interactions and the magnitude of microbial structure shift, and generates microbial disorder values; Threshold calibration module: Through iterative training with clinical samples, it calibrates and defines the threshold range for hand-foot-mouth disease-specific microbial dysbiosis.

[0012] Furthermore, the intelligent identification module includes: a disease stage judgment module, a severe illness risk warning module, and an identification result output module; Disease progression assessment module: Based on the degree of bacterial flora disorder and the characteristics of bacterial flora structure, determine the incubation period, mild stage, severe progression stage, and recovery stage of hand-foot-mouth disease; Severe illness risk warning module: Identifies the characteristics of near disappearance of the core protective gut microbiota and complete disruption of the microbiota interaction network, triggering a warning of severe complications; The identification result output module integrates information on microbial community characteristics, disorder scores, and disease course determination to generate a standardized auxiliary identification report.

[0013] Furthermore, the three-stage gradient centrifugation is performed at 800 r / min for 5 minutes, 3000 r / min for 10 minutes, and 12000 r / min for 15 minutes.

[0014] The beneficial effects of this invention's fecal microbiota high-throughput sequencing-assisted identification system for hand-foot-mouth disease are as follows: During the high-throughput sequencing-assisted identification of fecal microbiota in hand-foot-mouth disease, parameter-free sequencing and microbiota interaction network analysis enable multi-dimensional mining of microbiota characteristics at different infection stages. This allows for specific identification of the test sample without external database dependence. Furthermore, it collects microbiota dysbiosis data at different disease stages and performs stratified analysis of the collected data to determine the infection stage and severity risk. Compared to existing hand-foot-mouth disease identification methods that often focus only on single viral target indicators, this invention more comprehensively reflects the microecological pathological characteristics throughout the disease process, making the identification more closely aligned with the actual disease progression. In addition, the identification process eliminates the need for reliance on public databases and manual comparison analysis, resulting in a high degree of automation and significantly improving early screening efficiency, thus meeting the needs of large-scale population screening. Attached Figure Description

[0015] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0016] Figure 1 This is a schematic diagram of the system module architecture. Detailed Implementation

[0017] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0018] According to one aspect of the invention, such as Figure 1 As shown, a high-throughput sequencing system for fecal microbiota is provided to assist in the identification of hand-foot-mouth disease. The system includes a sample preprocessing module for stratified removal of contaminating bacteria from fecal samples, unbiased amplification of the entire microbiota, and initial screening and noise reduction during sequencing to prepare purified microbial nucleic acid samples. This module includes: Impurity removal module: Removes food residue, exfoliated cells from the host intestinal tract, and free viral nucleic acid fragments from fecal samples layer by layer through three-stage gradient centrifugation; First, fresh fecal samples are mixed with sterile buffer to form a uniform suspension to avoid local impurities from accumulating and affecting the separation effect. Then, the suspension is added to a sterile centrifuge tube for the first stage of low-speed centrifugation (800 r / min for 5 minutes). The density difference is used to settle coarse impurities such as food residues with larger volume and higher density to the bottom of the tube. The upper mixture is carefully aspirated and transferred to a new centrifuge tube. Then, a second stage of medium-speed centrifugation (3000 r / min for 10 minutes) is performed to allow medium-sized impurities such as host intestinal cells to precipitate fully, and the supernatant is collected again to remove host cell contamination. Finally, a third stage of high-speed centrifugation (12000 r / min for 15 minutes) is performed to separate free viral nucleic acid fragments and tiny impurity particles by sedimentation. The supernatant containing intact intestinal microbial cells is retained. The centrifugation environment is strictly controlled to ensure sterility throughout the process, avoid the introduction of exogenous bacteria, and completely block the interference of viral fragments on subsequent microbial sequencing signals.

[0019] Microbial community amplification module: Employs whole-genome random primer amplification method to non-discriminatory amplify nucleic acids of all microorganisms in the sample; Specifically, the centrifuged microbial cells are first subjected to a gentle lysis process to release all nucleic acid material inside the cells. Liquefaction impurities are then removed using a low-interference nucleic acid purification process to obtain a complete mixed microbial nucleic acid template. Subsequently, a reaction system adapted to amplify the entire microbial community is prepared, and species-free random degenerate primers are selected (e.g., primer sequence 5'-NNNNNNNN-3', concentration 0.5µM). These primers can bind non-specifically to conserved regions of various microbial genomes and will not preferentially amplify specific pathogenic bacteria or dominant bacterial groups. Next, the nucleic acid template was mixed with the reaction system and placed in a gradient amplification instrument. A progressive temperature-controlled amplification program was executed: first, pre-denaturation at 95℃ for 3 minutes, then 25 cycles (denaturation at 95℃ for 30 seconds, annealing at 42℃ for 45 seconds, extension at 72℃ for 90 seconds), and finally final extension at 72℃ for 5 minutes. This ensured that low-abundance and unknown species of intestinal microbial nucleic acids could be stably amplified. No targeted screening steps were introduced throughout the process, and the genetic information of unknown disordered bacterial flora related to hand-foot-mouth disease was completely preserved. After amplification, the uniformity of the amplified product concentration was confirmed by micro-quantitative nucleic acid detection (concentration ≥10ng / µL, total amount ≥1µg) to avoid amplification deviation.

