A genomic-based bacterial risk annotation method and system

Abstract

The application provides a genome-based bacterial risk annotation method and system. The system integrates four dimensions of virulence / drug resistance annotation, genome stability analysis, harmful metabolism gene analysis and phenotype prediction into the same risk assessment framework. The multi-dimensional integrated architecture can significantly reduce the information omission probability in risk assessment, and provide more complete genotype-phenotype correlation evidence for bacterial safety evaluation.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics technology, specifically the field of bacterial genome analysis, and specifically relates to a genome-based bacterial risk annotation method and system. Background Technology

[0002] Bacterial safety risk assessment is a core component of environmental microbial monitoring, clinical infection control, food safety assurance, and pharmaceutical manufacturing process control. Traditional bacterial risk assessment primarily relies on phenotypic experiments such as culture identification, biochemical tests, antibiotic susceptibility testing, and virulence verification. While these methods provide reliable results, they suffer from drawbacks such as long operation cycles, low throughput, and high labor costs. Furthermore, they are limited by culturable conditions and experimental scale, resulting in a narrow phenotypic range that can be covered in a single assessment. This makes it difficult to meet the practical needs of rapid screening of large numbers of strains and the identification of potential risks in the era of high-throughput sequencing.

[0003] With the widespread adoption of high-throughput sequencing technology and the development of bioinformatics methods, directly obtaining bacterial risk and phenotypic information from genome sequences has become an important technological direction. Currently, researchers and institutions have developed several genome-based bacterial risk assessment or annotation methods. For example, some methods predict bacterial drug resistance characteristics at the genome level by comparing protein sequences with drug resistance gene databases to determine whether a strain carries known drug resistance genes. However, these methods only focus on the single dimension of drug resistance and do not address other important safety factors such as virulence factors and metabolite risks. Other methods focus on the identification and annotation of bacterial virulence factors, capable of identifying known virulence proteins encoded in the genome. However, they also suffer from the problem of a single assessment dimension, failing to reflect the potential risks of strains in terms of genome stability and metabolites. Therefore, it is necessary to provide a multi-dimensional, comprehensive risk annotation method. Summary of the Invention

[0004] To address the issue of existing genome-based bacterial risk annotation methods having limited assessment dimensions, this invention provides a bacterial risk annotation method and system that encompasses four dimensions: virulence / resistance, genome stability, harmful metabolic genes, and phenotypic prediction.

[0005] The technical solution adopted in this invention is:

[0006] In a first aspect, the present invention provides a genome-based bacterial risk annotation system, comprising: The virulence and resistance annotation module is used to align the predicted protein sequences from the genome to the virulence factor database and the resistance gene database, and to filter the alignment results based on the first similarity threshold and the first coverage threshold to obtain virulence factor annotation information and resistance gene annotation information. The genome stability analysis module is used to establish a mobile element database covering insert sequences, integrons, transposons, and replicons. The genome sequence is aligned to the mobile element database, and the alignment results are filtered based on a second similarity threshold and a second coverage threshold to obtain mobile element annotation information. The harmful metabolism gene analysis module is used to establish a harmful metabolism gene database covering ammonia metabolism genes, amine metabolism genes, phenol metabolism genes, bile acid metabolism genes, mucin degradation genes, carcinogen activation genes, and toxin-related genes. The genomic sequence is aligned to the harmful metabolism gene database to obtain harmful metabolism gene annotation information. The phenotypic prediction module is used to annotate protein families in genome coding sequences and predict phenotypes based on the phylogenetic distribution patterns of protein families, obtaining phenotypic prediction information including pathogenicity-related phenotypes, growth condition phenotypes, enzyme activity phenotypes, and sugar fermentation phenotypes. The risk annotation results module is used to summarize the virulence factor annotation information, drug resistance gene annotation information, mobile element annotation information, harmful metabolic gene annotation information, and phenotypic prediction information to obtain risk annotation results.

