A colorectal early cancer risk assessment system

Abstract

The application discloses a colorectal early cancer risk assessment system, aiming at solving the defects of poor adaptability of the existing technology general population benchmark, static detection lag, weak anti-interference ability and insufficient clinical guidance value. The core scheme is: based on the screening object's own time sequence flora data, an individual exclusive nonlinear steady state benchmark is constructed, the flora drift degree is adaptively weighted and quantified from the structure and function dimensions, through the three-layer collaborative anti-interference system of circadian rhythm noise purification, multi-source drift difference discrimination and population threshold adaptive calibration, the benign interference is removed, and finally the four-level dynamic risk stratification is realized through the normalized flora drift acceleration. The application realizes the non-invasive, individualized and dynamic precise evaluation of colorectal early cancer, can capture precancerous signals in advance, greatly improves the screening accuracy and specificity, and is suitable for large-scale population normalization early screening scene.

Description

Technical Field

[0001] This invention belongs to the field of smart healthcare and gut microbiota detection technology, specifically a risk assessment system for early colorectal cancer. Background Technology

[0002] In recent years, colorectal cancer has become the third leading cause of cancer incidence and the second leading cause of cancer death worldwide, with a significant trend towards younger age of onset. Clinical data shows that the prognosis of colorectal cancer is closely related to the stage of diagnosis; the 5-year survival rate for patients with early-stage colorectal cancer can reach over 90%, while the 5-year survival rate for patients with advanced-stage cancer is less than 10%. Therefore, achieving non-invasive, accurate, and routine screening for early-stage colorectal cancer and precancerous lesions is crucial to reducing colorectal cancer mortality and improving patient prognosis.

[0003] Currently, commonly used clinical methods for colorectal cancer screening mainly include four categories: colonoscopy, fecal occult blood test, fecal tumor marker detection, and gut microbiota testing. Among them, colonoscopy, as the gold standard in clinical practice, is an invasive procedure with problems such as painful bowel preparation, high examination risks, and strain on medical resources. In my country, colonoscopy screening compliance is less than 15%, making large-scale routine screening impossible. While fecal occult blood test and fecal tumor marker detection are non-invasive, their specificity is low and easily affected by factors such as diet, medication, and intestinal inflammation, with false positive rates generally exceeding 30%, potentially leading to over-treatment and unnecessary panic.

[0004] With the deepening research on the relationship between gut microbiota and tumor development, colorectal cancer screening technology based on gut microbiota biomarkers has gradually become a research hotspot. Among existing technologies, the patent with authorization announcement number CN108064273B, entitled "Biomarkers for Colorectal Cancer-Related Diseases," is one of the most representative technical solutions in this field. This patent constructs a risk prediction model for colorectal cancer and advanced adenocarcinoma by detecting the abundance differences of specific gut microbiota biomarkers in fecal samples, achieving non-invasive screening for colorectal cancer and improving screening accuracy to a certain extent compared to traditional fecal testing techniques.

[0005] However, in-depth analysis of this existing technical solution and similar microbial detection technologies reveals the following insurmountable technical shortcomings: First, using average gut microbiota data from a generally healthy population as a baseline completely ignores the natural differences in gut microbiota between individuals. The structure and function of gut microbiota are influenced by various factors such as genetic background, dietary habits, living environment, age, and physical condition, resulting in significant individual differences. A universal baseline can lead to physiological microbiota differences in some populations being misdiagnosed as pathological drift, while mild pathological drift in some high-risk groups is masked by the universal baseline, resulting in an overall assessment accuracy of less than 60%.

[0006] Secondly, static risk assessment based solely on the abundance of gut microbiota at a single time point cannot capture the dynamic evolution of the gut microbiota. The development of colorectal cancer is a gradual process that takes 5-10 years. The changes in gut microbiota during the precancerous lesion stage exhibit continuous and gradual drift characteristics. A single static test can only reflect the state of the gut microbiota at a certain moment and cannot distinguish between transient fluctuations and continuous pathological changes, which can easily lead to missed diagnoses and misdiagnoses.

[0007] Third, the impact of external interference factors on test results has not been systematically addressed. Gut microbiota is easily affected by various benign factors such as circadian rhythms, short-term dietary fluctuations, and common intestinal inflammation. Neither this patent nor existing similar technologies have established an effective anti-interference mechanism, resulting in a clinical false positive rate generally higher than 18%, which seriously affects the reliability of test results.

[0008] Fourth, it can only output static risk scores at a single time point and cannot quantify the rate of disease progression. Under the same static risk score, different rates of microbial drift correspond to completely different levels of cancer urgency and clinical intervention strategies. Static risk assessment cannot provide dynamic progression information for clinicians, making it difficult to achieve tiered and precise intervention, and thus has limited clinical guidance value.

[0009] In summary, existing early colorectal cancer screening technologies still suffer from core technical bottlenecks such as poor individualization adaptability, lag in static detection, weak anti-interference ability, and insufficient clinical guidance value. There is an urgent need to develop a colorectal cancer risk assessment technology that can achieve individualization, dynamics, high specificity, and high compliance. Summary of the Invention

[0010] The purpose of this invention is to provide a risk assessment system for early colorectal cancer. This invention eliminates benign interference through a three-layer synergistic anti-interference system, achieving non-invasive, individualized, and highly specific dynamic and accurate assessment of early colorectal cancer. It can capture precursor signals of cancer in advance, significantly improving the accuracy of screening and its clinical guidance value.

[0011] The technical solution adopted in this invention is as follows: A colorectal early cancer risk assessment system includes: The temporal microbiota acquisition module is used to acquire multiple sets of raw fecal microbiota sequencing data from different time points of the screening subjects; The individual homeostasis construction module is connected to the time-series microbiota acquisition module. It is used to construct a unique gut microbiota homeostasis benchmark that is adapted to individual characteristics based on the time-series microbiota data of the screening subject, abandoning the unified benchmark for the general population. The gut microbiota drift quantification module is used to quantify the degree of gut microbiota drift from the dimensions of microbiota structure and microbiota function, with reference to the individual-specific gut microbiota homeostasis benchmark, and to obtain microbiota temporal drift risk data. The noise purification module is used to remove interference noise caused by human physiological rhythm fluctuations in the microbial community time series data, and to purify and obtain effective signals of pathological microbial community drift. The threshold adaptive calibration module is used to dynamically match the differential microbial drift judgment threshold according to the population stratification attributes of the screening subjects, and to calibrate and correct the pathological microbial drift signal. The drift difference discrimination module is used to distinguish the homologous drift of the microbial community caused by different factors, remove benign non-cancerous interfering drift data, and obtain accurate static risk data. The dynamic risk prediction module combines microbial drift data from multiple time-series nodes to quantify the dynamic progression rate of microbial drift and output the dynamic risk level of early colorectal cancer for the screening subjects and the corresponding clinical intervention plan.

[0012] Preferably, the time-series microbiota collection module performs stratified differentiated time-series sampling and standardized clinical quality control; The screening target population is stratified into healthy young people, healthy middle-aged and elderly people, and people at high risk of underlying intestinal diseases and polyps. The sampling intervals for the three groups are configured as 3 months, 2 months and 1 month, respectively. The standardized clinical quality control includes: discontinuing antibiotics 14 days before sampling, discontinuing probiotics, PPI proton pump inhibitors, metformin and nonsteroidal anti-inflammatory drugs 7 days before sampling, adopting a light diet control mode for 3 days before sampling, and uniformly collecting the first stool sample in the morning; the sequencing data quality control standard is that the sequencing depth is not less than 5G clean data and the number of effective reads is not less than 200,000.

[0013] Preferably, the individual homeostasis construction module screens resident microbiota with a relative abundance of ≥0.1% in the sequencing data, removes transient exogenous bacteria data, and extracts the structural abundance characteristics of resident microbiota and the abundance characteristics of core metabolic pathways. The individual homeostasis construction module uses a Gaussian nonlinear fitting model to fit the extracted feature data with a nonlinear probability distribution, generates a homeostasis center value for the individual microbiome of the screening object, and constructs a 95% confidence interval for physiological fluctuations to form an individual-specific microbiome homeostasis baseline map.

[0014] Preferably, the microbial community drift quantification module is configured with a population stratification adaptive weighting mechanism and incorporates a microbial community comprehensive drift gradient coefficient algorithm; the formula for the microbial community comprehensive drift gradient coefficient algorithm is: ; Wherein, G is the microbial community comprehensive drift gradient coefficient with a value of 0~1, Gs is the normalized structural drift factor, Gf is the normalized functional drift factor, α is the structural drift weight coefficient, β is the functional drift weight coefficient, and α+β=1; the population adaptive weight configuration is as follows: healthy young people α=0.4, β=0.6, healthy middle-aged and elderly people α=0.35, β=0.65, and people with underlying intestinal diseases and high risk of polyps α=0.3, β=0.7.