[0020] Sequencing noise reduction module: Based on the signal-to-noise ratio and fragment length consistency of the sequencing fragments, invalid sequencing fragments are removed to complete data purification; The process begins by sending the amplified bacterial nucleic acid samples into a high-throughput sequencing platform (e.g., Illumina NovaSeq6000, using a paired-end 150bp mode) for sequencing, generating raw sequencing fragment data in batches. Then, each sequencing fragment is individually checked for signal quality, and the signal-to-noise ratio (SNR) (the ratio of signal intensity to background noise intensity) is calculated. A SNR threshold of ≥15dB is set, and invalid fragments with low SNR, weak signals, large fluctuations, or a high risk of base misinterpretation are removed. Next, the length of the selected fragments is verified: a lower length limit of 200bp and an upper length limit of 1500bp are set, removing short fragments shorter than 200bp (incomplete information) and abnormally long fragments longer than 1500bp (potentially chimeras). Then, Trimmomatic software was used for quality trimming, with a sliding window size of 4 bp and an average quality threshold of Q20. Low-quality bases were removed, and a small number of residual non-microbial impurity sequences were filtered out. The specific filtering method was as follows: the remaining fragments were compared with the host reference genome (human hg38 genome) using short sequence alignment (using BWA software with parameters mem-t4-k19-w100, setting a matching rate threshold of ≤80% to be considered non-host source and retained, and a matching rate of >80% to be considered host contamination and removed). Then, the sequences were compared with common environmental pollutant databases (such as PhiX174, E. coli bacteriophage λ vector sequences, etc.), and pollutant sequences were removed with a similarity threshold of ≥95%. A step-by-step progressive screening process was adopted to filter unqualified fragments in round by round to avoid accidentally deleting effective sequences of real microbial communities. Finally, effective microbial community sequencing fragments with reliable signals, compliant lengths, and pure sources were integrated and retained to complete the basic purification of sequencing data, providing high-quality raw data for subsequent microbial community feature mining.

[0021] The sequencing acquisition module performs high-throughput sequencing of the entire microbial community without relying on external microbial databases, performs clustering and grouping based on sequence characteristics, and screens effective microbial sequences to generate the original microbial community sequencing dataset. This module includes: Full-scale gene sequencing module: Performs unbiased high-throughput sequencing on the amplified bacterial community nucleic acids to obtain the complete bacterial community gene sequence; First, a standardized library was constructed from the purified microbial nucleic acid samples. The entire process adopted an unbiased library construction process, without setting species screening thresholds or enriching or removing specific microbial sequences, to ensure that the library uniformly contains nucleic acid fragments of high-abundance, low-abundance, and unknown species of intestinal microorganisms. The prepared library was then loaded onto a high-throughput sequencing platform, and the sequencing process was run in parameter-free sequencing mode. The sequencing reaction parameters were dynamically adjusted during the sequencing process to ensure balanced sequencing coverage of microbial nucleic acids with different GC contents and fragment lengths, avoiding overload of high-abundance bacterial communities and missing sequencing of low-abundance hand-foot-mouth disease-related disordered bacterial communities. Finally, the reaction status is monitored in real time during the sequencing process, and sequencing signal deviations are corrected in a timely manner. In the end, complete gene sequence data covering all gut microbiota are generated. The entire process does not rely on external microbial reference databases, and all genetic information of the gut microbiota after hand-foot-mouth disease infection is completely preserved, avoiding the loss of unknown disordered microbiota information.

[0022] Sequencing clustering module: Based on sequence similarity, gene fragment conservation, and base arrangement characteristics, it autonomously clusters and groups sequencing sequences. First, core feature information is extracted from the full sequencing sequence, including homology similarity between sequences (the global alignment score is calculated using the Needleman-Wunsch algorithm), conserved domain features of gene fragments (the known conserved domain database Pfam is searched using HMMER software), and base composition and arrangement patterns (e.g., GC content, k-mer frequency, k=4), forming a unique feature identifier for each sequence. Then, the clustering analysis process is initiated: using the UPARSE algorithm (USEARCH v11 software), with a sequence similarity threshold of 97%, operational taxonomic units (OTUs) are clustered. The specific steps are as follows: ① Sort all sequences in descending order of abundance and extract the sequence with the highest abundance as the representative OTU sequence; ② Compare other sequences with the representative sequence; those with a similarity of ≥97% are classified into this OTU. ③ Repeat the iteration until all sequences are assigned; at the same time, use the DADA2 algorithm to perform single nucleotide resolution clustering at the ASV (amplifier sequence variant) level to distinguish closely related strains; Through iterative optimization of dynamic similarity thresholds, for example, setting the initial similarity threshold to 85% for coarse clustering, and then fine-tuning the threshold based on sequence consistency within the cluster after the first round of coarse clustering (increasing by 1% each round), sequences with similarity greater than the threshold (finally ≥97%) are finally grouped into the same bacterial cluster. The comparison rules are dynamically optimized during the clustering process to reduce interference from sequence sequencing errors (e.g., using noise reduction algorithms to remove PCR chimeras, UCHIME algorithm), ensuring high homology of sequences within the same cluster group and significant differences in sequence characteristics between different cluster groups. This accurately identifies different clusters such as healthy bacteria and hand-foot-mouth disease-specific disordered bacteria, providing a clear grouping basis for subsequent bacterial feature mining.