[0007] Existing genome-based bacterial risk annotation methods generally suffer from a lack of dimensional assessment. Each approach focuses on either drug resistance genes or virulence factors, independently annotating movable elements and predicting phenotypes, resulting in a disconnect between these dimensions. In particular, systematic annotation of harmful metabolic genes (including genes related to ammonia, amine, phenol, bile acid metabolism, and toxin synthesis) is completely absent from existing technologies, even though the metabolites of these genes may pose substantial harm to humans or the environment. To address this, this invention constructs a four-dimensional integrated analysis system from genotype to phenotype. Virulence and drug resistance annotation identifies known virulence and drug resistance genes; genome stability analysis assesses genome instability and horizontal gene transfer potential through movable element analysis; harmful metabolic gene analysis, for the first time, incorporates harmful metabolic genes into the annotation system, covering the gaps in the metabolic risk dimension of existing methods; and phenotype prediction predicts the pathogenicity, growth conditions, and enzyme activity characteristics of strains at the phenotypic level, creating a complementary relationship between genotype annotation and actual phenotypic potential. The four dimensions mentioned above form a progressive analytical chain, from known risks to potential risks, and from genotype to phenotype. This upgrades risk assessment from static gene list judgment to dynamic genotype-phenotype association reasoning, significantly reducing the blind spots in risk assessment, which is something that existing single-dimensional or two-dimensional solutions cannot achieve.

[0008] Preferably, the harmful metabolic gene analysis module is further configured to: align the genome sequence to harmful metabolic gene data from multiple database sources; cross-validate the annotation results obtained from multiple sources for the same gene; assign a weighted score to the gene based on the weight of the database source and the weight of the functional category; accumulate the weighted score when the same gene is annotated by multiple database sources; and classify the annotated harmful metabolic genes according to the score, outputting a graded risk list containing entries of different risk levels; wherein the weight of the database source is set according to the database credibility, and the weight of the functional category is set according to the degree of harm of the functional category.

[0009] Preferably, the risk classification is as follows: a score greater than 0.7 is classified as high risk, a score greater than 0.5 and not greater than 0.7 is classified as medium risk, and a score not greater than 0.5 is classified as low risk.

[0010] Preferably, the weight coefficient of the database source weighted score is 0.15 to 0.25; the weight setting method of the database source includes: SwissProt weight is 0.9 to 1.0, eggNOG weight is 0.8 to 0.95, KEGG weight is 0.7 to 0.9, and the weight of SwissProt is not lower than the weight of eggNOG, and the weight of eggNOG is not lower than the weight of KEGG; the weight coefficient of the functional category weighted score is 0.25 to 0.35; the weight setting method of the functional category includes: secondary bile acid metabolism and carcinogen activation category weight is 0.9 to 1.0, mucin degradation and phenol metabolism category weight is 0.8 to 0.95, amine metabolism and ammonia metabolism category weight is 0.6 to 0.85, and toxin-related category weight is 0.5 to 0.7.

[0011] Preferably, the first similarity threshold is 85% to 98%, and the first coverage threshold is 75% to 95%; more preferably, the first similarity threshold is 90%, and the first coverage threshold is 80%.

[0012] Preferably, the virulence factor database is the VFDB (Virulence Factor Database), and the drug resistance gene database is the CARD (The Comprehensive Antibiotic Resistance Database); the virulence and drug resistance annotation module uses the ABRicate tool for the comparison.

[0013] Preferably, the second similarity threshold is 75% to 95%, and the second coverage threshold is 50% to 80%; more preferably, the second similarity threshold is 80%, and the second coverage threshold is 60%.

[0014] Preferably, the mobile element database is constructed based on at least one of ISFinder, INTEGRALL, PlasmidFinder, TnCentral, and NCBI drug resistance gene databases, and more preferably, it is constructed in combination; the genome stability analysis module uses Blast software to perform the alignment; the genome stability analysis module is also used to generate a visual map of the mobile element regions in the genome based on the location information of the mobile element annotation regions.

