Screening method of cecal multi-omics characteristics of meat rabbits for non-antibiotic feeding effect discrimination

Abstract

This invention relates to a method for screening multi-omics features of rabbit cecum for judging the effectiveness of antibiotic-free feeding. Cecum samples are collected from rabbits in antibiotic-free and control feeding groups to obtain microbiome, metabolome, and host response omics data. Candidate microbiome features, candidate metabolite features, and candidate host response features are screened based on inter-group differences. Closed-loop matching of the candidate features is performed according to preset correspondence rules to obtain closed feature groups, and candidate features that do not form closed feature groups are eliminated. Then, based on the inter-group specificity and intra-group consistency of the closed feature groups between the antibiotic-free and control feeding groups, multi-omics features of rabbit cecum for judging the effectiveness of antibiotic-free feeding are obtained. This invention significantly improves the accuracy, stability, and biological interpretability of judging the effectiveness of antibiotic-free feeding by constructing closed feature groups and combining uniqueness, breakpoints, and host permission gating screening.

Description

Technical Field

[0001] This invention belongs to the field of animal antibiotic-free feeding evaluation and bioinformatics analysis technology, specifically a method for screening multi-omics features of the cecum in meat rabbits for judging the effect of antibiotic-free feeding. Background Technology

[0002] Existing screening methods for judging the effects of antibiotic-free feeding, while beginning to shift from simple microbial abundance analysis to the combined use of microbial and metabolic information, still largely remain at the level of screening based on parallel differential characteristics and statistical correlation. For example, patent application CN114974432A's core idea is to establish representative strain genome sequence libraries and metabolite gene cluster sequence libraries, then use metagenomic sequencing data to calculate strain abundance and metabolite abundance respectively, and then screen for significantly different strains and significantly different metabolites. Strains and metabolites with significantly different characteristics are used as biomarkers, and heterogeneity analysis and predictive models are combined as needed to complete subsequent screening. This type of approach can indeed improve screening efficiency at both the microbial and metabolic levels and reduce the time and cost of directly detecting metabolites. However, its technical approach is essentially still based on a two-layer joint discrimination based on the correspondence between bacterial species and metabolites, focusing on finding candidates with statistical differences, but failing to further address the continuous transmission relationship from microbial changes to metabolic changes and then to host functional responses. Regarding the effects of antibiotic-free feeding, the real challenge lies not only in identifying which microbiota and metabolites change, but also in determining whether these changes can form a stable, continuous, and host-specific chain of action within the rabbit cecum. Screening solely based on bacterial abundance, metabolite abundance, and their correlations can easily include statistically significant but biologically insignificant characteristics, leading to biomarkers that are more superficially correlated than accurately revealing the true impact of antibiotic-free feeding on the cecal microecology and host homeostasis.

[0003] Furthermore, existing methods suffer from insufficient constraints on inter-group specificity, intra-group stability, and host outcomes. While CN114974432A proposes using heterogeneity indicators to retain candidate strains with low heterogeneity in multi-cohort scenarios and inputting screening results into a classifier to construct a predictive model, its stability control primarily targets statistical consistency between different cohorts. It does not structurally discriminate whether the same feature chain maintains a consistent closed state within the treatment group or exhibits fixed breakpoints in the control group, nor does it eliminate spurious features that can be substituted, compensated for by bypassing, or are only formally valid. Especially for discriminating the effects of antibiotic-free feeding, without gating constraints at the host response level, even if there are significant differences in the microbiota and metabolism, it is impossible to confirm whether these differences ultimately lead to the maintenance of the cecal mucosal barrier and improvement of local homeostasis, or to increased inflammatory stimulation and increased risk of mucosal damage. In other words, existing methods struggle to distinguish between truly effective feature chains with positive discriminative value and spurious closed relationships that are only statistically significant but lack host-permitted significance. Furthermore, these methods generally do not specifically identify the reversal relationships of results for the same front-end characteristics in different groups, nor do they verify whether the host response results are unique. Therefore, when faced with the complex microecological remodeling effects between the antibiotic-free group and the control group, they are prone to problems such as redundant screening results, incomplete interpretation chains, and significant impact of random fluctuations on discrimination accuracy. In other words, while existing technologies can complete the initial screening at the microbiome and metabolite levels, they still lack a three-layer closed screening mechanism for antibiotic-free rabbit feeding scenarios. It is difficult to balance biological continuity, inter-group differences, intra-group consistency, and host functional effectiveness. This is the main shortcoming and drawback of these methods in practical applications. Summary of the Invention

[0004] The purpose of this invention is to provide a method for screening multi-omics features of the cecum in meat rabbits for judging the effect of antibiotic-free feeding, thereby solving some of the drawbacks and shortcomings pointed out in the background art.

[0005] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: a method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding, comprising: collecting cecum samples of rabbits in antibiotic-free feeding group and control feeding group, obtaining microbiome data, metabolome data and host response omics data; and screening candidate microbial features, candidate metabolite features and candidate host response features based on inter-group differences.

[0006] According to the preset correspondence rules, the candidate features are matched in terms of the preceding, middle and following terms to obtain closed feature groups, and candidate features that do not form closed feature groups are removed; then, based on the inter-group specificity and intra-group consistency of the closed feature groups between the antibiotic-free feeding group and the control feeding group, multi-omics features of rabbit cecum used for judging the effect of antibiotic-free feeding are screened.

[0007] Furthermore, after the closure matching, a uniqueness test of the closed structure is performed on the closed feature group; the preceding or middle feature is fixed, and the remaining features are replaced and rematched; if a closed feature group is still formed after replacement, it is determined to be a replaceable closed feature group and is eliminated; only non-replaceable closed feature groups are retained.

[0008] Furthermore, during the specific screening based on the groups, the breakpoints of the anterior, middle, and posterior feature chains corresponding to the closed feature group in the control feeding group are determined; if there is an interruption between the anterior and middle features, it is determined to be an anterior break; if there is an interruption between the middle and posterior features, it is determined to be a posterior break; only closed feature groups that have fixed breakpoints in the control feeding group and remain closed in the antibiotic-free feeding group are retained.