[0023] Effective sequence screening module: Removes low-quality, short fragments and non-microbial sequences, and retains complete and effective bacterial community sequences; First, a multi-dimensional quality screening was conducted on the clustered sequence data: FastQC software was used to evaluate the base quality score, and samples with a Q30 score (error rate ≤ 0.1%) or higher of ≥ 80% were retained. Then, fragment length verification was performed, with the retention length set at 250bp~1200bp, and fragments with a length < 250bp and abnormal chimeric sequences with a length > 1200bp were removed (the complete length was confirmed by comparing the positions of the primers at both ends). Then, sequence origin tracing and identification were performed: Kraken2 software (database: standard microbial reference genome plus host genome GRCh38) was used to classify and annotate each sequence, with a classification confidence threshold of 0.5. Sequences annotated as bacteria, archaea, or fungi were retained, while sequences annotated as host (human) or unknown were removed. Further, the sequence was compared with common environmental microbial pollution databases (including common pollutants in soil, water sources, and laboratories such as Bacillus and Pseudomonas). Sequences with a coverage threshold of <10% and a similarity of <85% were determined to be of microbial origin and retained; otherwise, they were removed. The screening process employs a gradient filtering strategy: first, filtering by quality score (removing sequences below Q20), then by length, and finally by source. After each step, the number of retained sequences is counted to ensure that the final number of retained sequences is no less than 60% of the original number of sequences, avoiding over-screening that could lead to the loss of low-abundance disordered bacterial sequences related to hand-foot-mouth disease. Finally, reliable, complete, and pure bacterial sequences are selected and retained to form a standardized original bacterial sequencing dataset (an average of 50,000-80,000 valid sequences per sample), ensuring the accuracy of subsequent bacterial quantification and identification analysis.

[0024] The gut microbiota feature mining module constructs a baseline map of healthy gut microbiota, screens for hand-foot-mouth disease-specific dysbiosis clusters, and extracts specific pathological feature tags. This module includes: Healthy gut microbiota modeling module: Constructs a baseline structural map of healthy gut microbiota based on clustering data from infection-free healthy samples; First, 80 fecal microbiota sequencing samples from healthy children with no history of hand-foot-mouth disease infection, no chronic digestive tract diseases, and whose age matched the target population (e.g., 3-6 years old) were selected as the modeling baseline sample to ensure that the samples covered different regions (e.g., East China, South China, North China), daily diet and living environment, and to eliminate the limitations of a single population sample. Then, the microbial clustering data (OTU / ASV table) of all benchmark samples were integrated, the data format was unified and batch correction was performed (the ComBat-seq algorithm was used to eliminate sequencing batch differences) to ensure data consistency and comparability. Then, core features were extracted from multiple dimensions such as microbial composition structure (calculating α diversity: Shannon index, Chao1 index), inter-microbial association patterns (calculating β diversity: Bray-Curtis distance), microbial abundance distribution pattern (relative abundance ≥1% is dominant microbial community), etc. Key indicators such as the composition of dominant microbial communities in healthy gut (e.g., Firmicutes / Bacteroidetes ratio is about 1.5:1), the proportion range of core microbial communities (Bacteroidetes 10%~25%, Faecalibacterium 5%~15%), and microbial interaction balance relationship (co-occurrence network positive and negative connectivity ratio is about 1.2) were analyzed. Finally, a visualized baseline structure map of healthy gut microbiota is constructed based on multi-dimensional features: This map consists of two parts: Heatmap: Shows the relative abundance distribution of the top 50 bacterial genera in healthy samples; the shade of color represents the abundance level. Network diagram: Nodes represent bacterial genera, and edges represent significant co-occurrence or mutual exclusion relationships (Spearman correlation coefficient |r|>0.6 and p<0.05). Red edges indicate positive correlation, and blue edges indicate negative correlation. This diagram fully presents the distribution characteristics and balance of the gut microbiota in a healthy state. After completion, it was repeatedly validated and calibrated using multiple independent healthy samples (an additional 20 cases were selected) to ensure the stability, universality, and accuracy of the baseline map, providing a reliable reference for subsequent comparisons of samples to be tested.