[0015] Preferably, the harmful metabolic gene database is constructed based on at least one of the CAZy (Carbohydrate-Active enZYmesdatabase), KEGG (Kyoto Encyclopedia of Genes and Genomes), SwissProt (UniProtKB / Swiss-Prot), and eggNOG (evolutionary genealogy of genes: Non-supervised Orthologous Groups) databases; the harmful metabolic gene analysis module uses at least one of diamond software, hmmscan software, and emapper software to align the genome sequence to the harmful metabolic gene database.

[0016] Preferably, the phenotypic prediction module uses HMMer software to annotate the protein family, which is the Pfam protein family; the phenotypic prediction module uses a pre-trained model built into the Traitar software, and performs phenotypic prediction based on a dual-model strategy of the phypat model and the phypat+PGL model. When the prediction results of the two models are consistent, it is judged as positive.

[0017] Preferably, the pathogenicity-related phenotypes include hemolysis and coagulase production; the growth condition phenotypes include temperature, salt concentration, culture medium, bile sensitivity or tolerance, carbon source utilization, nitrogen source utilization, glucose gas production, and fermentation reactions; and the enzyme activity phenotypes include hydrolases and oxidoreductases.

[0018] Secondly, the present invention provides a genome-based bacterial risk annotation method, comprising: Virulence and resistance annotation steps: Align the predicted protein sequences from the genome to the virulence factor database and the resistance gene database. Based on the first similarity threshold and the first coverage threshold, filter the alignment results to obtain virulence factor annotation information and resistance gene annotation information. Genome stability analysis steps: Establish a mobile element database covering insert sequences, integrons, transposons, and replicons; align genome sequences to the mobile element database; filter alignment results based on a second similarity threshold and a second coverage threshold; and obtain mobile element annotation information. Harmful metabolic gene analysis steps: Establish a harmful metabolic gene database covering ammonia metabolism genes, amine metabolism genes, phenol metabolism genes, bile acid metabolism genes, and toxin-related genes; align the genome sequence to the harmful metabolic gene database to obtain harmful metabolic gene annotation information; Phenotypic prediction steps: Protein family annotation is performed on the genome coding sequence, and phenotypic prediction is performed based on the phylogenetic distribution pattern of the protein family to obtain phenotypic prediction information including pathogenicity-related phenotypes, growth condition phenotypes, enzyme activity phenotypes, and sugar fermentation phenotypes. Risk annotation result summary steps: Summarize the virulence factor annotation information, drug resistance gene annotation information, mobile element annotation information, harmful metabolic gene annotation information, and phenotypic prediction information to obtain risk annotation results.

[0019] Thirdly, the present invention provides a genome-based bacterial risk annotation device, including a memory and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code to implement the method described in the second aspect above.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: First, this invention is the first to integrate four dimensions—virulence / resistance annotation, genomic stability analysis, harmful metabolic gene analysis, and phenotype prediction—into the same risk assessment framework. Compared to existing methods that only focus on a single or a few dimensions, the multi-dimensional integrated architecture of this invention can significantly reduce the probability of information omission in risk assessment and provide more complete genotype-phenotype association evidence for bacterial safety evaluation.

[0021] Second, this invention introduces systematic annotation of harmful metabolic genes, overcoming the shortcomings of existing methods that focus only on classical virulence factors while neglecting atypical harmful metabolic risks. By constructing a database covering ammonia, amines, phenols, bile acids, and toxin-related metabolic genes, strains that do not express known virulence proteins but still pose potential safety threats are identified, thus expanding the coverage of risk annotation.

[0022] Third, the phenotypic prediction module of this invention not only covers pathogenicity-related phenotypes (such as hemolysis and coagulase production), but also growth and culture conditions, enzyme activity and biochemical phenotypes, as well as sugar fermentation phenotypes, providing multi-faceted information support for the assessment of the physiological characteristics of strains and the prediction of their behavior in specific environments, making the risk assessment results more practically valuable. Attached Figure Description

[0023] Figure 1 This is an overall architecture diagram of a genome-based bacterial risk annotation method proposed in an embodiment of the present invention.