[0009] Furthermore, the latter feature is used for host permission gating discrimination; if the latter feature indicates the maintenance of the cecal mucosal barrier or the maintenance of local homeostasis, the corresponding closure feature group is determined to have passed the host permission gating; if the latter feature indicates mucosal damage, homeostasis imbalance or increased inflammatory stimulation, the corresponding closure feature group is determined to be a pseudo-closure feature group and is removed.

[0010] Furthermore, the term "fixed breakpoint" means that the same feature chain is interrupted between the preceding and middle features, or between the middle and following features, in the control feeding group; if both the preceding and following segments are interrupted, it is determined to be a feature chain with a drifting breakpoint and is removed.

[0011] Furthermore, bypass compensation is performed on the feature chains of the front or rear fracture; the features on the front side of the fracture site are kept unchanged, and the rear features are replaced and the closure matching is performed again; if a new closed feature group is formed after replacement, it is determined to be a compensable fracture feature chain and is removed; only the uncompensable fracture feature chains are retained.

[0012] Furthermore, the intragroup consistency refers to the same closed feature group maintaining a consistent closed state within the antibiotic-free feeding group and maintaining a consistent breakpoint or consistent host permission gating result within the control feeding group; closed feature groups with inconsistent intragroup states are removed.

[0013] Furthermore, a uniqueness test is performed on the closed feature group that passes the host permission gating; keeping the first and second features unchanged, the host permission gating is re-performed after replacing the second feature; if it still passes the host permission gating after replacement, it is determined to be a non-unique closed feature group and is removed.

[0014] Furthermore, permission inversion discrimination was performed on closed feature groups with the same antecedent and middle features in the antibiotic-free feeding group and the control feeding group; if the antibiotic-free feeding group passed the host permission gating but the control feeding group failed, it was determined to be a permission inversion closed feature group and retained; if the host permission gating results of the two groups were the same, they were removed.

[0015] Furthermore, a uniqueness test for license inversion is performed on the license inversion closed feature group; keeping the first and second features unchanged, the license inversion is re-evaluated after replacing the second feature; if the license inversion is still satisfied after replacement, it is determined to be a non-unique license inversion closed feature group and is removed.

[0016] The beneficial effects of this invention are as follows: By integrating microbiome, metabolome, and host response omics data, this invention constructs closed-loop feature sets of antecedent, intermediate, and consequent terms, achieving a shift from single-difference screening to multi-level correlation screening. Through closed-loop matching, microbial changes, metabolic changes, and host responses are organically linked, reflecting the continuous biological process from microbial-driven to host-effect interactions. Compared to traditional methods relying solely on single-omics or simple correlation analysis, this significantly improves the biological interpretability and discriminative accuracy of feature screening. Simultaneously, through inter-group specificity and intra-group consistency screening, unstable or indifferential features are effectively eliminated, making the screening results more stable and reliable.

[0017] Furthermore, this invention introduces a uniqueness test for closed structures, a determination of the fixity of fracture sites, a bypass compensation test, and a host permission gating mechanism to constrain and optimize the closed feature set from multiple levels. By eliminating substitutable, compensable, and non-unique feature combinations with permission results, and retaining feature sets that show stable fracture differences and permission reversal relationships between the antibiotic-free feeding group and the control group, the specificity, uniqueness, and functional relevance of the screened multi-omics features are significantly improved, thereby more accurately reflecting the true impact of antibiotic-free feeding on the cecal microecology and host status of rabbits. Attached Figure Description

[0018] Figure 1 This is a functional relationship diagram of the multi-omics feature screening method for the cecum of meat rabbits used in the present invention for judging the effect of antibiotic-free feeding; Figure 2 This is a comparison chart of the candidate feature screening results between the antibiotic-free feeding group and the control feeding group in Example 1 of the present invention; Figure 3 This is a graph showing the closure correlation index and uniqueness screening results of candidate closed feature chains in Embodiment 1 of the present invention; Figure 4 This is a diagram showing the structural consistency of chains A and C in the antibiotic-free feeding group and the control group in Example 1 of the present invention. Figure 5 This is a comparison chart of the core detection data of candidate closed feature groups in Embodiment 2 of the present invention; Figure 6 This is a line graph showing the host permission gating score and uniqueness test results in Embodiment 2 of the present invention. Figure 7 This is a radar chart showing the permitted inversion intensity index and inversion uniqueness results in Embodiment 2 of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] Combined with appendix Figure 1 This invention relates to a method for screening multi-omics features of the cecum in rabbits to determine the effectiveness of antibiotic-free feeding. Rabbits at the same growth stage and in the same health state were selected and divided into an antibiotic-free feeding group and a control feeding group. Cecal contents and cecal tissue samples were collected from each group after a predetermined feeding cycle. To ensure the authenticity and comparability of the test results, the sampling time and sampling site were kept consistent, and the samples were preserved at low temperatures and uniformly numbered. Subsequently, multi-omics tests were performed on the collected samples. Cecal contents were used to obtain microbiome and metabolome data, while cecal tissue samples were used to obtain host response omics data. Microbiome data was used to characterize changes in the composition and abundance of the cecal microbiota, metabolome data was used to characterize changes in the types and relative contents of metabolites in the cecum, and host response omics data was used to characterize changes in the local barrier status, immune response status, and tissue homeostasis of the cecum.

[0021] After obtaining the aforementioned multi-omics data, inter-group difference analysis was performed between the antibiotic-free feeding group and the control feeding group. For the microbiome data, the differences in bacterial abundance between the two groups were compared, and genera, species, or functional groups that showed significant changes under antibiotic-free feeding conditions were screened as candidate microbial characteristics. For the metabolome data, the differences in metabolite abundance between the two groups were compared, and differentially expressed metabolites associated with antibiotic-free feeding were screened as candidate metabolite characteristics. For the host response omics data, differentially expressed indices related to mucosal barrier maintenance, inflammatory response regulation, and local homeostasis were compared between the two groups as candidate host response characteristics.