[0015] Preferably, the formula for calculating the normalized structure drift factor Gs is: ; Where n is the number of species in the core resident microbial community of an individual, Si is the relative abundance of the i-th microbial community currently detected, S0i is the standard abundance of the i-th microbial community under the steady-state baseline of the individual, ΔT is the time interval between two samplings in months, and the calculation results are truncated to the 0-1 interval after being normalized to the maximum and minimum values; the formula for calculating the normalized functional drift factor Gf is: ; Where m is fixed at 4, corresponding to the four core KEGG metabolic pathways associated with colorectal cancer mutations: short-chain fatty acid synthesis pathway, bile acid metabolism pathway, intestinal toxin metabolism pathway, and intestinal inflammation regulation metabolic pathway. Fj is the abundance of the j-th metabolic pathway currently detected, and F0j is the abundance of the j-th metabolic pathway under the individual homeostatic baseline. The calculation results are truncated to the 0~1 interval after being normalized to the maximum and minimum values.

[0016] Preferably, the noise purification module is pre-trained with a 24-hour four-segment circadian rhythm baseline model, which divides the 24 hours into four time periods: 0:00-6:00 AM, 6:00-12:00 PM, 12:00-6:00 PM, and 6:00-12:00 AM. The noise purification module identifies periodic small fluctuations in bacterial species abundance with an amplitude of ≤8% and metabolic pathway abundance with an amplitude of ≤10% as physiological noise and removes them, while retaining pure pathological bacterial community drift signals that are non-periodic and continuously shift in one direction.

[0017] Preferably, the threshold adaptive calibration module has a built-in quantization drift threshold library for three types of populations; Healthy young adults: structural drift threshold 0.20, functional drift threshold 0.35; Healthy middle-aged and elderly adults: structural drift threshold 0.18, functional drift threshold 0.32; High-risk groups with underlying intestinal diseases and polyps: structural drift threshold 0.15, functional drift threshold 0.28; Among them, for people with underlying intestinal diseases and those at high risk of polyps, active polyps are distinguished from inactive polyps, and the threshold for inactive polyps is increased by 10% overall.

[0018] Preferably, the drift differential discrimination module combines the duration of microbial drift, the fluctuation slope, and the detection threshold of fecal calprotectin to achieve accurate differentiation of multi-source drift; Drifts that fluctuate continuously for 10-20 days, have a gentle slope, and can spontaneously decline are identified as dietary-induced physiological drifts and are excluded. Drifts with sustained fluctuations for ≥20 days, calprotectin levels ≥50 μg / g, and no sustained worsening slope were classified as benign inflammatory drifts and excluded. Drifts that fluctuate continuously for ≥25 days, have a continuously increasing slope, show no spontaneous decline trend, and have a calprotectin detection value <50μg / g are identified as pathological drifts that are precursors to cancer, and a precise static microbiota drift risk value is output.

[0019] Preferably, the dynamic risk prediction module has a built-in normalized drift accelerometer. Calculation formula and four-level clinical risk stratification mechanism; The formula for calculating the normalized drift acceleration is: ; Where A is the normalized monthly acceleration of the microbial community drift, Gt is the comprehensive drift gradient coefficient of the current period, Gt-1 is the comprehensive drift gradient coefficient of the previous period, ΔT is the sampling interval in months, GO is the initial coefficient of the individual steady-state baseline, and the acceleration A is forcibly truncated in the range of [-0.1, 0.1]; The four-level risk stratification standard is: A≤0.02 corresponds to low risk, Corresponding to medium risk High risk, each risk level is matched with a corresponding annual screening, 6-month follow-up, 3-month intervention follow-up, and immediate colonoscopy detailed examination plan.

[0020] A method for assessing the risk of early colorectal cancer, applied to the aforementioned early colorectal cancer risk assessment system, includes the following steps: S1. Match the corresponding sampling interval according to the stratification type of the screening target population, complete the standardized quality control of pre-medication, diet and sampling time, collect multiple sets of morning fecal samples, and obtain standardized time-series microbial sequencing data through metagenomic sequencing and quality control filtering. S2. Screen the resident microbial community characteristic data in the sequencing data, remove exogenous bacteria, and construct an individual microbial community homeostasis benchmark map specific to the screened object through a Gaussian nonlinear fitting model; S3. Match the adaptive weight coefficients of the population, calculate the structural drift factor and the functional drift factor respectively, and obtain the weighted comprehensive drift gradient coefficient of the microbiome to obtain the initial time-series drift risk value. S4. Combine the diurnal rhythm baseline model to remove physiological rhythm noise and purify to obtain pure pathological microbial community drift signal; S5. Match a specific drift threshold based on the population attributes of the screening target, and adaptively calibrate and correct pathological drift signals; S6. Differentiate multi-source homology drift by drift duration, fluctuation slope and calprotectin threshold, remove benign interference data and obtain accurate static risk value; S7. Calculate the normalized monthly acceleration of microbial community drift, determine the dynamic risk level by combining the four-level risk stratification standard, and output a matching clinical screening and intervention plan.

[0021] The beneficial effects of this invention are as follows: 1. Most existing technologies default to using average gut microbiota data from healthy individuals as a unified control benchmark. The industry has long believed that a universal benchmark can meet screening needs, failing to recognize the inherent differences in gut microbiota between individuals due to factors such as genetics, diet, and living environment, which can lead to systematic assessment bias. This invention abandons the industry-standard group control logic and constructs a nonlinear gut microbiota steady-state benchmark based on the time-series data of the screening subjects. It uses the individual's own health status as a reference to identify abnormal changes, significantly eliminating the inherent defects of universal benchmarks that "normal individuals are misjudged as abnormal, and abnormal individuals are masked as normal," achieving truly personalized and accurate assessment.

[0022] 2. Existing microbial screening technologies are limited to detecting the structure and abundance of microbial communities at a single time point, reflecting only the microbial community state at a particular moment. They cannot distinguish between transient fluctuations and persistent pathological changes, nor can they predict the progression trend of lesions. This invention not only simultaneously quantifies the drift of microbial community structure and function in two dimensions, but also innovatively introduces the concept of dynamic acceleration into the field of microbial ecological risk assessment. By normalizing the drift acceleration, it quantifies the progression rate of microbial community abnormalities, achieving a fundamental leap from "static risk assessment" to "dynamic trend prediction." It can identify rapidly progressing precancerous lesions in advance, even before static indicators show significant abnormalities.

[0023] 3. Existing technologies generally suffer from weak anti-interference capabilities. Comparative studies and similar solutions have not established a complete interference removal mechanism, leading to misjudgments of gut microbiota caused by benign factors such as circadian rhythms, dietary fluctuations, and common inflammation as cancer risk, resulting in persistently high false positive rates in clinical practice. This invention employs a three-layer synergistic mechanism: "circadian rhythm noise purification - multi-source drift differential discrimination - individualized threshold calibration," systematically removing benign interference from three dimensions: physiological rhythm, causal tracing, and population differences, rather than simply adding a single anti-interference method. This effectively addresses the core pain point of insufficient specificity that has long plagued early gut microbiota screening.