[0025] Disordered microbiota screening module: Compares the microbiota structure of the sample to be tested with the healthy baseline map to screen for hand-foot-mouth disease-specific disordered microbiota clusters; First, the microbial clustering data of the samples to be tested are standardized and normalized (using Z-score standardization or total abundance scaling) to ensure that the data dimensions and statistical standards are completely consistent with the baseline map of healthy gut microbiota, avoiding data interference caused by individual sample differences and sequencing batch differences, and ensuring the accuracy of comparative analysis. Subsequently, the bacterial community structure of the test sample was comprehensively compared with the health baseline map from three core dimensions: bacterial community composition (Jaccard similarity), relative abundance distribution of the bacterial community (Wilcoxon rank-sum test), and inter-microbial association (network modularity coefficient). Differential bacterial communities were screened using LEfSe (linear discriminant analysis effect size) software. For example, an LDA threshold of >2.0 and p<0.05 were set. The specific steps were as follows: first, the Kruskal-Wallis test was used to screen differential bacteria between the test sample and the healthy control group; then, the Wilcoxon test was used to verify the significantly different bacteria; finally, the LDA score was calculated, and bacteria with LDA>2.0 were retained as candidate specific disordered bacterial communities. Simultaneously, by combining the clinical and pathological correlation patterns of hand-foot-mouth disease (HFMD), specific identification was performed on deviation characteristics: excluding non-HFMD-induced dysbiosis such as common diarrhea and indigestion (e.g., diarrhea patients often show transient increases in Prevotella spp., but this is not significant in HFMD), and distinguishing between general intestinal flora fluctuations and pathological dysbiosis caused by HFMD infection; finally, identifying bacterial clusters that are stably present only in HFMD-infected samples (abundance change fold ≥2 in more than 80% of HFMD samples) and not significantly present in healthy samples and common digestive tract disease samples (including abnormal proliferation of Enterococcus spp., significant reduction of Bifidobacterium spp., etc.), thus completing the precise screening of HFMD-specific dysbiosis bacterial clusters; To illustrate the screening process more clearly, a fecal sample from a child with hand-foot-mouth disease in the incubation period is used as an example: After processing by the sequencing acquisition module, 72,418 valid sequences were obtained from this sample, clustered into 1,247 OTUs. The OTU abundance table of the sample was compared with the healthy baseline map, and LEfSe analysis (LDA threshold > 2.0, p < 0.05) was used to screen out 47 differentially expressed bacterial groups. Further, considering the clinicopathological association patterns of hand-foot-mouth disease, 13 fluctuating bacterial groups commonly found in ordinary diarrhea samples (such as Prevotella and Bilitrophus) were removed, retaining 34 candidate specific dysbiotic bacterial groups for hand-foot-mouth disease. Based on this, the samples with an abundance change ≥ 2 and consistent with the same disease pattern observed in our hospital were further selected. The screening criterion was a frequency of ≥80% in 20 hand-foot-mouth disease samples. Ultimately, 11 core specific dysbiosis bacterial groups were identified. Among them, the relative abundance of Enterococcus increased from 0.8%±0.3% of the healthy baseline to 5.2% of the sample, a fold change of 6.5; Bifidobacterium decreased from 12.5%±3.2% of the healthy baseline to 3.1% of the sample, a fold change of -4.0; Dorea increased from 2.1% to 7.8%, a fold change of 3.7; and Bacteroides decreased from 18.5% to 6.2%, a fold change of -3.0.

[0026] Disease feature extraction module: Extracts hand-foot-mouth disease-specific feature tags based on microbial community structure shift, core beneficial microbial community attenuation, and abnormal proliferation of pathogenic microbial communities; First, we conducted an in-depth analysis of the selected hand-foot-mouth disease-specific disordered bacterial clusters, quantified the deviation magnitude and pattern of the overall bacterial structure compared to the healthy baseline map, and identified the specific structural changes such as decreased bacterial diversity and disruption of bacterial distribution uniformity. Then, focusing on the core beneficial gut microbiota, we statistically analyzed the abundance decline rate and the degree of damage to the population integrity of each microbiota, sorted out the types of beneficial microbiota that significantly declined after hand-foot-mouth disease infection and their decline patterns, formed the characteristics of the decline of core beneficial microbiota, and at the same time investigated the proliferation of pathogenic microorganisms in the microbiota, identified the categories, proliferation amplitude and distribution characteristics of potential pathogenic microbiota with abnormal proliferation, and determined the specific indicators of abnormal proliferation of pathogenic microbiota. Finally, the three core characteristics of microbial community structure shift, decline of core beneficial bacteria, and abnormal proliferation of pathogenic bacteria are integrated and refined to form a unique and highly identifiable pathological feature label for hand-foot-mouth disease. The label can cover the differences in microbial characteristics at different stages of hand-foot-mouth disease from early infection to severe disease progression, providing accurate feature basis for subsequent microbial quantification and intelligent identification.

[0027] The gut microbiota quantification module constructs a bidirectional interaction network of the gut microbiota, calculates a quantitative score for microbiota dysbiosis, and calibrates the specific identification threshold for hand-foot-mouth disease. This module includes: Microbiome Relationship Construction Module: Based on clustered microbiome groups, calculate the abundance linkage, survival dependence, and competitive inhibition relationships among microbiome groups to construct a bidirectional interaction network of gut microbiota; First, standardized data processing was performed on all bacterial clusters divided after sequencing clustering to unify the statistical standards for data such as bacterial abundance and distribution ratio, eliminating data fluctuations caused by differences in sequencing batches and individual samples, and providing a consistent data foundation for the analysis of relationships between bacterial communities. Then, two random bacterial clusters were selected one by one, and dynamic association characteristics such as synchronous changes in bacterial abundance, survival dependence, and competitive suppression were continuously tracked from multiple association dimensions such as long-term symbiosis, mutual promotion, inhibition and antagonism, and independent coexistence. Different types of interaction patterns such as positive synergy and negative inhibition between the two types of bacterial communities were sorted out. Then, using each microbial cluster as network node and the relationship between microbial communities as connecting edge, a two-way interaction network framework is built. The association type and association strength are labeled for each connecting edge, clearly presenting the direction and strength of various interactions such as symbiosis, antagonism, inhibition, and proliferation among different microbial communities, forming an intuitive and data-driven intestinal microbial community interaction network map, which fully restores the dynamic balance or imbalance between microbial communities in healthy and infected states. Finally, the initially constructed interaction network was refined and optimized. Weak associations that were formed accidentally or had poor stability were eliminated, while core interaction relationships that were stable in both healthy and infected samples and had pathological correlation value were retained. At the same time, network nodes and connecting edges were standardized and encoded to generate structured interaction network data that could be directly used for subsequent quantitative analysis, ensuring the stability, reliability and pathological correlation of the network.