[0022] After obtaining candidate microbial features, candidate metabolite features, and candidate host response features, closed-loop matching of the candidate features (preceding, intermediate, and subsequent terms) is performed according to preset correspondence rules. Specifically, candidate microbial features are considered the preceding features, candidate metabolite features the intermediate features, and candidate host response features the subsequent features. The preset correspondence rules are constructed based on the association between microorganisms and metabolites, and the association between metabolites and host responses. Correlation analysis and association strength calculations are performed on multi-omics data to establish the correspondence between preceding and intermediate features, as well as between intermediate and subsequent features. Based on this, feature combinations that simultaneously satisfy significant correlation between the preceding and intermediate terms and significant correlation between the intermediate and subsequent terms are selected to construct closed-loop feature groups consisting of sequentially connected preceding, intermediate, and subsequent features.

[0023] Candidate feature combinations that do not simultaneously meet the above association conditions and cannot form a continuous association relationship are judged as candidate features that do not form a closed structure and are removed from the candidate feature set. Through this step, only closed feature groups with complete association paths are retained, so that the selected features can reflect the continuous biological process of metabolic changes triggered by microbial changes and further acting on the host response.

[0024] After obtaining the closed-loop characteristic groups, further screening was conducted based on the inter-group specificity and intra-group consistency between the antibiotic-free feeding group and the control feeding group. Inter-group specificity refers to the stable closed-loop state of the same closed-loop characteristic group in the antibiotic-free feeding group, while it shows weakened association or broken connection in the control feeding group, thus reflecting the difference in response to antibiotic-free feeding conditions. Intra-group consistency refers to the consistent association structure and trend of the same closed-loop characteristic group among samples within the same treatment group. Closed-loop characteristic groups with significant differences or unstable changes within the group were removed.

[0025] Through the above-described processing procedure based on closed-loop matching and consistency screening, a multi-omics feature set that simultaneously possesses clear biological association pathways, significant inter-group differences, and good intra-group stability is finally obtained. This feature set can be used to identify and evaluate the effect of antibiotic-free feeding on meat rabbits.

[0026] After obtaining closed feature groups, a uniqueness test for their closed structure is performed to improve the stability and reliability of the features. Specifically, a fixing and replacement operation is performed on each closed feature group. First, the first or middle feature is fixed and kept unchanged. Then, the remaining features are replaced, and the association matching is re-performed according to the established closure matching rules to determine whether a complete closed structure can still be formed.

[0027] During the replacement process, features with similar expression levels or correlations within the same category of candidate features are prioritized for replacement to ensure the rationality of the replacement operation. After the replacement is completed, the new feature combination is subjected to correlation analysis again. If the correlation conditions between the preceding and middle items, as well as between the middle and following items, are still satisfied, thus forming a new closed feature group, then the original closed feature group is determined to be substitutable.

[0028] Feature combinations deemed replaceable closed feature groups are excluded because their structures lack uniqueness and are susceptible to feature substitution, thus reducing their stability in determining the effectiveness of antibiotic-free feeding. Feature combinations that cannot reform a closed structure under substitution conditions are deemed non-replaceable closed feature groups, and are retained because they possess stable and specific correlations.

[0029] After constructing and uniquely verifying the closed feature groups, the closed feature groups were further screened based on inter-group specificity. Specifically, using the control feeding group as a reference, the association status analysis was performed on the feature chain consisting of the preceding, middle, and following features corresponding to each closed feature group to determine its connection integrity and breakage location in the control feeding group.

[0030] The association strength between the preceding and middle features, and between the middle and following features, is calculated separately. When the association strength of a segment falls below a preset threshold, the connection is considered broken. If the association between the preceding and middle features is broken, it is considered a break in the preceding segment; if the association between the middle and following features is broken, it is considered a break in the following segment. This method clarifies the breakpoints of each closed feature group in the control feeding group.

[0031] Furthermore, the consistency of breakage patterns of the same closed feature group across samples in the control feeding group was assessed. If the feature chain showed breakage at the same location in all or most samples, the breakage site was identified as a fixed breakage site. If the breakage location varied between different samples, it was determined to be an unstable feature chain and was removed.

[0032] Based on this, the closed feature groups with fixed breakpoints in the control feeding group were compared with the corresponding feature chains in the antibiotic-free feeding group. Only the closed feature groups that maintained an intact closed structure in the antibiotic-free feeding group and had fixed breakpoints in the control feeding group were retained.

[0033] After identifying breakpoints and selecting closure feature groups that meet intergroup specificity, a subsequent feature is introduced to perform host permission gating discrimination on the closure feature groups. This subsequent feature is derived from host response proteomics data and reflects the local physiological state and immune response level of the cecum. For each closure feature group, its corresponding subsequent feature is extracted, and the biological effect characterized by this feature is quantitatively evaluated.

[0034] Specifically, the functional attributes of subsequent features are determined based on the direction and intensity of changes in host response indicators. When a subsequent feature manifests as enhanced mucosal barrier integrity, stable tight junction structures, or maintenance of local immune homeostasis, it is considered to have a positive regulatory effect, and the corresponding closure feature group is determined to have passed the host-permitted gating. Conversely, when a subsequent feature manifests as damaged mucosal structure, increased barrier permeability, or upregulated expression of inflammatory factors, it is considered to have a negative regulatory effect, and the corresponding closure feature group is determined to have failed to pass the host-permitted gating. For closure feature groups that have not passed the host-permitted gating, although they form a structural closure, they fail to demonstrate biological effects conducive to host homeostasis, and therefore are defined as pseudo-closure feature groups and eliminated.