[0024] 4. Existing colorectal cancer screening systems exhibit significant polarization. Invasive colonoscopy suffers from extremely low compliance, hindering large-scale screening, while non-invasive testing employs a uniform sampling model, resulting in resource misallocation due to insufficient monitoring of high-risk groups and over-screening of low-risk groups. This invention utilizes a non-invasive fecal sampling method with differentiated sampling intervals for different risk groups. This ensures high-frequency dynamic monitoring of high-risk groups while avoiding the waste of medical resources for low-risk groups, achieving routine and large-scale dynamic screening for early colorectal cancer and filling a gap in the existing screening system. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the architecture of the colorectal early cancer risk assessment system of the present invention; Figure 2 This is a flowchart illustrating the method for assessing the risk of early colorectal cancer according to the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0027] See Figure 1 and 2 A colorectal early cancer risk assessment system, comprising: This invention aims to address the technical deficiencies in existing colorectal cancer early risk assessment technologies. Current clinical colorectal cancer screening methods primarily rely on colonoscopy imaging, single fecal tumor marker detection, and general population gut microbiota control assessment models, which have several limitations in practical application. Firstly, colonoscopy is an invasive procedure with low patient acceptance and poor screening compliance, making it impossible to achieve high-frequency, routine, and time-series monitoring, and difficult to capture the dynamic evolution of the gut microbiota. Secondly, traditional fecal tumor marker detection has low specificity and is easily affected by external factors such as diet, intestinal inflammation, and circadian rhythms, resulting in a high probability of false positives and false negatives, and failing to accurately distinguish between physiological fluctuations and pathological precancerous changes. Thirdly, existing microbiota assessment schemes all use gut microbiota homeostasis data from a generally healthy population as a control benchmark, ignoring the differences between individuals and different population groups. The individual differences in gut microbiota mean that there are significant differences in the baseline structure and metabolic function of gut microbiota among healthy young people, healthy middle-aged and elderly people, and people at high risk of underlying intestinal diseases and polyps. Using a uniform benchmark for drift assessment will lead to technical biases such as lenient assessment thresholds for high-risk groups and over-screening of low-risk groups, which will significantly reduce the accuracy and clinical suitability of the assessment. Fourth, the existing technology can only achieve static risk assessment at a single time point and cannot quantify the dynamic drift rate and progression trend of gut microbiota, making it difficult to achieve prospective prediction and graded intervention for early colorectal cancer.

[0028] To overcome the shortcomings of the existing technologies, this invention abandons the universal steady-state benchmark for the general population and constructs an individual gut microbiota steady-state map specific to the screening subjects. It uses adaptive weighted quantification of microbiota drift through structural and functional dimensions, combined with circadian rhythm noise purification, population-differentiated threshold calibration, and multi-source drift type differential discrimination technology to eliminate various benign interference factors. Finally, it achieves dynamic risk stratification prediction through microbiota drift acceleration, realizing non-invasive, high-frequency, accurate, and individualized early colorectal cancer risk screening throughout the process, filling the technical gap of individualized time-series dynamic assessment in the existing technology.

[0029] The overall technical architecture of this invention comprises eight layers: a hardware acquisition layer, a data quality control layer, a model construction layer, a feature quantification layer, a noise purification layer, a threshold calibration layer, a differential discrimination layer, and a risk prediction layer. The functional modules at each layer operate collaboratively, forming a closed-loop assessment system covering the entire process from sample collection, data processing, feature analysis to risk output and clinical intervention. The core functional modules of the system include a time-series microbiota acquisition module, an individual homeostasis construction module, a microbiota drift quantification module, a noise purification module, a threshold adaptive calibration module, a drift differential discrimination module, and a dynamic risk prediction module. These modules are interconnected and share data, working together to achieve accurate risk assessment of early colorectal cancer.

[0030] First, this invention utilizes a time-series microbiota collection module to standardize and quality-control the collection of raw gut microbiota data from the screening subjects. This module forms the data foundation of the entire evaluation system; the standardization of data collection and the rigor of quality control directly determine the accuracy of subsequent microbiota homeostasis construction and drift assessment. Unlike traditional uniform sampling methods, this invention innovatively employs a stratified, differentiated time-series sampling mechanism. Based on the clinical incidence patterns of colorectal lesions and the stability characteristics of the gut microbiota, the screening subjects are divided into three core population groups: healthy young adults, healthy middle-aged and elderly individuals, and individuals at high risk of underlying intestinal diseases and polyps. Sampling intervals are configured differently to adapt to the microbiota fluctuation characteristics and lesion risk levels of different population groups from the outset.

[0031] The specific criteria and configuration standards for stratified sampling are as follows: Young healthy individuals (18-40 years old, no underlying intestinal diseases, no family history of cancer, and no history of polyps) have good gut microbiota homeostasis, with a high proportion of short-term physiological fluctuations and an extremely low probability of pathological malignant transformation. Therefore, a 3-month sampling interval is configured to avoid resource waste caused by high-frequency sampling and effectively capture annual trends in gut microbiota dynamics. Middle-aged and elderly healthy individuals (41-65 years old, no organic intestinal lesions, and no history of colorectal polyps) experience changes in gut mucosa with age. As the barrier function gradually declines and the stability of the gut microbiota decreases, the risk of colorectal lesions gradually increases. Therefore, the sampling interval is shortened to 2 months to improve the continuity of time-series data. Individuals at high risk of underlying intestinal diseases and polyps (any age group, with underlying intestinal diseases such as chronic enteritis, ulcerative colitis, and colorectal polyps) are at high risk of early colorectal cancer. Their gut microbiota is prone to pathological drift and progresses rapidly. Therefore, a high-frequency sampling interval of 1 month is configured to achieve real-time monitoring of gut microbiota dynamics and timely capture of precancerous signals.

[0032] To completely avoid interference from external human, dietary, and drug factors on gut microbiota sequencing data, this invention has established a rigorous and practical standardized clinical quality control system. All screening subjects must strictly adhere to pre-screening quality control standards before sample collection. These standards are based on general standards for clinical microbial sequencing experiments and research findings on gut microbiota stability, and have sufficient medical experimental basis. Specific quality control requirements are as follows: All antibiotics must be strictly discontinued 14 days prior to sampling. Antibiotics indiscriminately kill both beneficial and harmful bacteria in the gut, completely destroying the original structure of the gut microbiota and causing sequencing data distortion. A 14-day self-repair period is necessary to ensure the microbiota returns to its natural homeostasis. Probiotics, proton pump inhibitors (PPIs), metformin, and nonsteroidal anti-inflammatory drugs (NSAIDs) must be discontinued 7 days prior to sampling. These drugs directly regulate gut microbiota metabolism, alter the gut acid-base environment and inflammation levels, significantly interfering with the detection of microbiota functional characteristics. A 7-day discontinuation period can completely eliminate the impact of drug residues. For 3 days prior to sampling, the screening subjects must adopt a light diet, avoiding high-fat, high-sugar, spicy, and raw / cold foods. High-fat diets promote bile acid metabolism disorders, and high-sugar diets induce the proliferation of harmful gut bacteria. A light diet can minimize temporary fluctuations in the gut microbiota caused by dietary factors.

[0033] In terms of sample collection and sequencing data quality control, this invention uniformly collects the first stool sample of the screening subjects in the morning. Morning stool, after accumulating through intestinal metabolism overnight, best reflects individual homeostatic characteristics in terms of microbiota structure and metabolic state, avoiding interference from microbiota fluctuations caused by daytime eating and activity. The sequencing process employs metagenomic high-throughput sequencing technology, with uniform sequencing quality control standards requiring a sequencing depth of at least 5G clean data and at least 200,000 effective reads. These parameters are clinically accepted and effective standards for intestinal metagenomic sequencing, ensuring the completeness and accuracy of microbiota species detection and avoiding missed or false detections of microbiota characteristics due to insufficient sequencing depth or low effective data volume. The final output is standardized, multi-time-node fecal microbiota sequencing raw data that can be used for subsequent analysis.

[0034] After completing the standardized gut microbiota data collection, the individual homeostasis construction module constructs a specific gut microbiota homeostasis baseline for the screening subject. This module is the core key that distinguishes this invention from traditional general baseline assessment technologies, completely solving the industry pain points of poor adaptability and large assessment bias of general population baselines. Traditional microbiota risk assessment technologies all use the average microbiota data of a large sample of healthy people as the homeostasis control standard, ignoring the individual differences in microbiota caused by individual genes, dietary habits, gut environment, age and physical condition. This leads to physiological microbiota differences in some populations being misjudged as pathological drift, and slight pathological drift in some high-risk populations being masked by general baselines, resulting in an assessment accuracy of less than 60%. This invention constructs a specific homeostasis baseline based on individual time-series sequencing data, using the screening subject's own historical microbiota state as a reference, accurately identifying individual-specific microbiota changes, and significantly improving assessment accuracy.

[0035] In the specific implementation process, the individual homeostasis construction module first performs data screening and impurity removal on the standardized sequencing raw data output by the time-series microbiota acquisition module. Since fecal samples contain a large number of transient exogenous bacteria, which originate from the external environment and ingested food, and are not part of the individual's resident gut microbiota, nor do they participate in the individual's homeostatic metabolic processes, their inclusion in the analysis would severely interfere with the accuracy of the homeostasis baseline. Therefore, this invention sets a screening threshold, retaining only bacterial species with a relative abundance ≥0.1% in the sequencing data as the individual's resident microbiota, and completely removing transient exogenous bacteria data with a relative abundance below this threshold. This threshold has been validated through a large number of clinical samples and can accurately distinguish between resident homeostatic microbiota and transient bacteria.