[0028] Disorder numerical calculation module: quantifies the degree of imbalance in microbial interactions and the magnitude of microbial structure shift, and generates microbial disorder values; Based on the constructed bidirectional interaction network of gut microbiota, quantitative analysis is carried out from two core dimensions: interaction imbalance and structural deviation. First, the interaction network of the microbiota of the sample to be tested is compared with the interaction network of the healthy baseline in a comprehensive manner. The deviation of various interaction relationships is statistically analyzed, the scope and depth of imbalance of relationships such as synergy and antagonism between microbiota are determined, and the overall state of the disruption of microbiota interaction balance is accurately captured. Next, the overall structure of the microbial community of the test sample was compared and analyzed with the healthy baseline structure. The deviation of core characteristics such as the types of microbial composition, the distribution of dominant microbial communities, and microbial diversity was quantified to clarify the severity of the overall deviation of the microbial community from the healthy balance. Then, the quantitative results of the two dimensions of microbial interaction imbalance and microbial structure deviation were integrated according to the pathological influence weight of hand-foot-mouth disease to form a single, horizontally comparable microbial disorder value. The value directly corresponds to the severity of microbial disorder and accurately reflects the overall level of intestinal microecological abnormalities caused by hand-foot-mouth disease infection. Finally, the initially generated disorder values ​​were quality checked and error corrected. Abnormal values ​​caused by sequencing noise and trace impurities in the sample were removed, and the fluctuation range of the values ​​was calibrated to ensure that the disorder values ​​are stable and reliable, and can truly and accurately reflect the actual disorder state of the intestinal flora of the sample to be tested, providing a reliable quantitative basis for subsequent threshold calibration and clinical identification.

[0029] Threshold calibration module: Through iterative training with clinical samples, the threshold range for hand-foot-mouth disease-specific microbiota dysbiosis is calibrated and defined; First, sufficient clinical stool samples were collected from various groups, including patients with different stages of hand-foot-mouth disease and patients with common digestive tract diseases. The samples covered different ages, regions, and physical characteristics to ensure sample diversity and representativeness, eliminate data bias from a single group, and provide comprehensive clinical data support for threshold calibration. Then, the entire process of microbial sequencing, feature mining, and disorder calculation was completed for all clinical samples in sequence, generating microbial disorder values ​​for each group of samples in batches. Combined with the clinical diagnosis results, the values ​​of each group were classified, statistically analyzed, and distributed. The differences in the distribution range of disorder values ​​among healthy groups, hand-foot-mouth disease groups, and groups with common digestive tract diseases were identified, and the specific disorder characteristics of hand-foot-mouth disease were distinguished from non-specific microbial fluctuations caused by other diseases were identified. Finally, through iterative training with 320 clinical samples (including 80 healthy controls, 40 cases in the incubation period of hand-foot-mouth disease, 60 cases in the mild stage, 35 cases in the severe progressive stage, 25 cases in the recovery stage, and 80 controls with common gastrointestinal diseases), the optimal cutoff value was determined using ROC curves, and a threshold system for hand-foot-mouth disease-specific flora dysbiosis was defined: dysbiosis score <30 for healthy individuals, 30-45 for the incubation period, 46-65 for the mild stage, >65 for the severe progressive stage, and 35-55 for the recovery stage. The dysbiosis score for common gastrointestinal diseases was mostly concentrated in the 20-40 range, showing a clear distinction from the hand-foot-mouth disease threshold range (AUC=0.95, 95% CI). The threshold range of hand-foot-mouth disease infection (≥30 points) and disease course stratification was defined by 0.92-0.98. The thresholds were repeatedly cross-validated using multiple batches of independent clinical samples (an additional 100 cases were selected), and the threshold boundaries were dynamically fine-tuned (with an adjustment range not exceeding ±2 points) to ensure that the thresholds could accurately distinguish between healthy status, hand-foot-mouth disease infection status and different disease stages, while effectively excluding interference from common gastrointestinal diseases. Finally, a stable, accurate and clinically applicable threshold system for determining the disorder of the hand-foot-mouth disease flora was formed.

[0030] The intelligent identification module completes the four-level stratification of the disease course of hand-foot-mouth disease, from the incubation period to the recovery period, provides intelligent early warning of the risk of severe complications, and outputs standardized auxiliary identification results. This module includes: Disease progression assessment module: Based on the degree of bacterial flora disorder and the characteristics of bacterial flora structure, determine the incubation period, mild stage, severe progression stage, and recovery stage of hand-foot-mouth disease; Specifically, the process first retrieves the bacterial community disorder value, bacterial community structure shift amplitude, and specific disordered bacterial community distribution data calculated by the bacterial community quantification module. These data are then precisely matched with the four-level disease-specific threshold intervals defined by the judgment threshold calibration module to initially pinpoint the disease range. Simultaneously, key structural features such as the degree of decline of core beneficial bacteria, the scale of pathogenic bacteria proliferation, and changes in bacterial community diversity are retrieved to cross-validate the preliminary judgment results, eliminating misjudgments caused by fluctuations in critical values ​​and ensuring the accuracy of the judgment. It also includes the ability to accurately capture weak pathological signals under low disorder levels, targeting the characteristics of no obvious clinical symptoms, only mild dysbiosis and weak specific bacterial shift signals during the incubation period; to clearly determine the characteristics of moderate disorder, no large-scale decay of core flora and no breakage of interaction network during the mild stage; to accurately identify the core characteristics of severe disorder, near disappearance of core protective flora and complete collapse of flora interaction network during the progression of severe disease; and to match the dynamic characteristics of gradual decline in disorder, flora structure tending towards healthy baseline and gradual recovery of beneficial flora during the recovery period, so as to achieve accurate, stable and complete determination of the four-level disease course.