[0024] Figure 2 This is a visual schematic diagram of a movable element in an embodiment of the present invention. Detailed Implementation

[0025] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention.

[0026] It should be noted that, unless otherwise specified, the following embodiments and features can be combined with each other.

[0027] Example 1: Bacterial Risk Annotation Method Based on Staphylococcus aureus ATCC 12600 Genome This embodiment uses the genome sequence of Staphylococcus aureus ATCC 12600 strain (GCA_006094915.1) as an example to illustrate the specific implementation process of the method of the present invention.

[0028] Sample preparation: The complete genome sequence (GCA_006094915.1) of Staphylococcus aureus ATCC 12600 was obtained from the NCBI database.

[0029] Virulence and resistance annotation steps: First, Prokka software was used to predict and annotate the genome sequence to obtain the predicted protein sequences. Then, ABRicate was used to align the protein sequences to the VFDB virulence factor database and the CARD resistance gene database, respectively. After alignment, the results were filtered based on thresholds of at least 90% similarity and at least 80% coverage, yielding the virulence factor annotation results and resistance gene annotation results for this strain. The annotation results are presented in tabular form; some results are shown in Tables 1 and 2.

[0030] Table 1. Annotations on the drug resistance of Staphylococcus aureus ATCC 12600 strains (partial)

[0031] Table 2. Virulence annotation results of Staphylococcus aureus ATCC 12600 strains (partial)

[0032] Referring to Table 1, three drug resistance genes, tet(38), mepR, and mepA, were annotated for this strain, with 100% coverage and over 99% similarity. tet(38) mediates tetracycline resistance. mepR and mepA are adjacent in the genome, forming a regulator-multidrug efflux pump gene pair, which can mediate efflux resistance to fluoroquinolones and disinfectants. It can be seen that the drug resistance mechanism of this strain is mainly based on efflux pumps, suggesting that its active efflux capacity for multiple drugs should be the focus in risk assessment. The annotation results achieved the expected screening accuracy and reliability. Referring to Table 2, virulence factors such as adsA were annotated for this strain. adsA encodes adenosine synthase A, which belongs to the immunomodulatory virulence factor class. It can convert AMP into adenosine, inhibit the bactericidal function of host neutrophils, and help bacteria achieve immune escape. The annotation results not only provide gene names and genomic locations, but also clarify the functional classification and source strains of virulence factors through product description fields. This indicates that the virulence annotation results of this invention are complete and can be directly used for the classification of virulence types and the division of risk levels, providing clear virulence evidence for subsequent comprehensive risk assessment.

[0033] Genome stability analysis steps: Sequence data for inserts (IS), integrons, transposons, and replicons were downloaded from six databases: ISFinder, INTEGRALL, TnCentral, PlasmidFinder, ICEberg 2.0, and NCBI. These four types of mobile element sequences were merged to create a comprehensive mobile element database covering inserts, integrons, transposons, and replicons. The genome sequence of *Staphylococcus aureus* ATCC 12600 was aligned to this mobile element database using Blast software. Alignment results were filtered based on similarity thresholds of at least 80% and coverage thresholds of at least 60% to obtain mobile element annotation results. Partial annotation results for the mobile elements of this strain are shown in Table 3.

[0034] Furthermore, using the matplotlib package in Python, a visual map of the distribution of mobile elements in the genome is plotted based on the annotation location information of mobile elements on the contig sequence. This visually displays the location and density of each mobile element in different regions of the genome, facilitating rapid assessment of genome stability (see [link to documentation]). Figure 2 ).

[0035] Table 3. Annotation results of movable elements in Staphylococcus aureus ATCC 12600 strain (partial).

[0036] Referring to Table 3, this strain was annotated with two types of mobile elements: the inserted sequence ISSep3 and the replicon rep7c. Both similarity and coverage exceeded the set screening thresholds. ISSep3 originated from *Staphylococcus epidermidis*, and rep7c originated from the plasmid backbone of strain MSSA476. Both are exogenous genetic elements, suggesting traces of horizontal gene transfer in the strain's genome. This indicates that this strain has the potential to acquire and spread exogenous drug resistance and virulence genes, posing a potential risk to genome stability. Figure 2 The visualized map shown can intuitively present the distribution and density of each element on the genome, making it easy to quickly assess the risk to genome stability.