[0036] After obtaining the individual core resident microbiota dataset, the module simultaneously extracts two types of core feature data: resident microbiota structure abundance features and core metabolic pathway abundance features. The microbiota structure abundance features refer to the relative abundance percentage of each core resident microbiota species, reflecting the species composition structure of the gut microbiota. The core metabolic pathway abundance features specifically refer to the abundance of four KEGG core metabolic pathways highly associated with colorectal cancer, including short-chain fatty acid synthesis pathways, bile acid metabolism pathways, intestinal toxin metabolism pathways, and intestinal inflammation regulation pathways. These four metabolic pathways directly participate in intestinal mucosal repair, inflammation regulation, toxin degradation, and cell proliferation regulation, and are core functional targets for the occurrence and progression of early colorectal cancer. Selecting these four pathways as functional features provides sufficient basis for tumor molecular mechanism research.

[0037] After feature data extraction, this invention employs a Gaussian nonlinear fitting model to perform nonlinear probability distribution fitting on structural abundance features and metabolic pathway abundance features. Since the physiological fluctuations of the human gut microbiota are not linearly predictable, traditional linear fitting models cannot adapt to the nonlinear fluctuation characteristics of the microbiota. However, the Gaussian nonlinear fitting model can accurately fit the normal distribution of biological feature data, adapting to the physiological fluctuation characteristics of the gut microbiota. After model fitting, a microbiota homeostasis center value specific to the screening subject is generated. This center value represents the optimal homeostasis feature value of the individual's gut microbiota structure and function, representing the baseline state of the microbiota in a healthy state for the screening subject. Simultaneously, a 95% confidence interval for physiological fluctuations is constructed based on the fitted data. This confidence interval is a statistically common range of biological data fluctuations, fully covering the normal physiological fluctuation range of the individual's microbiota. Microbiota fluctuations within the interval are considered normal physiological changes, while deviations outside the interval are considered abnormal drift changes. Finally, an individual-specific microbiota homeostasis baseline map containing both structural and functional dimensions is generated, providing a precise reference standard for subsequent microbiota drift quantification.

[0038] Based on the constructed individual-specific gut microbiota homeostasis baseline map, the microbiota drift quantification module performs refined quantitative calculations of the degree of microbiota drift. This module innovatively sets up a population-stratified adaptive weighting mechanism to quantify microbiota drift from both structural and functional dimensions, solving the problems of one-sidedness and low accuracy in traditional single-dimensional quantitative assessments. Both structural disorder and functional abnormality of the gut microbiota are important causes of colorectal cancer. Structural disorder represents species imbalance in the microbiota, while functional abnormality represents disruption of intestinal metabolic homeostasis. Moreover, the weights of cancer-causing factors differ significantly among different populations. Younger populations have stronger microbiota structural stability, and functional metabolic abnormalities have a greater impact on cancer. The risk of disease caused by microbiota structural imbalance is significantly increased in the elderly and high-risk groups. Therefore, differentiated configuration of weighting coefficients can greatly improve the scientific nature of the assessment.

[0039] The core calculation logic of this module is the comprehensive microbial community drift gradient coefficient, which is used to uniformly quantify the overall drift degree of the microbial community. The coefficient ranges from 0 to 1, with a larger value indicating a higher degree of microbial community drift and a higher risk of cancer. The specific calculation formula is as follows: ; Wherein, G is the comprehensive gut microbiota drift gradient coefficient, used to comprehensively characterize the overall abnormal shift in the structure and function of the gut microbiota of the screening subjects, with a fixed value range of 0~1, where 0 represents that the microbiota is completely in a steady state with no abnormal drift, and 1 represents that the microbiota has completely deviated from the steady state and has serious pathological abnormalities; Gs is the normalized structural drift factor, used to quantify the abnormal shift in the species structure dimension of the microbiota; G is the normalized functional drift factor, used to quantify the abnormal shift in the metabolic function dimension of the microbiota; a is the structural drift weight coefficient, used to characterize the contribution ratio of structural drift to the overall carcinogenesis risk; β is the functional drift weight coefficient, used to characterize the contribution ratio of functional drift to the overall carcinogenesis risk, and the condition of a+B=159 renormalization is satisfied to ensure that the contribution ratio of the two dimensions fully covers the overall risk.

[0040] Based on extensive clinical controlled trial data, this invention configures differentiated adaptive weight coefficients for three screening populations, fully adapting to the differences in pathogenic mechanisms among different populations: In healthy young adults, the gut microbiota species structure is stable, and carcinogenesis is mainly induced by metabolic dysfunction; therefore, the structural drift weight is reduced and the functional drift weight is increased, with a=0.4 and β=0.6. In healthy middle-aged and elderly individuals, the intestinal structural stability is slightly reduced, and structural imbalance and functional dysfunction synergistically cause disease; therefore, the weight ratio is finely adjusted, with a=0.35 and β=0.65. In individuals with underlying intestinal diseases and those at high risk of polyps, the intestinal mucosa has underlying damage, and microbiota imbalance is a core factor in inducing precancerous lesions; therefore, the functional weight is further tilted, with a=0.3 and β=0.7. This weight ratio has been optimized through training with thousands of clinical samples to maximize the matching of pathological characteristics in different populations.

[0041] To achieve accurate quantification of the two-dimensional drift factor, this invention constructs dedicated calculation formulas for the normalized structure drift factor and the normalized functional drift factor, respectively. All calculation results are normalized to their maximum and minimum values ​​and truncated to the 0-1 range to ensure data dimensionality consistency and horizontal comparability. The calculation formula for the normalized structure drift factor Gs is as follows: ; The core mechanism of this formula is to statistically analyze the relative abundance shift rate of all core resident microbial communities of an individual, and combine this with sampling time interval normalization to quantify the degree of microbial community structure drift per unit time. The specific interpretations of each parameter are as follows: n is the number of species in the core resident microbial community of an individual, which is determined by the total number of resident microbial communities with a relative abundance of 20.196 obtained by the steady-state construction module; Si is the relative abundance of the i-th core resident microbial community detected at the current time node; SOi is the standard steady-state abundance of the i-th core resident microbial community stored in the individual's dedicated steady-state baseline map; T is the time interval between two time series samplings, in months, used to eliminate data bias caused by different sampling periods and achieve unified quantification of time series drift rate; the overall formula obtains the average offset rate of a single species through mean calculation, and obtains the standardized structural drift degree after time normalization, and finally normalizes and truncates to the 0~1 interval.

[0042] Normalized functional drift factor G focuses on four core metabolic pathways in carcinogenesis to quantify the degree of abnormal shift in intestinal metabolic function. The calculation formula is as follows: ; The parameters are specifically defined as follows: m is the number of core metabolic pathways, which is fixed at 4 in this invention, corresponding to four types of core pathways associated with colorectal cancer: short-chain fatty acid synthesis pathway, bile acid metabolism pathway, intestinal toxin metabolism pathway, and intestinal inflammation regulation metabolic pathway; FOj is the abundance value of the j-th core metabolic pathway detected at the current time node; FOj is the steady-state standard abundance value of the j-th type of core metabolic pathway stored in the individual-specific steady-state baseline map; the formula quantifies the degree of abnormal drift of the overall intestinal metabolic function by calculating the average relative offset rate of the four types of core metabolic pathways. The calculation results are also normalized to the maximum and minimum values ​​and truncated to the 0~1 interval to ensure that they are consistent with the dimension of the structural drift factor and can be weighted and fused.

[0043] After obtaining the comprehensive gut microbiota drift gradient coefficient through the weighted calculations described above, the initial temporal drift risk data for the screening subjects can be obtained. However, this initial data still contains noise from the human body's physiological diurnal rhythm fluctuations. This noise needs to be removed through a noise purification module to obtain pure pathological drift signals. The human gut microbiota exhibits natural diurnal periodic fluctuations, influenced by eating, rest, and intestinal peristalsis rhythms. This is a normal physiological phenomenon and not a pathological precursor to cancer. Failure to remove this type of noise will lead to a significant increase in the false positive rate of risk assessment. Therefore, noise purification in this module is a crucial step in improving the accuracy of the assessment.

[0044] This module pre-trains a 24-hour four-segment circadian rhythm baseline model of the gut microbiota using a large amount of time-series monitoring data. Based on the rhythm of human intestinal metabolism, the 24 hours are divided into four independent time periods: 0:00-6:00 AM, 6:00-12:00 PM, 12:00-6:00 PM, and 6:00-12:00 AM. For each time period, the physiological fluctuation baseline range of microbial species abundance and metabolic pathway abundance is independently trained to accurately adapt to the microbial fluctuation characteristics of different time periods.