[0031] Severe illness risk warning module: Identifies the characteristics of near disappearance of the core protective gut microbiota and complete disruption of the microbiota interaction network, triggering a warning of severe complications; First, high-risk core indicators such as the extreme value of disorder, the abundance ratio of core protective bacteria, the integrity of the bacterial community interaction network, and the abnormal proliferation scale of pathogenic bacteria are captured in real time from the bacterial community quantification module. This allows for the precise identification of high-risk feature combinations such as the disappearance of core bacteria, the breakage of the interaction network, and the excessive disorder. Then, the severity risk grading criteria are matched: ordinary severity tendency (disorder score 65-80, bifidobacteria 0.1%-1%), high severity risk (disorder score 80-90, bifidobacteria <0.1%, network edge count reduced by 50%-70%), and extremely high risk warning (disorder score >90, bifidobacteria disappear, network edge count reduced by >70%). The warning mechanism is immediately triggered for extremely high risk signals, and the warning trigger time, core abnormal indicators, degree of abnormality of indicators and corresponding pathological correlation evidence are recorded simultaneously and completely to form an unalterable warning log, ensuring that the entire warning process is traceable and verifiable. Finally, critical illness warning information is pushed to the physician's dedicated management portal in real time, along with an interpretation of the risk level, explanation of core abnormal indicators, and clinical intervention reference suggestions. At the same time, the warning data is stored to provide reliable data support for subsequent clinical intervention guidance and system model iteration and optimization, so as to achieve early identification, early warning, and early intervention of critical illness complications.

[0032] The identification result output module integrates information on microbial community characteristics, disorder scores, and disease course determination to generate a standardized auxiliary identification report. Specifically, the report summarizes five core pieces of information about the sample: the characteristics of the microbial community structure shift, information on specific disordered microbial communities, quantitative values ​​of microbial community disorder, four-level disease course determination results, and the level of severe disease risk. It strictly follows the medical document specifications to construct a standardized auxiliary identification report framework, covering the core sections of basic sample information, testing process description, core data display, identification conclusion, and risk warning, to ensure that the report structure is clear, the information is complete, and the professional standards are met. The report combines textual descriptions with visual charts to intuitively present a comparison of healthy and tested microbiota, disorder levels, disease progression criteria, and severe risk indicators. It generates both printable paper reports and encrypted electronic reports, supporting clinical archiving, data traceability, and cross-platform viewing. The assessment conclusions are rigorously worded and well-supported, fully complying with the usage guidelines and archiving requirements for clinical auxiliary diagnosis.

[0033] To illustrate the effects achieved using this technology, we conducted the following experiment: Experimental materials Experimental Sample: A total of 320 stool samples were selected from pediatric outpatients and inpatients of a certain hospital. The samples were grouped as follows: Healthy control group: 80 healthy children aged 3-6 years with no history of hand-foot-mouth disease infection, no chronic digestive tract diseases, and no recent medication history; Hand, foot and mouth disease group: 160 cases, diagnosed by clinical symptoms, signs and traditional enterovirus nucleic acid testing, covering four stages of the disease: incubation period (40 cases), mild stage (60 cases), severe progression stage (35 cases), and recovery stage (25 cases); Control group of common digestive tract diseases: 80 children diagnosed with common diarrhea, indigestion, bacterial enteritis and other non-hand-foot-mouth disease digestive tract diseases.

[0034] All samples were aseptically collected in the morning on an empty stomach, transported to the laboratory at low temperature within 2 hours after collection, and stored at -80℃ for later use.

[0035] Main reagents: sterile phosphate buffer, fecal microbiota lysis buffer, unbiased whole genome amplification kit, high-throughput sequencing library preparation kit, sequencing quality control reagents, sterile centrifugation consumables, nucleic acid purification reagents, etc., all of which are commercially available reagents that meet microbiological testing standards.

[0036] The main instruments include: high-speed gradient centrifuge, constant temperature metal bath, gradient PCR amplification instrument, high-throughput sequencing platform, bioinformatics analysis workstation, low-temperature freezer, sterile operating table, micro-volume nucleic acid quantification instrument, etc., all of which are routine instruments in microbiology testing laboratories and are used after being calibrated.

[0037] Experimental steps Sample pretreatment: All fecal samples to be tested are processed according to the sample pretreatment module of this system, and impurity removal, bacterial amplification, and sequencing noise reduction are completed in sequence to prepare pure bacterial nucleic acid samples. The entire process strictly follows aseptic operation procedures to avoid exogenous contamination.

[0038] Microbial community sequencing and feature analysis: The pre-processed nucleic acid samples are imported into the sequencing acquisition module to complete unbiased high-throughput sequencing, autonomous sequence clustering (using the UPARSE algorithm, similarity 97%), and effective sequence screening (Q30≥80%, length 250-1200bp); then, the microbial community feature mining module is used to construct a health baseline map, use LEfSe (LDA>2.0) to screen for hand-foot-mouth disease-specific disordered microbial communities, and extract specific pathological feature tags.

[0039] Microbiome quantification and identification: Feature data is input into the microbiome quantification module to construct a bidirectional interaction network of gut microbiota, calculate the microbiome disorder degree (formula above) and match specific thresholds (30-45 points for incubation period, 46-65 points for mild symptoms, and >65 points for severe symptoms); finally, through the intelligent identification module, the hand-foot-mouth disease infection determination, disease course stratification and severe risk warning of the sample are completed, and a standardized auxiliary identification report is generated.