[0037] Hazardous metabolic gene analysis steps: Based on four databases—CAZy, KEGG, SwissProt, and eggNOG—retrieved using keywords related to harmful metabolism. Search keywords included, but were not limited to, urea amidolyase, phenolic, bileacid, mucinase, beta-glucuronidase, and toxin, covering genes related to ammonia metabolism, amine metabolism, phenolic metabolism, bile acid metabolism, mucin degradation, carcinogen activation, and toxin synthesis. The retrieved harmful metabolic gene sequences were compiled to construct a comprehensive harmful metabolic gene database. The genome sequence of *Staphylococcus aureus* ATCC 12600 was aligned to the aforementioned harmful metabolic gene database using Diamond, HMMSCAN, and Emapper software, respectively. The annotation results from the three software programs were combined to obtain the classification and annotation information of harmful metabolic genes for this strain. Partial annotation results are shown in Table 4. Multiple harmful metabolic-related genes were annotated for this strain, covering categories such as ammonia metabolism, mucin degradation, carcinogen activation, and toxins. The urease accessory protein gene ureE indicates that this bacterium has ammonia-producing potential, and the nitroreductase gene suggests that it may activate procarcinogens into mutagenic products, thus being rated as high-risk. The annotation results were cross-validated across multiple databases, and scores and risk classifications were provided. High-risk entries were retained, while low-risk results were not simply discarded, forming a graded risk list. This table fully demonstrates the invention's coverage of atypical harmful metabolic risks such as ammonia, amines, phenols, and carcinogen activation, filling a gap in traditional toxicity annotation.

[0038] Furthermore, in this embodiment, the harmful metabolic gene analysis module also performs multi-database cross-validation and risk grading on the annotation results. Specifically, the three software programs Diamond, HMMSCAN, and Emapper align the genome sequence to harmful metabolic gene data from databases such as SwissProt, KEGG, and eggNOG, respectively. For annotation results obtained from multiple sources for the same gene, cross-validation is performed: the score is a weighted sum of the database source weighted score (weight coefficient 0.2) and the functional category weighted score (weight coefficient 0.3), and normalized to a value no greater than 1; the database source is assigned a value according to its reliability: SwissProt is 1, eggNOG is 0.9, and KEGG is 0.8. When the same gene is annotated by multiple databases, the weighted score is accumulated; the functional category is assigned a value according to its hazard level: secondary bile acid metabolism and carcinogen activation are 1.0, mucin degradation and phenol metabolism are 0.9, amine metabolism is 0.7, ammonia metabolism is 0.8, and toxin-related is 0.6. Subsequently, the annotated harmful metabolic genes are risk-classified based on the scores, and a risk classification list is output, retaining entries of different levels. Specifically, the risk classification is based purely on the score thresholds: scores greater than 0.7 are high risk, scores greater than 0.5 but not greater than 0.7 are medium risk, and scores not greater than 0.5 are low risk.

[0039] Taking Table 4 as an example, gene CP035101.1_2083 (urease accessory protein UreE) is annotated by three databases: SwissProt, KEGG, and eggNOG, with consistent functional descriptions from all sources. Its score is 0.85, classifying it as high-risk. Gene CP035101.1_2310 (nitroreductase) is consistently annotated by both SwissProt and eggNOG as being related to oncogen activation, with a score of 0.78, also classifying it as high-risk. Gene CP035101.1_1170 (M16 peptidase) is only annotated by the single eggNOG database, with a lower score of 0.45, but it is still retained in the list and marked as low-risk. This tiered risk list ensures that high-risk items are based on consistent evidence from multiple sources while avoiding the simple discarding of weak risk signals, preserving complete clues for subsequent manual review and risk tracing.