[0035] After identifying the breakpoints, the stability of these breakpoints is further defined. Specifically, the breakpoints of the feature chains corresponding to the same closed feature group are statistically analyzed in each sample within the control feeding group. When the feature chain breaks between the preceding and middle features in all or most samples, it is considered a breakpoint at the preceding segment, and the breakpoint is fixed. When the feature chain breaks between the middle and following features in all or most samples, it is considered a breakpoint at the following segment, and the breakpoint is fixed. If the same feature chain exhibits both preceding and following segment breaks in different samples, the breakpoint of the feature chain is considered unstable, defined as a breakpoint drift feature chain, and discarded.

[0036] Based on this, a bypass compensation test is performed on feature chains identified as having an anterior or posterior fracture. Specifically, features preceding the fracture site are kept unchanged; for an anterior fracture, the preceding features remain unchanged; for a posterior fracture, features preceding the fracture site in the preceding and middle features remain unchanged. Features following the fracture site are replaced within the candidate feature range, and association matching is performed again according to the closure matching rules. If a new closed feature group can be formed after replacement, it indicates that the fracture relationship can be compensated through feature replacement, and the chain is identified as a compensable fracture feature chain and discarded. Feature chains that still cannot form a closed structure after replacement are identified as uncompensable fracture feature chains, considered to have stable and specific fracture relationships, and are retained.

[0037] After determining the breakpoint and host permission gating, intragroup consistency screening was performed on the closed feature groups. Specifically, the antibiotic-free feeding group and the control feeding group were used as evaluation objects, and the consistency analysis of the structural state of each closed feature group in different samples within each group was performed.

[0038] In the antibiotic-free feeding group, the association structure of the same closed feature group was examined to ensure that it remained completely closed across all samples, meaning that a stable association was maintained between the preceding and middle features, as well as between the middle and subsequent features. If weakened associations or broken connections were observed in some samples, the closed feature group was determined to lack a consistent closed state within the antibiotic-free feeding group.

[0039] In the control feeding group, the status of the feature chains of the same closed feature group was determined, requiring them to show consistent breakpoints or consistent host permission gating results across all samples. Specifically, when all or most samples broke at the same location, or all failed to pass the host permission gating, the closed feature group was considered to have a consistent status within the control feeding group. If there were differences in breakpoint locations or inconsistent permission gating results among different samples, the group was considered to be unstable.

[0040] If any group in the antibiotic-free feeding group or the control feeding group has a closed characteristic group with inconsistent state, it is considered that its stability is insufficient and it is difficult to use as a reliable basis for judgment, so it is excluded.

[0041] After completing the host permission gating judgment, the uniqueness of the permission result is further checked for closed feature groups that have passed the host permission gating. Specifically, for each closed feature group, its first and second features remain unchanged, and the second feature is replaced within the range of candidate host response features, and the host permission gating judgment is re-executed. If the replaced feature combination still passes the host permission gating, it means that the permission result of the closed feature group can be achieved by different second features, lacking uniqueness, and is judged as a non-unique closed feature group and eliminated. For closed feature groups that cannot pass the host permission gating again after replacement, their permission result is determined to be unique and retained.

[0042] Based on this, permission inversion discrimination was performed on closed feature groups with the same antecedent and median features in both the antibiotic-free and control feeding groups. Specifically, the host permission gating results of corresponding closed feature groups in the two groups were compared. When the closed feature group passed the host permission gating in the antibiotic-free feeding group but failed to pass the host permission gating in the control feeding group, it was identified as a permission inversion closed feature group, considered to reflect the differential impact of feeding method changes on host status, and was retained. If the host permission gating results were the same in both groups, it indicated that the feature combination lacked discriminatory power and was discarded.

[0043] For the retained permissible reversal closed feature groups, a further permissible reversal uniqueness test is performed. Specifically, keeping the antecedent and middle features unchanged, substitutions are made within the range of candidate consequent features, and the permissible reversal discrimination is re-executed. If the permissible reversal condition of passing for the antibiotic-free feeding group and failing for the control feeding group is still met after substitution, it indicates that the reversal relationship can be realized by multiple consequent features, and it is determined to be a non-unique permissible reversal closed feature group and is removed. For closed feature groups that no longer meet the permissible reversal condition after substitution, their reversal relationship is determined to be unique and they are retained.

[0044] Example 1:

[0045] In this embodiment, a standardized meat rabbit breeding base plans to replace the antibiotic-containing growth-promoting formula with an antibiotic-free formula in the spring fattening batch. To avoid the production risks brought about by direct population expansion, the base first conducted a cecal multi-omics validation experiment in the same batch of Ira meat rabbits. The goal was to screen out a closed characteristic group that could stably indicate the effect of antibiotic-free feeding from changes in microorganisms, metabolism, and host response. During the experiment, the same technician continuously recorded feed intake, fecal characteristics, and mental state. On day 18, three meat rabbits in the control group showed softer feces, but their feed intake did not decrease significantly. Therefore, they continued to be fed according to the established plan until the end of the cycle to ensure that the samples were under the natural fluctuation conditions of a real business scenario.

[0046] Thirty-two healthy Ira rabbits of uniform body condition, having entered the fattening stage after weaning, were selected. Their initial weight ranged from 1.12 kg to 1.18 kg. They were randomly divided into an antibiotic-free feeding group and a control feeding group, with 16 rabbits in each group, based on ear tags. Both groups received the same basal diet in terms of energy and crude protein levels. The antibiotic-free feeding group received no antibiotic growth promoters throughout the feeding period, but instead used acidifiers, plant extracts, and compound enzymes as synergistic substitutes. The control feeding group followed the existing conventional feeding program at the base. The indoor temperature was maintained between 19℃ and 23℃, and the relative humidity between 55% and 68%. Rabbits had free access to food and water for 42 consecutive days.

[0047] On day 42, from 9:00 AM to 10:00 AM, all animals were slaughtered and sampled. Cecal contents and cecal tissue were collected from each animal, maintaining consistent sampling sites. Immediately after sampling, samples were flash-frozen in liquid nitrogen and stored at -80°C. Cecal contents were used for 16S sequencing and untargeted metabolomics analysis, while cecal tissue was used for tight junction protein and inflammation-related marker detection. Microbiome analysis was performed using relative abundance at the genus level, metabolomics analysis used normalized peak intensity to represent relative metabolite content, and host response proteomics analysis used relative expression levels. To ensure comparability for subsequent matching, all samples were uniformly numbered, prepared in the same batch, and processed in the same batch, with the same normalization process used to output association analysis matrices.