[0045] The model's judgment mechanism and purification rules are based on clinical gut microbiota rhythm research findings: Under normal physiological conditions, the diurnal rhythm fluctuation range of the gut microbiota is limited; fluctuations exceeding this range are considered abnormal changes caused by external factors. This invention sets a precise noise judgment threshold, uniformly classifying periodic small-amplitude fluctuations in microbiota species abundance ≤8% and metabolic pathway abundance ≤10% as physiological rhythm noise and completely eliminating them. Conversely, microbiota changes without a fixed period, exhibiting unidirectional continuous shift characteristics, and exceeding the above thresholds are judged as non-physiological abnormal changes and retained as valid signals of purely pathological microbiota drift. This threshold has been validated by time-series monitoring data from multiple groups of healthy individuals, achieving 100% coverage of normal physiological fluctuation ranges, accurately distinguishing rhythm noise from pathological abnormalities, and completely eliminating the interference of diurnal rhythms on the evaluation results.

[0046] After noise purification, a secondary calibration and correction of pathological drift signals is required through a threshold adaptive calibration module to further eliminate assessment biases caused by individual differences in the population. Because there are significant differences in intestinal tolerance and gut microbiota homeostasis resilience among different population groups, a uniform drift judgment threshold cannot be adapted to the entire population. Young, healthy individuals have strong gut microbiota homeostasis resilience, and minor drifts can self-repair, while high-risk individuals have fragile gut microbiota homeostasis, and even small drifts can induce lesions. Therefore, this module incorporates a quantitative drift threshold library specifically for three population groups to achieve adaptive and accurate threshold matching.

[0047] Specific population-specific threshold standards have been optimized and calibrated using clinical samples, with the following configurations: Young, healthy individuals exhibit the best homeostatic resilience and the widest tolerance range for physiological drift, with a structural drift threshold of 0.20 and a functional drift threshold of 0.35; middle-aged and elderly healthy individuals experience a slight decline in intestinal function, resulting in a slightly lower tolerance threshold, with a structural drift threshold of 0.18 and a functional drift threshold of 0.32; individuals with underlying intestinal diseases and those at high risk for polyps have underlying intestinal damage, making their gut microbiota homeostasis highly susceptible to imbalance, resulting in the lowest tolerance threshold, with a structural drift threshold of 0.15 and a functional drift threshold of 0.28. Furthermore, high-risk individuals are finely stratified, distinguishing between active and inactive polyps. Individuals with inactive polyps have a relatively lower risk of disease, therefore their thresholds are increased by 10% to avoid over-assessment. Individuals with active polyps are directly assessed using the baseline high-risk thresholds to ensure no high-risk signals are missed. The system can automatically match the corresponding thresholds to complete signal calibration based on the age, intestinal history, and polyp activity status entered by the screening subjects, correcting for population-specific assessment biases.

[0048] Even after threshold calibration, pathological drift signals still contain drift interference caused by benign factors such as diet and inflammation. A drift differential discrimination module is needed to accurately distinguish and eliminate multi-source homologous drift, further purifying the pathological drift data related to precancerous lesions. Clinical studies have shown that gut microbiota drift has multiple causes. Besides precancerous lesions, short-term dietary changes and common intestinal inflammation can trigger microbiota drift. Such benign drift does not require tumor risk intervention; however, if it cannot be accurately distinguished, it can lead to misjudgments in clinical intervention plans and increase the screening burden. This module innovatively combines three core indicators—duration of microbiota drift, fluctuation slope, and fecal calprotectin detection—to construct a multi-dimensional differential discrimination system, achieving accurate classification of drift caused by different factors.

[0049] The specific criteria and clinical basis are as follows: The first category is dietary-induced physiological drift, which is induced by short-term high-fat, spicy, or binge-eating dietary behaviors. It is characterized by short fluctuation periods, gentle slopes, and spontaneous recovery. This invention classifies gut microbiota drift that lasts 10-20 days, has a gentle slope, and spontaneously returns to baseline without intervention as dietary-induced physiological drift and completely excludes it. This type of drift poses no risk of disease and is not included in the risk assessment. The second category is inflammatory-induced benign drift, induced by common intestinal inflammation or mild intestinal microbiota infection, accompanied by elevated intestinal inflammatory markers. This invention uses the clinically accepted inflammatory threshold of 50 μg / g as the calprotectin criterion. Drifts that last ≥20 days are classified as dietary-induced physiological drift. The first type is benign inflammatory drift with a calprotectin level ≥50 μg / g and no continuously worsening slope. This type of drift is a temporary abnormality in the gut microbiota caused by inflammation, and the microbiota can return to homeostasis after the inflammation subsides. It is not a precursor signal of cancer. The third type is pathological drift that is a precursor to cancer. It is characterized by continuous progression, no spontaneous decline, and no obvious inflammatory response. This invention identifies drifts with continuous fluctuations ≥25 days, continuously increasing fluctuation slopes, no spontaneous decline trend, and calprotectin levels <50 μg / g as pathological drifts that are precursors to cancer. This type of drift is a specific precursor signal for the occurrence of early colorectal cancer. Finally, based on this type of effective signal, a precise static gut microbiota drift risk value is output to complete the accurate determination of static risk.

[0050] Based on accurate static risk data, the dynamic risk prediction module quantifies the dynamic progression rate of gut microbiota drift and stratifies clinical risk, achieving an upgrade from static numerical judgment to dynamic trend prediction. The occurrence of early colorectal cancer is a dynamic process of continuous deterioration and gradual drift of the gut microbiota. A single static numerical value cannot reflect the rate of disease progression. This module calculates the gut microbiota drift acceleration through multi-time-series drift data, quantifies the rate of progression of abnormal gut microbiota, and, combined with a four-level clinical risk stratification mechanism, outputs precise clinical intervention plans adapted to different risk levels.

[0051] The core formula for calculating normalized drift acceleration in this module is as follows: ; The specific definitions and technical basis of each parameter are as follows: A is the normalized monthly acceleration of gut microbiota drift, which is mainly used to characterize the monthly progression rate of pathological abnormalities in the gut microbiota. The positive and negative values ​​represent the drift trend. Positive values ​​represent continuous deterioration of the microbiota and increased risk, while negative values ​​represent gradual recovery of the microbiota and reduced risk. Gt is the comprehensive drift gradient coefficient of the microbiota detected in the current assessment period. Gt-1 is the comprehensive drift gradient coefficient of the microbiota in the previous adjacent assessment period. ΔT is the sampling time interval between the two assessment periods, in months, to ensure that the acceleration is a unified monthly dimension indicator. GO is the initial coefficient of the steady-state baseline of the individual microbiota of the screened object, which is used to normalize and eliminate individual baseline differences, and to achieve horizontal comparison of acceleration between different individuals and different populations. At the same time, in order to avoid interference from extreme abnormal data, this invention performs forced truncation processing on the acceleration A in the range of [-0.1, 0.1]. Extreme values ​​outside the range are uniformly corrected to the critical value of the range to ensure data stability and assessment reliability.

[0052] Based on the numerical range of normalized drift acceleration A, and combined with follow-up validation of thousands of clinical screening data, a four-level clinical risk stratification standard was constructed. Each level is precisely matched with a corresponding clinical intervention plan to achieve graded precision diagnosis and treatment. The specific stratification and intervention rules are as follows: Low risk level: A≤0.02. This range represents that the gut microbiota drift tends to be stable, there is no continuous deterioration trend, the gut homeostasis is basically normal, the risk of early colorectal cancer is extremely low, and it corresponds to the routine annual clinical screening program without the need for additional intervention. Medium risk level: 0.02 < A ≤ 0.05. This range indicates that there is a small and persistent pathological drift in the gut microbiota, with slight disruption of homeostasis and potential risk of disease. Monitoring frequency needs to be increased, and a special review plan should be implemented every 6 months to dynamically track the trend of changes in the gut microbiota. High risk level: 0.05 < A ≤ 0.08. This range represents a continuous pathological deterioration and drift of the gut microbiota, significant disruption of intestinal homeostasis, and a high risk of precancerous lesions. Timely intervention is required to block the progression of the disease. The corresponding treatment is a short-term follow-up examination every 3 months and a special gut microbiota intervention program, which uses diet and probiotics to restore gut microbiota homeostasis. Extremely high risk level: A>0.08. This range indicates that the microbiota is undergoing rapid pathological drift and is on the verge of collapse. The probability of early colorectal cancer is extremely high, and immediate clinical diagnosis and screening are required. The corresponding immediate colonoscopy intervention plan can achieve early detection, early diagnosis and early treatment.