[0040] Table 4. Annotation results of harmful metabolic genes in Staphylococcus aureus ATCC 12600 strain (partial).

[0041] Phenotypic prediction steps: The hmmsearch program in HMMer software was used to align the protein sequences of Staphylococcus aureus ATCC12600 to the Pfam-A HMM database to obtain Pfam protein family annotation information for each coding sequence. Subsequently, using Traitar software, based on a dual-model strategy consisting of the phypat model and the phypat+PGL model, the presence / deletion of 67 phenotypes was predicted according to the distribution pattern of the protein family in phylogeny. When the prediction results of the two models were consistent, it was judged as positive. The predicted phenotypic information included: (1) pathogenicity-related phenotypes, such as hemolysis and coagulase production; (2) growth condition phenotypes, such as temperature range, salt tolerance, culture medium adaptability, bile sensitivity, carbon / nitrogen source utilization, glucose gas production and fermentation reaction; (3) enzyme activity and biochemical phenotypes, such as hydrolases and oxidoreductase activities; (4) sugar fermentation phenotypes. The phenotypic prediction results of this strain are shown in Table 5.

[0042] Table 5. Phenotypic prediction results of Staphylococcus aureus ATCC 12600 strain

[0043] Referring to Table 5, the phenotypic prediction results of this strain cover four dimensions: In terms of pathogenicity, it is positive for both β-hemolysis and coagulase production, exhibiting typical pathogenic characteristics; in terms of growth conditions, it can grow at 42°C and 6.5% NaCl and is a facultative anaerobe, demonstrating strong environmental adaptability; in terms of enzyme activity, it is positive for multiple enzymes including urease and catalase; and in terms of sugar fermentation, it can utilize various carbon sources such as glucose, lactose, and sucrose. Notably, the cocci-like structure, clustered arrangement, Gram-positive appearance, and yellow pigment morphology perfectly match the known characteristics of Staphylococcus aureus, validating the accuracy of the prediction; simultaneously, the positive urease phenotype corroborates the annotation results of the ureE gene in Table 4, demonstrating the comprehensive evaluation value of combining genotype and phenotype.

[0044] Risk annotation results summary steps: Based on the analysis results of the above four steps, a multi-dimensional risk annotation report for Staphylococcus aureus ATCC 12600 strain was obtained: This strain carries multiple drug resistance genes and virulence factors (Step 2), and its genome contains mobile elements such as insertion sequences and transposons, suggesting a certain degree of genomic instability and horizontal gene transfer potential (Step 3). Several harmful metabolism-related genes were also annotated (Step 4). Phenotypic prediction results indicate that it is hemolytic, can grow over a wide temperature and salt concentration range, and possesses multiple sugar fermentation capabilities (Step 5). The above multi-dimensional risk information provides systematic genotype-phenotype evidence for a comprehensive assessment of the safety risks of this strain.

[0045] This invention also provides a genome-based bacterial risk annotation system. This embodiment illustrates the specific structure of the genome-based bacterial risk annotation system provided by this invention. The system comprises five parts: a virulence and resistance annotation module, a genome stability analysis module, a harmful metabolic gene analysis module, a phenotypic prediction module, and a risk annotation result module.

[0046] The virulence and resistance annotation module receives the genome sequence of the bacteria to be tested. First, it uses gene prediction software (such as Prokka) to obtain the predicted protein sequences. Then, it aligns the protein sequences to a virulence factor database (such as the VFDB database) and a resistance gene database (such as the CARD database). Based on a first similarity threshold and a first coverage threshold, the alignment results are filtered to obtain virulence factor annotation information and resistance gene annotation information. In one specific embodiment, the first similarity threshold is 90%, the first coverage threshold is 80%, and the alignment is performed using the ABRicate tool, as described in step two of Example 1.