[0048] Using the criterion of a clear direction of change between two groups and a significance level less than 0.05, three candidate anterior features, three candidate middle features, and four candidate posterior features were first obtained. In the antibiotic-free feeding group, the relative abundance of *Bacteroides* and *Lactobacillus* increased, while the relative abundance of *Clostridium* decreased. Among metabolites, butyric acid and indolepropionic acid increased, while deoxycholic acid decreased. In the host response, ZO-1, MUC2, and Occludin increased, while IL-6 decreased. The screening results are shown in Table 1. Figure 2 The differences in the first, middle, and last characteristics between the antibiotic-free feeding group and the control group were presented in a grouped horizontal bar chart. Among them, Bacteroides increased from 6.1% to 9.8%, Lactobacillus increased from 4.5% to 7.4%, Clostridium decreased from 5.7% to 3.2%, butyric acid increased from 2.41 to 3.82, indolepropionic acid increased from 0.84 to 1.46, deoxycholic acid decreased from 1.37 to 0.91, ZO-1 increased from 1.01 to 1.72, MUC2 increased from 1.11 to 1.64, Occludin increased from 1.05 to 1.69, and IL-6 decreased from 1.36 to 0.78. Figure 2 The study demonstrated the intergroup differences in candidate features, which can intuitively reflect the overall changes in the antibiotic-free feeding group in terms of microbial community structure, metabolite composition, and some barrier-related response indicators that are conducive to the maintenance of cecal homeostasis.

[0049]

[0050] Association matching was performed using microbial characteristics as the antecedent, metabolite characteristics as the middle term, and host response characteristics as the consequent. Stable association between the two segments of each chain and a closure association index of 1.45 or higher were used as criteria for advancement to the next round of screening. Three candidate closed feature chains were obtained: chain A consisted of Bacteroides, butyric acid, and ZO-1; chain B consisted of Lactobacillus, indolepropionic acid, and MUC2; and chain C consisted of Clostridium, deoxycholic acid, and IL-6. Chains A and B showed complete closure within the antibiotic-free group and consistent association directions. Although chain C reached the initial closure index, its stability was low. Specific parameters are shown in Table 2. Figure 3 The closure correlation index and structural stability of each chain are displayed using a bubble scatter plot. Chain A has a closure correlation index of 1.675, chain B 1.557, and chain C 1.511, all exceeding the entry threshold of 1.45. However, subsequent replacement operations on chain A failed to form a new closure group. Chain B, even after replacing MUC2 with Occludin, still achieved a closure correlation index of 1.516, indicating that its closure structure is substitutable. Although chain C passed the initial screening, its structural stability was the lowest among the three candidate chains, and subsequent breakpoint drift was observed. Figure 3 It can intuitively show the progressive relationship from passing the initial screening to the uniqueness test and then to the subsequent retention or rejection.

[0051]

[0052] Immobilization and replacement operations were performed on chains A, B, and C that entered the initial screening. For chain A, after immobilizing Bacteroides, butyric acid was replaced with indolepropionic acid and deoxycholic acid, and ZO-1 was replaced with MUC2 and Occludin, but no complete closed structure was formed again. After immobilizing butyric acid, the former was replaced with Lactobacillus and Clostridium, but again no new closed group was formed, indicating that chain A is not substitutable. For chain B, after immobilizing Lactobacillus and indolepropionic acid, replacing MUC2 with Occludin, the correlation strength of the latter segment still reached 0.68, the structural stability was 0.84, and the recalculated new closure correlation index was 1.516, which still met the closure requirements, indicating that chain B can be replaced by similar subsequent features, and the closed structure is not unique, so it was removed. For chain C, after immobilizing the former or middle term and replacing the rest, no closure of the same level appeared, but its subsequent breakpoint drift was observed in the control group, and it failed to enter the final retention set.

[0053] In the control group, chain A showed anterior segment breakage in all 16 samples, meaning the association strength between Bacteroides and butyric acid decreased to 0.12 to 0.21, while the association strength between butyric acid and ZO-1 remained at a moderate level of 0.61 to 0.68, thus indicating a fixed anterior segment breakage. Subsequently, keeping Bacteroides unchanged, the middle term was replaced sequentially with indolepropionic acid and deoxycholic acid, followed by the replacement of MUC2 and Occludin. Three consecutive rounds of reassortment failed to restore complete closure, with zero successful reclosure attempts, indicating that the breakage could not be compensated for by posterior feature replacement. In the control group, chain C showed anterior segment breakage in 8 samples and posterior segment breakage in 8 samples, indicating breakage site drift. Furthermore, while keeping Clostridium unchanged, one round of posterior replacement with butyric acid and ZO-1 resulted in one reconstruction closure, indicating that the breakage relationship was compensable; therefore, chain C was excluded.

[0054] In the antibiotic-free feeding group, all 16 samples of chain A maintained complete closure. The association strength of the initial segment ranged from 0.76 to 0.87, and the association strength of the subsequent segment ranged from 0.72 to 0.83, with consistent direction and no reversal. In the control group, all 16 samples of chain A maintained initial segment breakage, and the breakage site did not drift, demonstrating good inter-group specificity and intra-group consistency. Although chain B showed relatively uniform closure in the antibiotic-free feeding group, it was excluded due to failure of the uniqueness test. In the antibiotic-free feeding group, only 13 samples of chain C maintained complete closure, while the association of the subsequent segment in the other 3 samples was significantly weakened. In the control group, mixed breaks of the initial and subsequent segments were observed, indicating inconsistency within the group. Therefore, chain C was not used as a reliable criterion for judgment. Figure 4A consistency heatmap was used to identify the sample-level structural status of chains A and C in the antibiotic-free feeding group and the control group. In the antibiotic-free feeding group, all 16 samples of chain A showed complete closure, while in the control group, all 16 samples showed anterior segment breakage, indicating stable inter-group specificity. In the antibiotic-free feeding group, only 13 samples of chain C remained completely closed, while the other 3 samples showed weakened posterior segment association and were recorded as posterior segment breakage. In the control group, 8 samples showed anterior segment breakage and 8 samples showed posterior segment breakage, exhibiting typical breakage site drift. This heatmap further supports the conclusion that chain A was preserved and chain C was eliminated.