[0053] Based on the hardware logic and data processing principles of the aforementioned system modules, this invention simultaneously provides a fully implementable method for assessing the risk of early colorectal cancer. This method is fully compatible with the aforementioned assessment system, with coherent steps and a closed-loop logic. Those skilled in the art can completely replicate and implement it based on the content disclosed in this embodiment. The specific execution steps are as follows: Step 1: Stratified Sampling and Standardized Data Collection. First, the population is stratified based on age, intestinal history, and polyp status, and corresponding differentiated sampling intervals are matched: 3 months for healthy young adults, 2 months for healthy middle-aged and elderly individuals, and 1 month for high-risk individuals. Strict pre-sampling quality control protocols are followed: antibiotics are discontinued 14 days prior to sampling, related interfering drugs are discontinued 7 days prior, and a light diet is maintained for 3 days prior. First stool samples are collected upon waking each morning, completing sample collection across multiple time points. Metagenomic high-throughput sequencing is performed on the collected samples. Valid data with at least 5G clean data and no less than 200,000 effective reads are selected, and sequencing impurities and contaminating data are removed to obtain standardized time-series microbiota sequencing data.

[0054] Step 2: Construction of Individual-Specific Microbial Community Homeostasis Benchmark Map. Standardized sequencing data were screened, retaining core resident microbiota with a relative abundance ≥0.1%, while removing exogenous transient contaminant data. Structural abundance features of the core resident microbiota and abundance features of core metabolic pathways associated with four types of carcinogenesis were extracted. A Gaussian nonlinear fitting model was used to fit the nonlinear probability distribution of these two types of feature data to generate individual microbial community homeostasis center values. A 95% confidence interval for physiological fluctuations was constructed, ultimately forming an individual-specific microbial community homeostasis benchmark map adapted to the individual characteristics of the screening subject, serving as a control benchmark for subsequent drift assessment.

[0055] Step 3: Quantitative Calculation of Two-Dimensional Microbiome Drift. Based on the stratification type of the screened population, the corresponding adaptive structural and functional weight coefficients are matched and substituted into the calculation formulas for the normalized structural drift factor and the normalized functional drift factor, respectively, to quantify the two-dimensional drift values. The comprehensive microbiome drift gradient coefficient is then calculated using a weighted fusion formula to obtain the initial time-series drift risk value of the screened population, thus completing the preliminary quantification of the degree of microbiome drift.

[0056] Step 4: Circadian rhythm noise purification. The time periods of the time-series microbial community drift data are matched one by one with the 24-hour four-segment circadian rhythm baseline model. The physiological fluctuation thresholds of microbial species abundance and metabolic pathway abundance in each time period are compared. Species fluctuations with an amplitude ≤8% and metabolic fluctuations with an amplitude ≤10% are identified as physiological rhythm noise and removed. Pure pathological microbial community drift signals with no periodicity and unidirectional shift are retained, thus completing the noise purification process.

[0057] Step 5: Adaptive Calibration of Population-Differentiated Thresholds. The system retrieves the stratified attribute parameters of the screening subjects, including age range, underlying intestinal disease status, and polyp activity status. It automatically matches the system's built-in dedicated quantitative drift threshold, performs secondary calibration and correction on the purified pathological drift signal, and adjusts the threshold upwards based on the polyp status of high-risk groups. This completely eliminates baseline bias in assessments across different populations and improves signal accuracy.

[0058] Step Six: Multi-source Drift Differential Judgment and Benign Interference Removal. Simultaneously collect the duration of gut microbiota drift, fluctuation slope, and fecal calprotectin levels of the screening subjects. According to preset differential judgment rules, identify and remove two types of benign gut microbiota drift data: dietary-related and inflammatory-related. Accurately screen out pathological drift data with specific precancerous symptom characteristics, and output the precise static gut microbiota drift risk value for the screening subjects.

[0059] Step 7: Dynamic Risk Stratification and Clinical Intervention Plan Output. Retrieve the comprehensive microbial drift gradient coefficients for multiple consecutive assessment periods of the screened subjects, substitute them into the normalized drift acceleration calculation formula, calculate the monthly microbial drift progression rate, and perform acceleration extreme value truncation to ensure data validity. Based on the four-level clinical risk stratification standard, determine the dynamic risk level corresponding to the screened subjects, and finally output the matched personalized clinical screening cycle and intervention plan, completing the full-process risk assessment.

[0060] Furthermore, traditional fixed abundance thresholds (0.1%) directly eliminate low-abundance bacterial communities in the 0.01%–0.1% range, while the characteristic bacterial community changes in early colorectal cancer and precancerous lesions often occur in this range. Using a uniform threshold would result in the loss of very early signals. Therefore, this embodiment adds a dynamic retention mechanism based on individual homeostasis, enabling the system to accurately capture weak pathological fluctuations without introducing interference from environmental bacteria.

[0061] After performing the time-series sampling, data quality control, and resident microbiome screening steps, the low-abundance microbiome dynamic retention threshold subunit is activated: First, all low-abundance microbiome groups with relative abundance in the 0.01%–0.1% range from the time-series sequencing data are extracted. The mean abundance, fluctuation range, and dispersion of each low-abundance microbiome group during the individual's homeostasis are then statistically analyzed. A 95% physiological fluctuation confidence interval specific to each low-abundance microbiome group is calculated, and an individual low-abundance microbiome dynamic retention threshold table corresponding to the individual's homeostasis baseline map is generated. This threshold table does not rely on general population standards; it is entirely generated from the screening subject's own time-series data and possesses highly individualized characteristics.

[0062] When entering the microbial community drift quantification stage, the system automatically calls the dynamic threshold table and executes a two-layer discrimination rule: for core resident microbial communities with a relative abundance ≥ 0.1%, the structure drift factor calculation formula of Example 1 is used; for low abundance microbial communities in the range of 0.01%–0.1%, they are only included in the structure drift calculation when the current abundance exceeds its own 95% confidence interval and the relative change is ≥ 2 times; if they are still within the confidence interval, they are judged as physiological fluctuations and removed to avoid interference from other microorganisms.

[0063] After drift quantization is completed, subsequent noise purification, threshold adaptive calibration, drift differential discrimination, and dynamic risk prediction all follow the same process.

[0064] Without increasing sequencing costs, altering operational procedures, or enhancing system complexity, this method significantly improves the ability to identify low-abundance pathogenic microbiota, particularly capturing very early, low-abundance, and minor anomalous shifts. This allows the system to identify precancerous trends even when static risk is not significantly increased and functional pathways are not significantly abnormal. This embodiment further enhances the detection sensitivity for occult precancerous lesions while maintaining the stability, specificity, and anti-interference capabilities of the original system, making it directly applicable to routine screening of high-risk clinical populations.

[0065] To further clarify and fully illustrate the specific implementation methods, technical implementation process, and actual technical effects of this invention, the following demonstration, using a real clinical high-risk screening subject, illustrates the entire process of this invention, from sample pre-screening, time-series sampling, steady-state modeling, two-dimensional drift quantification, noise removal, threshold calibration, multi-source drift discrimination to dynamic risk grading output. The entire process strictly adheres to all parameters, formulas, judgment criteria, and mechanistic logic defined above in this invention, demonstrating that the solution of this invention can be reproduced without difference by those skilled in the art and can stably achieve the invention's objective of non-invasive, accurate, and dynamic risk assessment of early colorectal cancer.

[0066] 1. Basic information of case screening subjects and determination of population stratification This practical case study selected a 48-year-old patient with a history of colorectal polyps, 8 months post-polyp surgery, belonging to the high-risk group for quiescent polyps, with no history of long-term antibiotic use or severe chronic intestinal inflammation, meeting the stratification criteria for underlying intestinal diseases and high-risk polyps described in this invention. According to the population stratification adaptation rules of this invention, this screening subject was matched with a 1-month time-series sampling interval, while also being adapted to the baseline drift threshold for high-risk groups. Furthermore, due to the quiescent polyp state, the overall threshold was increased by 10%; the corresponding adaptive weighting coefficients were configured as structural drift weight α = 0.3 and functional drift weight β = 0.7.