[0047] The genome stability analysis module is used to jointly acquire sequence data of inserts, integrons, transposons, and replicons from drug resistance gene databases in ISFinder, INTEGRALL, PlasmidFinder, TnCentral, and NCBI, establishing a mobile element database covering these four types of elements. Genome sequences are aligned to this mobile element database, and alignment results are filtered based on a second similarity threshold and a second coverage threshold to obtain mobile element annotation information. Based on the location information of the annotated mobile element regions, a visual map of the mobile element regions in the genome is generated. In one specific embodiment, the second similarity threshold is 80%, the second coverage threshold is 60%, the alignment is performed using Blast software, and the visualization map is drawn using the matplotlib package in Python. The specific implementation process is as described in step three of Example 1.

[0048] The harmful metabolism gene analysis module is used to retrieve sequence data covering ammonia metabolism genes, amine metabolism genes, phenol metabolism genes, bile acid metabolism genes, toxin-related genes, mucin degradation genes, and carcinogen activation genes from at least one of the CAZy, KEGG, SwissProt, and eggNOG databases by searching for harmful metabolism-related keywords, thus establishing a harmful metabolism gene database. Genomic sequences are then aligned to the harmful metabolism gene database to obtain harmful metabolism gene annotation information. In one specific embodiment, the alignment is performed using Diamond, HMMSCAN, and Emapper software, as described in step four of Example 1. In another specific embodiment, this module also performs multi-database cross-validation and risk grading on the annotation results. Specifically, Diamond, HMMSCAN, and Emapper software align genomic sequences to harmful metabolism gene data from databases such as SwissProt, KEGG, and eggNOG, respectively, and cross-validate the annotation results obtained from multiple sources for the same gene: the specific process of weighted scoring and risk grading is as described in step four of Example 1.

[0049] The phenotypic prediction module is used to annotate protein families in genomic coding sequences (e.g., using HMMer software to align protein sequences to the Pfam-A HMM database) and predict phenotypes based on the phylogenetic distribution patterns of protein families (e.g., using Traitar software, based on a dual-model strategy of the phypat model and the phypat+PGL model; when the prediction results of the two models are consistent, it is judged as positive). Phenotypic prediction information including pathogenicity-related phenotypes, growth condition phenotypes, enzyme activity phenotypes, and sugar fermentation phenotypes is obtained. The specific implementation process is as described in step five of Example 1.

[0050] The risk annotation results module is used to summarize the virulence factor annotation information, drug resistance gene annotation information, mobile element annotation information, harmful metabolic gene annotation information, and phenotypic prediction information to generate a multi-dimensional risk annotation results report. The specific implementation process is as described in step six of Example 1.

[0051] Each of the above modules can be implemented by a computer program running on one or more processors, and the computer program is stored in memory; data transfer between modules can be accomplished through shared memory or data interfaces. Those skilled in the art should understand that the above module division is only a logical functional division, and in actual implementation, there may be other division methods, such as multiple modules being combined or integrated into another system.

[0052] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope of the present invention.

Claims

1. A genome-based bacterial risk annotation system, characterized in that, include: The virulence and resistance annotation module is used to align the predicted protein sequences from the genome to the virulence factor database and the resistance gene database, and to filter the alignment results based on the first similarity threshold and the first coverage threshold to obtain virulence factor annotation information and resistance gene annotation information. The genome stability analysis module is used to establish a mobile element database covering insert sequences, integrons, transposons, and replicons. The genome sequence is aligned to the mobile element database, and the alignment results are filtered based on a second similarity threshold and a second coverage threshold to obtain mobile element annotation information. The harmful metabolism gene analysis module is used to establish a harmful metabolism gene database covering ammonia metabolism genes, amine metabolism genes, phenol metabolism genes, bile acid metabolism genes, mucin degradation genes, carcinogen activation genes, and toxin-related genes. The genomic sequence is aligned to the harmful metabolism gene database to obtain harmful metabolism gene annotation information. The phenotypic prediction module is used to annotate protein families in genome coding sequences and predict phenotypes based on the phylogenetic distribution patterns of protein families, obtaining phenotypic prediction information including pathogenicity-related phenotypes, growth condition phenotypes, enzyme activity phenotypes, and sugar fermentation phenotypes. The risk annotation results module is used to summarize the virulence factor annotation information, drug resistance gene annotation information, mobile element annotation information, harmful metabolic gene annotation information, and phenotypic prediction information to obtain risk annotation results.