[0040] Control test: All samples from the same batch were simultaneously tested using the traditional enterovirus nucleic acid detection method (detecting common hand-foot-mouth disease viruses such as EV71 and Coxsackie A16, using an RT-PCR kit, with a cycle threshold Ct<35 for a positive result), and the test results, disease course assessment, and severe case warning were recorded. To clarify the methods for determining each accuracy rate, the following provisions are made: ①Infection identification accuracy: The final clinical diagnosis is the gold standard (i.e., a comprehensive assessment of clinical symptoms, signs, disease progression, and discharge diagnosis, and independent confirmation by two or more associate chief physicians). The results of this system or traditional nucleic acid testing are compared with the gold standard, and the proportion of correctly identified samples to the total number of samples is calculated. ② Accuracy of disease course stratification: The final disease course staging made by clinicians based on the onset time, symptom severity (divided into incubation period, mild period, severe progression period, and recovery period according to the hand-foot-mouth disease diagnosis and treatment guidelines) and complete medical records during hospitalization is used as the gold standard. The stratification results of this system are compared with this gold standard, and the proportion of correctly stratified samples to the total number of samples in the hand-foot-mouth disease group is calculated. Traditional nucleic acid testing methods cannot directly stratify the disease, and only use the combination of virus positivity or negativity and clinical symptoms to roughly distinguish between mild and severe cases as a reference. ③ Accuracy of severe illness early warning: The gold standard is whether the child develops clinical deterioration (defined as serious complications such as new-onset encephalitis, myocarditis, pulmonary edema, or transfer to the intensive care unit or mechanical ventilation) within 48 hours from the sampling time point. The early warning results of this system are compared with the gold standard to calculate the proportion of the number of severe illness samples with correct early warning to the actual number of severe illness samples. Traditional methods do not have early warning functions, so they are not included in the statistics. ④ Specificity: Taking the common digestive tract disease group as the subject, calculate the proportion of samples that were judged negative (non-hand-foot-mouth disease) by this system to the total number of samples in the group.

[0041] Data statistical analysis: Organize experimental data of this system and traditional detection methods, calculate core indicators such as identification accuracy, disease course stratification accuracy, severe case warning accuracy, specificity, and window period detection rate, and compare and analyze the differences between the two methods.

[0042] Experimental results Infection identification accuracy: The infection identification accuracy of 320 samples was 98.75% (316 / 320), with 100% accuracy (80 / 80) in the healthy control group, 98.75% accuracy (158 / 160) in the hand-foot-mouth disease group (2 cases misdiagnosed as common gastrointestinal diseases) in the common gastrointestinal disease group, and 97.5% accuracy (78 / 80) in the common gastrointestinal disease group (2 cases misdiagnosed as hand-foot-mouth disease incubation period due to high bacterial flora disorder). The accuracy of traditional viral nucleic acid testing was 86.25% (276 / 320), with a detection rate of only 50% (20 / 40) in the incubation period, 91.7% (55 / 60) in the mild stage, 100% (35 / 35) in the severe stage, and 44% (11 / 25) in the recovery stage. The false positive rate in the common gastrointestinal disease group was 7.5% (6 / 80).

[0043] Accuracy of disease course stratification: The accuracy of disease course stratification for 160 samples of hand-foot-mouth disease was 95.0% (152 / 160), with an accuracy of 92.5% for the incubation period (37 / 40, 3 cases misclassified as healthy or mild), 95.0% for the mild stage (57 / 60, 3 cases misclassified as incubation period or severe), 97.1% for the severe progression stage (34 / 35, 1 case misclassified as mild), and 96.0% for the recovery stage (24 / 25, 1 case misclassified as mild). Traditional detection methods cannot achieve disease course stratification and can only roughly distinguish between mild and severe cases based on clinical symptoms, with a stratification consistency of only 68.1% (109 / 160) with the gold standard.

[0044] Accuracy of severe case warning: For 35 samples in the progressive stage of severe illness (24-48 hours before significant deterioration of clinical symptoms), the early warning accuracy rate was 97.14% (34 / 35), with only 1 case failing to trigger a warning due to the bacterial flora disorder not reaching the threshold (64 points measured, threshold 65 points). The false alarm rate for non-severe samples (mild cases + recovery period) was 2.35% (2 / 85). Traditional viral nucleic acid testing can only indirectly indicate the severity of illness through elevated viral load after the onset of severe symptoms (such as signs of encephalitis), and cannot provide early warning; therefore, the accuracy rate of early warning is clinically insignificant.

[0045] Clinical specificity: It has a good exclusion effect on samples from common gastrointestinal diseases, with a specificity of 97.5% (78 / 80). It can effectively distinguish the flora imbalance caused by hand-foot-mouth disease from the flora fluctuations caused by other gastrointestinal diseases (for example, the flora imbalance score of bacterial enteritis is mostly 25-40, which partially overlaps with the incubation period threshold of 30-45 in this system, but can be distinguished by additional feature labels). The specificity of traditional viral nucleic acid detection is 92.5% (74 / 80), with a false positive rate of 7.5% (6 cases), mainly due to cross-contamination or non-specific amplification.

[0046] Experimental conclusions Compared to traditional viral nucleic acid testing, this high-throughput sequencing system for fecal microbiota in hand-foot-mouth disease improves the accuracy of infection identification by 14.5 percentage points (98.75% vs 86.25%), achieves a 95.0% accuracy rate in disease progression stratification (traditional methods cannot stratify), and has a 97.14% accuracy rate in severe case warning (traditional methods do not have this function). It effectively compensates for the shortcomings of traditional detection methods, such as missed detections during the window period, inability to finely stratify disease progression, and difficulty in early warning of severe cases. It can provide reliable technical support for early screening, clinical diagnosis and treatment, and severe case prevention and control of hand-foot-mouth disease, and has good prospects for clinical application.