[0055] The closed correlation index is expressed by the following formula:

[0056]

[0057] In this embodiment, the following is taken , .

[0058] The calculation process of chain A is as follows:

[0059] The calculation process of B-chain is as follows:

[0060] The calculation process of the C-chain is as follows:

[0061] The calculation results above show that chains A, B, and C all reached a value above 1.45 in the initial screening stage, therefore they can all proceed to the next round of screening. However, considering... Figure 3 Further analysis reveals that chain A has the highest closure correlation index and no longer closes after replacement. Although chain B passed the initial screening, it can still maintain closure after replacement. Although chain C reached the threshold, its overall stability is relatively low. This is consistent with the results of subsequent uniqueness screening and consistency analysis.

[0062] The non-compensable fracture discrimination coefficient is adopted by the following formula:

[0063]

[0064] In this embodiment, the following is taken , .

[0065] When chain A breaks at the fixed front segment in the control group, the correlation strength on the front side before the break is taken. Theoretical reassortment correlation strength after replacing the side features Number of successful reclosures ,but

[0066] The value is low and This indicates that the anterior segment fracture in the control group could not be repaired by posterior replacement and was an uncompensable fracture; therefore, the A chain was retained.

[0067] C-chain was taken in the control group. , , ,but

[0068] The presence of one reconstruction closure and the drift of the breakpoint between samples indicates that the C chain has a bypass compensation space and does not possess the stable breakpoint characteristics of the control group; therefore, it was excluded. Figure 4 The sample-level state distribution shown indicates that chain A corresponds to a fixed front-end fracture that is uncompensable, while chain C corresponds to a drifting fracture that is potentially compensable.

[0069] After candidate difference screening, closure matching, closure structure uniqueness test, break site identification, bypass compensation test, and intragroup consistency screening, this embodiment ultimately retained only the A chain, namely the multi-omics closure feature group of Bacteroides, butyric acid, and ZO-1. This feature group showed stable closure in the antibiotic-free feeding group and fixed anterior segment breakage in the control feeding group, and the breakage could not be compensated by replacing the posterior feature. Therefore, it simultaneously possesses a clear biological link, stable intergroup differences, and good intragroup consistency.

[0070] Example 2:

[0071] In this embodiment, a rabbit breeding and fattening integrated farm conducted a 45-day production validation of an antibiotic-free formula during the autumn-winter transition. Previous antibiotic-free conversion trials at the farm showed that, while some chains in the characteristic chains screened based solely on closed-structure analysis maintained statistically continuous correlations in their antecedent, middle, and consequent terms, they did not simultaneously demonstrate mucosal benefits during actual necropsy. Therefore, the farm's technical team decided to introduce host permission gating and permission reversal screening in this batch to improve the consistency between the discrimination results and actual production performance. During the experiment, the same technician continuously recorded fecal pellet morphology, feeding rhythm, and cecal mucosal observations. After the nighttime temperature drop on day 24, three rabbits in the control feeding group exhibited delayed morning feeding and slightly sticky fecal pellet edges, while the overall condition of the antibiotic-free feeding group remained relatively stable. Therefore, sampling and verification were continued according to a unified protocol after the feeding cycle ended.

[0072] In this embodiment, 30 healthy Ira rabbits from the same batch were randomly divided into an antibiotic-free feeding group and a control feeding group, with 15 rabbits in each group, based on ear number. Both groups had the same basal nutritional level. The antibiotic-free feeding group received a regimen of plant extracts, acidifiers, and compound enzymes instead of conventional antibiotics, while the control feeding group used the original formula from the farm. Cecal contents and cecal tissue were collected uniformly between 9:00 AM and 10:00 AM on day 45. Cecal contents were used for 16S sequencing and non-targeted metabolomics detection, while cecal tissue was used to detect indicators such as Occludin, MUC2, ZO-1, TNF-α, and IL-6. Based on the basic closure screening in this embodiment, two candidate closure feature groups were obtained. Permitted strand P consisted of *Rosebria*, butyric acid, and Occludin; permitted strand Q consisted of *Bacteroides*, indolepropionic acid, and MUC2. To verify false closure, a control strand that could structurally close but whose subsequent components were unfavorable to the host was also verified. This strand consisted of *Bacteroides*, indolepropionic acid, and TNF-α. The core testing data are shown in Table 3. Figure 5 The data in Table 3 are presented in a single-figure grouped bar chart format. In the antibiotic-free feeding group, the permitted chain P showed 6.8% of Rosbrinodia, 3.94% of butyrate, and 1.88% of Occludin, while in the control feeding group, the corresponding values ​​were 4.1%, 2.36%, and 1.07%, respectively. The differences in the first, middle, and last terms were 2.7 percentage points, 1.58%, and 0.81%, respectively. The corresponding differences in permitted chain Q were only 0.7 percentage points, 0.15%, and 0.08%. Although the control chain R was consistent with permitted chain Q in the first and middle terms, its last term TNF-α was 1.21% in the antibiotic-free feeding group and 1.39% in the control feeding group, indicating a continuous increase in inflammatory stimulation. Figure 5 Therefore, it can intuitively show that the permission chain P has the greatest range of changes across the three levels, which is more in line with the retention direction of subsequent host permission gating and permission reversal screening.