[0067] 2. Pre-sampling quality control and standardized time-series sampling execution Before formal sampling and sequencing, the clinical quality control guidelines disclosed in this invention were strictly followed. Screening subjects had not used any antibiotics within 14 days prior to sampling, and had strictly discontinued probiotics, proton pump inhibitors (PPIs), metformin, and nonsteroidal anti-inflammatory drugs (NSAIDs) 7 days prior to sampling to completely avoid any interference with the gut microbiota structure and metabolic function. For 3 days prior to sampling, a light diet was maintained, avoiding high-fat fried foods, high-sugar desserts, spicy and stimulating foods, and raw or cold foods to maintain a stable dietary structure and prevent sudden fluctuations in dietary microbiota.

[0068] Following a one-month sampling cycle, two morning stool samples were collected in month 0 and month 1, respectively. All samples were taken on an empty stomach during the first bowel movement of the morning to avoid rhythmic interference from daytime eating and intestinal peristalsis. Metagenomic high-throughput sequencing was performed on both time-series samples, with sequencing depths exceeding 5G clean data and effective read counts greater than 200,000. After data quality control filtering, contaminated, low-quality, and adapter-contaminated data were removed, resulting in two sets of standardized time-series microbial sequencing data suitable for homeostasis modeling.

[0069] 3. Construction of Individual-Specific Microbial Community Steady-State Baseline Maps Two sets of time-normalized sequencing data were retrieved, and resident microbiota screening rules were applied. Microbial species with a relative abundance ≥0.1% were retained, while all transient exogenous bacteria, food residue bacteria, and environmental bacteria were removed. This resulted in the final set of core resident microbiota species specific to the screened individual. Two types of core feature data were further extracted: firstly, the relative abundance structure of each resident microbiota; and secondly, the functional abundance characteristics of four core KEGG metabolic pathways: short-chain fatty acid synthesis pathway, bile acid metabolism pathway, intestinal toxin metabolism pathway, and intestinal inflammation regulation metabolic pathway.

[0070] The two sets of time-series feature data were input into a Gaussian nonlinear fitting model. Nonlinear normal distributions were used to fit the abundance distribution of microbial community structure and the abundance distribution of metabolic pathway function, respectively, to generate a microbial community homeostasis center value specific to the screening subject. At the same time, a 95% confidence interval for physiological fluctuations was constructed based on the statistical normal distribution rule, which fully covers the physiological fluctuation range of the normal microbial community of the screening subject. Finally, an individual-specific microbial community homeostasis benchmark map independent of the general population standard was generated, which serves as the sole control benchmark for all subsequent drift quantification calculations.

[0071] 4. Two-dimensional microbial community drift quantification calculation Within this one-month time-series monitoring period, the current detection data of the first month and the individual steady-state baseline data are retrieved, and the normalization calculation of the structural drift factor and functional drift factor is completed respectively. Then, the comprehensive drift gradient coefficient is obtained through adaptive weight fusion.

[0072] First, substitute the formula for calculating the normalized structure drift factor: ; In this case, n represents the number of core resident bacteria in the screened individual, which is 38 species. Month; Substitute the current abundance Si and steady-state abundance SOi of each bacterial community into the input, calculate the relative shift rate of a single species and obtain the average value, combine with time normalization, and then truncate to the 0~1 interval after maximum and minimum value normalization, finally calculate the structure drift factor Gs=0.12 in this case.

[0073] Then, substituting into the normalized functional drift factor calculation formula: In this case, m is fixed at a value of 4, corresponding to the four core carcinogenic metabolic pathways. The current abundance Fi of each of the four pathways is compared with the steady-state baseline abundance FOj, and the average relative shift rate is calculated. After normalization and truncation within the 0-1 interval, the functional drift factor Gf = 0.24 is obtained. Finally, it is substituted into the formula for the comprehensive drift gradient coefficient: This case involves a high-risk polyp population. Taking a=0.3 and β=0.7, the calculation yields: That is, the comprehensive drift gradient coefficient for this period is 0.204.

[0074] 5. Circadian rhythm noise purification The sample collection time was 6:10 AM, corresponding to the daytime period of 6:00 AM to 12:00 PM in the four-segment circadian rhythm model of this invention. Preset physiological fluctuation thresholds were retrieved for this period: species abundance fluctuation ≤8%, metabolic pathway abundance fluctuation ≤10%. Upon comparison, all original small-amplitude fluctuations in the data were within these thresholds and were identified as physiological noise in the diurnal rhythm, which was then completely removed. The remaining drift changes were continuous unidirectional shifts without rhythmic repetition characteristics, and were ultimately purified to obtain pure pathological drift signals without physiological noise interference.

[0075] 6. Population-adaptive threshold calibration This case screening targeted individuals at high risk for quiescent polyps. The initial high-risk thresholds were 0.15 for structural drift and 0.28 for functional drift. Following a rule, these thresholds were increased by 10%, resulting in actual thresholds of 0.165 for structural drift and 0.308 for functional drift. Comparing the calculated values ​​of Gs (0.12) and Gf (0.24) with the calibrated thresholds, both drift values ​​did not exceed their corresponding limits. Therefore, the initial assessment is that the current static drift is within the critical abnormal range, without severe structural or functional imbalance.

[0076] 7. Multi-source homogeneous drift differential discrimination The duration, fluctuation trend, and fecal calprotectin test results of this drift were collected simultaneously. The duration of this continuous unidirectional shift of the microbiota was 28 days, which meets the pathological duration standard of ≥25 days. The fluctuation slope showed a continuous and slow increasing trend with no signs of spontaneous decline. The clinical fecal calprotectin value was 32 μg / g, which is less than the clinical critical threshold of 50 μg / g.

[0077] According to the discrimination rules of this invention, this drift does not meet the characteristics of short-term fluctuations and spontaneous decline of dietary-origin drift, nor does it meet the characteristics of high calcium protectant protein and inflammation-dominated inflammatory drift. It fully meets the criteria for determining pathological drift in the precursor of carcinogenesis. Therefore, this drift signal is retained, all benign interference drifts are eliminated, and the accurate static microbiota drift risk result is output: there is a low-degree persistent pathological microbiota drift in the precursor of carcinogenesis.

[0078] 8. Calculation of microbial community drift acceleration and dynamic risk stratification Retrieve the overall drift coefficient G from the previous cycle t -1=0.112, current period Gt=0.204, interval ΔT=1 month, steady-state baseline initial coefficient G0=0.98, substituting into the acceleration calculation formula: ; The calculation yields: A = (0.204 - 0.112) The interval was truncated to [-0.1, 0.1], with a value of A = 0.094. According to the four-level risk stratification standard, A > 0.08, indicating an extremely high risk level. Although the single static drift threshold was not severely exceeded, the monthly drift rate of the microbiota was rapid, and the microbiota homeostasis showed a rapid deterioration trend, consistent with the dynamic progression characteristics of early precancerous lesions. Following clinical intervention rules, a clinical intervention plan for detailed colonoscopy was immediately implemented.

[0079] This case successfully identified early-stage progressive precancerous microbiota drift, which is easily missed by traditional static screening, through the complete system of individualized steady-state modeling, dual-dimensional weighted quantification, noise purification, multi-source drift discrimination, and dynamic acceleration prediction. Traditional universal benchmark screening may determine the subject as low-risk due to individual baseline differences, resulting in missed detections. However, this invention relies on individual steady-state comparison and dynamic rate analysis to accurately capture the continuous deterioration trend of the microbiota, achieving early warning. Therefore, this invention effectively solves the shortcomings of existing technologies, such as strong universality, poor individualization, lagging static judgment, and susceptibility to missed and misdiagnosed cases. It fully realizes the invention's purpose of non-invasive, accurate, dynamic, and individualized risk assessment of early colorectal cancer. The entire process has clear parameters, fixed steps, and can be fully engineered and clinically implemented.

[0080] It should be noted that the above embodiments of the present invention are merely preferred embodiments and are not intended to limit the scope of protection of the present invention. Under the premise of the technical principles and core architecture of the present invention, those skilled in the art can make adaptive fine-tuning adjustments to the sampling interval values, fluctuation judgment thresholds, and weighting parameters according to actual clinical screening scenarios. Such fine-tuning does not require creative effort and falls within the scope of protection of the present invention. Specific experimental methods and data preprocessing methods not described in detail in this invention are all conventional technical means in the prior art, and those skilled in the art can directly refer to the prior art for implementation; therefore, they will not be elaborated upon here.