[0047] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention are also within the protection scope of the present invention.

Claims

1. A hand-foot-mouth disease-assisted identification system using high-throughput sequencing of fecal microbiota, characterized in that, include: Sample preprocessing module, sequencing acquisition module, microbial community feature mining module, microbial community quantification module, intelligent identification module; The sample preprocessing module performs stratified removal of contaminating bacteria, unbiased amplification of the entire microbial community, and initial screening and noise reduction of sequencing on fecal samples to prepare pure microbial nucleic acid samples. The sequencing acquisition module performs high-throughput sequencing of the entire microbial community without relying on external microbial databases, clustering and grouping based on the sequence characteristics themselves, and screening of effective microbial community sequences to generate the original microbial community sequencing dataset. The microbial feature mining module constructs a baseline map of healthy gut microbiota, screens hand-foot-mouth disease-specific disordered microbial clusters, and extracts exclusive pathological feature tags. The microbiome quantification module constructs a bidirectional interaction network of gut microbiota, calculates the quantification score of microbiome disorder, and calibrates the specific identification threshold for hand-foot-mouth disease. The intelligent identification module completes the four-level disease progression determination of hand-foot-mouth disease from the incubation period to the recovery period, intelligent early warning of the risk of severe complications, and standardized auxiliary identification result output.

2. The hand-foot-mouth disease-assisted identification system based on high-throughput sequencing of fecal microbiota according to claim 1, characterized in that: The sample preprocessing module includes: an impurity removal module, a microbial community amplification module, and a sequencing noise reduction module; Impurity removal module: Removes food residue, exfoliated cells from the host intestinal tract, and free viral nucleic acid fragments from fecal samples layer by layer through three-stage gradient centrifugation; Microbial community amplification module: Employs whole-genome random primer amplification method to non-discriminatory amplify nucleic acids of all microorganisms in the sample; Sequencing noise reduction module: Based on the signal-to-noise ratio and fragment length consistency of the sequencing fragments, invalid sequencing fragments are removed to complete data purification.

3. The hand-foot-mouth disease-assisted identification system based on high-throughput sequencing of fecal microbiota according to claim 1, characterized in that: The sequencing acquisition module includes: a full-genome sequencing module, a sequencing clustering module, and a valid sequence screening module; Full-scale gene sequencing module: Performs unbiased high-throughput sequencing on the amplified bacterial community nucleic acids to obtain the complete bacterial community gene sequence; Sequencing clustering module: Based on sequence similarity, gene fragment conservation, and base arrangement characteristics, it autonomously clusters and groups sequencing sequences. Effective sequence screening module: Removes low-quality, short fragments and non-microbial sequences, and retains complete and effective bacterial community sequences.

4. The hand-foot-mouth disease-assisted identification system based on high-throughput sequencing of fecal microbiota according to claim 1, characterized in that: The microbial community feature mining module includes: a healthy microbial community modeling module, a disordered microbial community screening module, and a disease feature extraction module. Healthy gut microbiota modeling module: Constructs a baseline structural map of healthy gut microbiota based on clustering data from infection-free healthy samples; Disordered microbiota screening module: Compares the microbiota structure of the sample to be tested with the healthy baseline map to screen for hand-foot-mouth disease-specific disordered microbiota clusters; Disease feature extraction module: Extracts hand-foot-mouth disease-specific feature tags based on microbial community structure shift, core beneficial microbial community attenuation, and abnormal proliferation of pathogenic microbial communities.

5. The hand-foot-mouth disease-assisted identification system based on high-throughput sequencing of fecal microbiota according to claim 1, characterized in that: The microbial community quantification module includes: a microbial community relationship construction module, a disorder numerical calculation module, and a judgment threshold calibration module; Microbiome Relationship Construction Module: Based on clustered microbiome groups, calculate the abundance linkage, survival dependence, and competitive inhibition relationships among microbiome groups to construct a bidirectional interaction network of gut microbiota; Disorder numerical calculation module: quantifies the degree of imbalance in microbial interactions and the magnitude of microbial structure shift, and generates microbial disorder values; Threshold calibration module: Through iterative training with clinical samples, it calibrates and defines the threshold range for hand-foot-mouth disease-specific microbial dysbiosis.

6. The hand-foot-mouth disease-assisted identification system based on high-throughput sequencing of fecal microbiota according to claim 1, characterized in that: The intelligent identification module includes: a disease stage judgment module, a severe illness risk warning module, and an identification result output module; Disease progression assessment module: Based on the degree of bacterial flora disorder and the characteristics of bacterial flora structure, determine the incubation period, mild stage, severe progression stage, and recovery stage of hand-foot-mouth disease; Severe illness risk warning module: Identifies the characteristics of near disappearance of the core protective gut microbiota and complete disruption of the microbiota interaction network, triggering a warning of severe complications; The identification result output module integrates information on microbial community characteristics, disorder scores, and disease course determination to generate a standardized auxiliary identification report.

7. The hand-foot-mouth disease-assisted identification system based on high-throughput sequencing of fecal microbiota according to claim 2, characterized in that: The three-stage gradient centrifugation was performed at 800 r / min for 5 minutes, 3000 r / min for 10 minutes, and 12000 r / min for 15 minutes.

Images