[0073]

[0074] Host permission gating is based on whether the local physiological state of the cecum changes in a favorable direction. In this embodiment, enhanced tight junction proteins, intact mucus layer, and reduced inflammatory stimulation are used as criteria for permission approval, while increased mucosal permeability risk, enhanced inflammatory factors, or impaired local homeostasis are used as criteria for permission rejection. In permission chain P, Occludin was 1.88 in the antibiotic-free group and 1.07 in the control group. Simultaneously, the IL-6 signal intensity was 0.73 in the antibiotic-free group and 1.26 in the control group. Autopsy revealed a more intact mucosal surface in the antibiotic-free group, indicating that this chain passed the host permission gating in the antibiotic-free group but not in the control group. In permission chain Q, MUC2 was 1.74 in the antibiotic-free group and 1.66 in the control group, with a small difference. No significant mucosal damage was observed in either group, indicating that this chain showed a certain host benefit trend in both groups, but the difference was insufficient. Although the control chain R was structurally closureable, its subsequent TNF-α levels were 1.21 in the antibiotic-free group and 1.39 in the control group, both indicating elevated inflammatory stimulation and failing to demonstrate barrier benefit. Therefore, it was defined as a pseudo-closure characteristic group and excluded. Based on a comprehensive score combining barrier indices, inflammatory indices, and tissue observation, the host permission gating score of the permission chain P was 2.91 in the antibiotic-free group and 0.62 in the control group, while the score of the permission chain Q was 2.14 in the antibiotic-free group and 1.95 in the control group.

[0075] In the uniqueness test of licensing results, only the antecedent and middle terms are fixed, without changing their source and direction, and then the subsequent term is replaced. For license chain P, keeping *Rosburyia* and butyric acid unchanged, after replacing Occludin with ZO-1, the host licensing gating score of the antibiotic-free feeding group was 1.86, and that of the control feeding group was 1.51; after further replacement with Claudin-1, the score of the antibiotic-free feeding group was 1.42, and that of the control feeding group was 1.18. This shows that the significant licensing advantage previously brought by Occludin could not be stably maintained after replacing the subsequent term, indicating that the licensing result of license chain P depends on a specific subsequent term and is unique. For license chain Q, keeping *Bacteroides* and indolepropionic acid unchanged, after replacing MUC2 with Occludin, the host licensing gating score of the antibiotic-free feeding group was 2.03, and that of the control feeding group was 1.81; after further replacement with ZO-1, the score of the antibiotic-free feeding group was 1.92, and that of the control feeding group was 1.73. This result indicates that the licensed state of licensed chain Q can be repeatedly implemented by different successors, and its licensed result is not unique; therefore, it is not retained as the final target chain. The uniqueness test of the licensed result is shown in Table 4. Figure 6The changes in gating scores for both the original and replaced chains in the two groups are simultaneously displayed as a line graph. The original licensed chain P had gating scores of 2.91 and 0.62 in the antibiotic-free and control feeding groups, respectively, with a difference of 2.29. After replacing with ZO-1, the difference decreased to 0.35, and after replacing with Claudin-1, it further decreased to 0.24. In contrast, the original difference for the licensed chain Q was only 0.19. After replacing with Occludin and ZO-1, it still maintained an approximate licensed state of 2.03 to 1.81 and 1.92 to 1.73, respectively, indicating that its licensed result is easily replicated by different subsequent terms. Figure 6 Therefore, it can be seen that the permission chain P has the uniqueness of the permission result, while the permission chain Q does not have the uniqueness.

[0076]

[0077] In the permission reversal discrimination, only the closed feature group, where the antibiotic-free feeding group passed the host permission gating and the control feeding group did not, was retained. In the antibiotic-free feeding group, the permission chain P showed increased abundance of *Rosbria*, enhanced butyrate accumulation, and a significant upregulation of Occludin, accompanied by improved cecal mucosal continuity. In the control feeding group, although the anterior and middle terms were still detectable, the increase in Occludin was limited, and the inflammatory background was high, thus forming a clear permission reversal. The permission chain Q showed a certain host benefit trend in both groups, and the gating results were close to the permission-passing state; therefore, this chain did not have significant inter-group discrimination ability. Although structural correlation could be calculated for the control chain R, its subsequent term consistently pointed to increased inflammation, so it was not included in the permission reversal comparison.

[0078] The permission chain P, which passed the permission reversal criterion, was further tested for uniqueness. Keeping *Rosburyia* and butyric acid unchanged, when the term *Occludin* was replaced with ZO-1, the gating scores for the antibiotic-free feeding group and the control feeding group were 1.86 and 1.51, respectively. Both groups were close to the permission boundary, and the clear opposition of one passing and one failing was no longer formed. Further replacement with MUC2 resulted in a score of 1.69 for the antibiotic-free feeding group and 1.47 for the control feeding group. Although a difference still existed, the control feeding group no longer consistently exhibited a permission failure. This indicates that the reversal relationship of permission chain P depends on the feature of the term *Occludin*, and the original reversal pattern cannot be continuously reproduced after replacement; therefore, its reversal relationship is unique. Permission chain Q, because it did not initially form a clear reversal, and both groups remained close to the permission boundary after replacement, was not included in the final retention set.

[0079] This embodiment uses the permitted reversal strength index to evaluate the inter-group discrimination ability of closed feature groups, and its formula is:

[0080]

[0081] in, For the permissible reversal strength index, The host permission gating score for the closed feature group corresponding to the antibiotic-free feeding group. The host permission gating score of the corresponding closed feature group in the control feeding group is used as an example. The strength of the association between the mid-term characteristic and the subsequent characteristic in the antibiotic-free feeding group. To compare the correlation strength between the middle and subsequent characteristics in the feeding groups, To adjust the parameters. In this embodiment, we take... .

[0082] The permission chain P is calculated as follows. Host permission gating score for the antibiotic-free feeding group. Host permission gating score compared to the feeding group The correlation strength between the middle term and the subsequent term in the antibiotic-free feeding group The correlation strength between the middle and subsequent terms in the control feeding group After substituting, we get

[0083] The results show that there is a clear permission reversal between the two groups of permission chain P, and the connection between the middle item and the next item is significantly different, demonstrating strong distinguishing ability.