Claims

1. A risk assessment system for early colorectal cancer, characterized in that, include: The temporal microbiota acquisition module is used to acquire multiple sets of raw fecal microbiota sequencing data from different time points of the screening subjects; The individual homeostasis construction module is connected to the time-series microbiota acquisition module. It is used to construct a unique gut microbiota homeostasis benchmark that is adapted to individual characteristics based on the time-series microbiota data of the screening subject, abandoning the unified benchmark for the general population. The gut microbiota drift quantification module is used to quantify the degree of gut microbiota drift from the dimensions of microbiota structure and microbiota function, with reference to the individual-specific gut microbiota homeostasis benchmark, and to obtain microbiota temporal drift risk data. The noise purification module is used to remove interference noise caused by human physiological rhythm fluctuations in the microbial community time series data, and to purify and obtain effective signals of pathological microbial community drift. The threshold adaptive calibration module is used to dynamically match the differential microbial drift judgment threshold according to the population stratification attributes of the screening subjects, and to calibrate and correct the pathological microbial drift signal. The drift difference discrimination module is used to distinguish the homologous drift of the microbial community caused by different factors, remove benign non-cancerous interfering drift data, and obtain accurate static risk data. The dynamic risk prediction module combines microbial drift data from multiple time-series nodes to quantify the dynamic progression rate of microbial drift and output the dynamic risk level of early colorectal cancer for the screening subjects and the corresponding clinical intervention plan.

2. The colorectal early cancer risk assessment system according to claim 1, characterized in that, The time-series microbiota collection module performs stratified differentiated time-series sampling and standardized clinical quality control. The screening target population is stratified into healthy young people, healthy middle-aged and elderly people, and people at high risk of underlying intestinal diseases and polyps. The sampling intervals for the three groups are configured as 3 months, 2 months and 1 month, respectively. The standardized clinical quality control includes: discontinuing antibiotics 14 days before sampling, discontinuing probiotics, PPI proton pump inhibitors, metformin and nonsteroidal anti-inflammatory drugs 7 days before sampling, adopting a light diet control mode for 3 days before sampling, and uniformly collecting the first stool sample in the morning; the sequencing data quality control standard is that the sequencing depth is not less than 5G clean data and the number of effective reads is not less than 200,000.

3. The colorectal early cancer risk assessment system according to claim 1, characterized in that, The individual homeostasis construction module screens resident microbiota with a relative abundance of ≥0.1% in the sequencing data, removes transient exogenous bacteria data, and extracts the structural abundance characteristics and core metabolic pathway abundance characteristics of resident microbiota. The individual homeostasis construction module uses a Gaussian nonlinear fitting model to fit the extracted feature data with a nonlinear probability distribution, generates a homeostasis center value for the individual microbiome of the screening object, and constructs a 95% confidence interval for physiological fluctuations to form an individual-specific microbiome homeostasis baseline map.

4. The colorectal early cancer risk assessment system according to claim 1, characterized in that, The microbial community drift quantification module is configured with a population stratification adaptive weighting mechanism and incorporates a comprehensive microbial community drift gradient coefficient algorithm; the formula for the comprehensive microbial community drift gradient coefficient algorithm is: ； Wherein, G is the microbial community comprehensive drift gradient coefficient with a value of 0~1, Gs is the normalized structural drift factor, Gf is the normalized functional drift factor, α is the structural drift weight coefficient, β is the functional drift weight coefficient, and α+β=1; the population adaptive weight configuration is as follows: healthy young people α=0.4, β=0.6, healthy middle-aged and elderly people α=0.35, β=0.65, and people with underlying intestinal diseases and high risk of polyps α=0.3, β=0.

7.

5. The colorectal early cancer risk assessment system according to claim 4, characterized in that, The formula for calculating the normalized structural drift factor Gs is as follows: ； Where n is the number of species in the core resident microbial community of an individual, Si is the relative abundance of the i-th microbial community currently detected, S0i is the standard abundance of the i-th microbial community under the steady-state baseline of the individual, ΔT is the time interval between two samplings in months, and the calculation results are truncated to the 0-1 interval after being normalized to the maximum and minimum values; the formula for calculating the normalized functional drift factor Gf is: ； Where m is fixed at 4, corresponding to the four core KEGG metabolic pathways associated with colorectal cancer mutations: short-chain fatty acid synthesis pathway, bile acid metabolism pathway, intestinal toxin metabolism pathway, and intestinal inflammation regulation metabolic pathway. Fj is the abundance of the j-th metabolic pathway currently detected, and F0j is the abundance of the j-th metabolic pathway under the individual homeostatic baseline. The calculation results are truncated to the 0~1 interval after being normalized to the maximum and minimum values.

6. The colorectal early cancer risk assessment system according to claim 1, characterized in that, The noise purification module is pre-trained with a 24-hour four-segment circadian rhythm baseline model, which divides the 24 hours into four periods: 0:00-6:00 AM, 6:00-12:00 PM, 12:00-6:00 PM, and 6:00-12:00 AM. The noise purification module identifies periodic small fluctuations in bacterial species abundance with an amplitude of ≤8% and metabolic pathway abundance with an amplitude of ≤10% as physiological noise and removes them, while retaining pure pathological bacterial community drift signals that are non-periodic and continuously shift in one direction.

7. The colorectal early cancer risk assessment system according to claim 1, characterized in that, The threshold adaptive calibration module has a built-in quantitative drift threshold library for three types of people. Healthy young adults: structural drift threshold 0.20, functional drift threshold 0.35; Healthy middle-aged and elderly adults: structural drift threshold 0.18, functional drift threshold 0.32; High-risk groups with underlying intestinal diseases and polyps: structural drift threshold 0.15, functional drift threshold 0.28; Among them, for people with underlying intestinal diseases and those at high risk of polyps, active polyps are distinguished from inactive polyps, and the threshold for inactive polyps is increased by 10% overall.

8. The colorectal early cancer risk assessment system according to claim 1, characterized in that, The drift differential discrimination module combines the duration of bacterial community drift, fluctuation slope and fecal calprotectin detection threshold to achieve accurate differentiation of multi-source drift; Drifts that fluctuate continuously for 10-20 days, have a gentle slope, and can spontaneously decline are identified as dietary-induced physiological drifts and are excluded. Drifts with sustained fluctuations for ≥20 days, calprotectin levels ≥50 μg / g, and no sustained worsening slope were classified as benign inflammatory drifts and excluded. Drifts that fluctuate continuously for ≥25 days, have a continuously increasing slope, show no spontaneous decline trend, and have a calprotectin detection value <50μg / g are identified as pathological drifts that are precursors to cancer, and a precise static microbiota drift risk value is output.

9. The colorectal early cancer risk assessment system according to claim 1, characterized in that, The dynamic risk prediction module has a built-in normalized drift accelerometer. Calculation formula and four-level clinical risk stratification mechanism; The formula for calculating the normalized drift acceleration is: ； Where A is the normalized monthly acceleration of the microbial community drift, Gt is the comprehensive drift gradient coefficient of the current period, Gt-1 is the comprehensive drift gradient coefficient of the previous period, ΔT is the sampling interval in months, GO is the initial coefficient of the individual steady-state baseline, and the acceleration A is forcibly truncated in the range of [-0.1, 0.1]; The four-level risk stratification standard is: A≤0.02 corresponds to low risk, Corresponding to medium risk High risk, each risk level is matched with a corresponding annual screening, 6-month follow-up, 3-month intervention follow-up, and immediate colonoscopy detailed examination plan.

10. A method for assessing the risk of early colorectal cancer, characterized in that, The colorectal early cancer risk assessment system according to any one of claims 1-9 comprises the following steps: S1. Match the corresponding sampling interval according to the stratification type of the screening target population, complete the standardized quality control of pre-medication, diet and sampling time, collect multiple sets of morning fecal samples, and obtain standardized time-series microbial sequencing data through metagenomic sequencing and quality control filtering. S2. Screen the resident microbial community characteristic data in the sequencing data, remove exogenous bacteria, and construct an individual microbial community homeostasis benchmark map specific to the screened object through a Gaussian nonlinear fitting model; S3. Match the adaptive weight coefficients of the population, calculate the structural drift factor and the functional drift factor respectively, and obtain the weighted comprehensive drift gradient coefficient of the microbiome to obtain the initial time-series drift risk value. S4. Combine the diurnal rhythm baseline model to remove physiological rhythm noise and purify to obtain pure pathological microbial community drift signal; S5. Match a specific drift threshold based on the population attributes of the screening target, and adaptively calibrate and correct pathological drift signals; S6. Differentiate multi-source homology drift by drift duration, fluctuation slope and calprotectin threshold, remove benign interference data and obtain accurate static risk value; S7. Calculate the normalized monthly acceleration of microbial community drift, determine the dynamic risk level by combining the four-level risk stratification standard, and output a matching clinical screening and intervention plan.

Images