[0084] The permission chain Q is calculated as follows. Host permission gating score for the antibiotic-free feeding group. Host permission gating score compared to the feeding group The correlation strength between the middle term and the subsequent term in the antibiotic-free feeding group The correlation strength between the middle and subsequent terms in the control feeding group After substituting, we get

[0085] The results indicate that although the permission chain Q remained closed and showed a certain host benefit trend in both groups, the difference in host permission gating between the antibiotic-free feeding group and the control feeding group was small, and no clear permission reversal was formed, thus the ability to distinguish was insufficient. Figure 7 The permission reversal strength index and gating score distribution of permission chains P and Q under original and replaced consequent conditions are presented in radar chart format. The chart shows that permission chain P exhibits a significant opening amplitude under the original consequent condition (2.91 in the antibiotic-free feeding group and 0.62 in the control feeding group), while under the replaced consequent condition, this opening amplitude narrows to 1.86 vs. 1.51 and 1.69 vs. 1.47, respectively. Permission reversal strength index of permission chain Q is only 0.11941, and the gating scores of the two groups are generally similar under different consequent conditions, indicating that although it has a certain host benefit trend, it lacks a clear and unique permission reversal characteristic.

[0086] After host permission gating discrimination, permission result uniqueness test, permission reversal discrimination, and permission reversal uniqueness test, this embodiment ultimately retains only the permission chain P, namely the permission reversal closure feature group of *Rosburyia*, butyric acid, and Occludin. Under antibiotic-free feeding conditions, this chain exhibits a continuous transmission relationship characterized by increased dominant bacterial genera, accumulation of key short-chain fatty acids, and enhanced tight junction proteins. Under control feeding conditions, it cannot complete the same degree of host permission closure, and the original reversal result cannot be stably reproduced after replacing the latter term. This indicates that it possesses both host benefit orientation and uniqueness of the reversal relationship.

Claims

1. A method for screening multi-omics features of the cecum in meat rabbits for judging the effect of antibiotic-free feeding, characterized in that, include: Cecal samples were collected from rabbits in the antibiotic-free feeding group and the control feeding group to obtain microbiome data, metabolome data and host response omics data; Candidate microbial characteristics, candidate metabolite characteristics, and candidate host response characteristics were screened based on intergroup differences. According to the preset correspondence rules, the candidate features are matched in terms of the antecedent, middle and posterior terms to obtain closed feature groups, and candidate features that do not form closed feature groups are removed; then, based on the inter-group specificity and intra-group consistency of the closed feature groups between the antibiotic-free feeding group and the control feeding group, multi-omics features of the rabbit cecum used to judge the effect of antibiotic-free feeding are screened.

2. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to claim 1, characterized in that, After the closure matching is performed, the uniqueness of the closed structure of the closed feature group is checked; the first or middle feature is fixed, and the remaining features are replaced and matched again; if a closed feature group is still formed after the replacement, it is determined to be a replaceable closed feature group and is removed. Only retain non-replaceable closed feature groups.

3. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to claim 1 or 2, characterized in that, When screening based on the specificity between groups, the breakpoints of the anterior, middle, and posterior feature chains corresponding to the closed feature group in the control feeding group are identified; if there is an interruption between the anterior and middle features, it is determined to be an anterior break; if there is an interruption between the middle and posterior features, it is determined to be a posterior break; only closed feature groups that have fixed breakpoints in the control feeding group and remain closed in the antibiotic-free feeding group are retained.

4. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to any one of claims 1 to 3, characterized in that, The latter feature is used for host permission gating discrimination; if the latter feature indicates that the cecal mucosal barrier is maintained or local homeostasis is maintained, the corresponding closure feature group is determined to have passed the host permission gating; if the latter feature indicates that the mucosal damage, homeostasis imbalance or increased inflammatory stimulation is indicated, the corresponding closure feature group is determined to be a pseudo-closure feature group and is removed.

5. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to claim 3, characterized in that, The term "fixed breakpoint" refers to the fact that the same feature chain is interrupted between the preceding and middle features, or between the middle and following features, in the control feeding group. If both the preceding and following breaks occur simultaneously, it is determined to be a feature chain with a drifting breakpoint and is removed.

6. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to claim 3 or 5, characterized in that, A bypass compensation test is performed on the feature chain with the front or rear segment fracture; the feature on the front side of the fracture site is kept unchanged, and the feature on the rear side is replaced and the closure matching is performed again; if a new closed feature group is formed after the replacement, it is determined to be a compensable fracture feature chain and is removed. Only retain the non-compensable fracture feature chains.

7. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to any one of claims 1 to 6, characterized in that, Intragroup consistency refers to the same closed feature group maintaining a consistent closed state within the antibiotic-free feeding group and maintaining a consistent breakpoint or consistent host permission gating result within the control feeding group; closed feature groups with inconsistent intragroup states are removed.

8. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to claim 4, characterized in that, Perform a uniqueness check on the closed feature group that passes the host permission gating; keep the first and second features unchanged, replace the second feature and re-perform the host permission gating judgment; if it still passes the host permission gating after replacement, it is determined to be a non-unique closed feature group and is removed.

9. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to claim 4 or 8, characterized in that, Permission reversal discrimination was performed on closed feature groups with the same antecedent and median features in the antibiotic-free feeding group and the control feeding group. If the antibiotic-free feeding group passed the host permission gating but the control feeding group failed, it was identified as a permission reversal closed feature group and retained. If the host permission gating results of the two groups were the same, they were removed.

10. The method for screening multi-omics features of rabbit cecum for judging the effect of antibiotic-free feeding according to claim 9, characterized in that, Perform a uniqueness test on the permission inversion closed feature group; keep the first and second features unchanged, replace the second feature and perform permission inversion judgment again; if the permission inversion judgment is still satisfied after replacement, it is determined to be a non-unique permission inversion closed feature group and is removed.

Images