2. The system according to claim 1, characterized in that, The harmful metabolic gene analysis module is further used for: aligning genome sequences to harmful metabolic gene data from multiple database sources; cross-validating the annotation results obtained from multiple sources for the same gene; assigning a weighted score to the gene based on the weight of the database source and the weight of the functional category; accumulating the weighted score when the same gene is annotated by multiple database sources; the weighted score is calculated by weighting the database source weighted score and the functional category weighted score according to their respective weight coefficients, and normalizing it to a value not greater than 1; and classifying the annotated harmful metabolic genes for risk based on the score, outputting a graded risk list containing entries of different risk levels; wherein, the weight of the database source is set according to the database reliability, and the weight of the functional category is set according to the degree of harm of the functional category.

3. The system according to claim 2, characterized in that, The risk classification is as follows: a score greater than 0.7 is classified as high risk, a score greater than 0.5 but not greater than 0.7 is classified as medium risk, and a score not greater than 0.5 is classified as low risk.

4. The system according to claim 2, characterized in that, The weighting coefficient of the weighted score of the database source is 0.15 to 0.25; the weighting method of the database source includes: the weight of SwissProt is 0.9 to 1.0, the weight of eggNOG is 0.8 to 0.95, and the weight of KEGG is 0.7 to 0.9, and the weight of SwissProt is not lower than the weight of eggNOG, and the weight of eggNOG is not lower than the weight of KEGG; The weight coefficient of the functional category weighted score is 0.25 to 0.35; the weight setting method of the functional category includes: the weight of secondary bile acid metabolism and carcinogen activation category is 0.9 to 1.0, the weight of mucin degradation and phenol metabolism category is 0.8 to 0.95, the weight of amine metabolism and ammonia metabolism category is 0.6 to 0.85, and the weight of toxin-related category is 0.5 to 0.

7.

5. The system according to claim 1, characterized in that, The virulence factor database is the VFDB database, and the drug resistance gene database is the CARD database; the virulence and drug resistance annotation module uses the ABRicate tool for the comparison.

6. The system according to claim 1, characterized in that, The mobile element database is constructed based on at least one of ISFinder, INTEGRALL, PlasmidFinder, TnCentral, and NCBI drug resistance gene database; the genome stability analysis module is also used to generate a visual map of the mobile element regions in the genome based on the location information of the mobile element annotation regions.

7. The system according to claim 1, characterized in that, The harmful metabolic gene database is constructed based on at least one of the CAZy, KEGG, SwissProt and eggNOG databases; the harmful metabolic gene analysis module uses at least one of diamond software, hmmscan software and emapper software to align the genome sequence to the harmful metabolic gene database.

8. The system according to claim 1, characterized in that, The phenotypic prediction module uses HMMer software to annotate the protein family, which is the Pfam protein family. The phenotypic prediction module uses a dual-model strategy based on the phypat model and the phypat+PGL model to predict the phenotypic pattern. When the prediction results of the two models are consistent, the result is considered positive.

9. A genome-based bacterial risk annotation method, characterized in that, include: Virulence and resistance annotation steps: Align the predicted protein sequences from the genome to the virulence factor database and the resistance gene database. Based on the first similarity threshold and the first coverage threshold, filter the alignment results to obtain virulence factor annotation information and resistance gene annotation information. Genome stability analysis steps: Establish a mobile element database covering insert sequences, integrons, transposons, and replicons; align genome sequences to the mobile element database; filter alignment results based on a second similarity threshold and a second coverage threshold; and obtain mobile element annotation information. Harmful metabolic gene analysis steps: Establish a harmful metabolic gene database covering ammonia metabolism genes, amine metabolism genes, phenol metabolism genes, bile acid metabolism genes, mucin degradation genes, carcinogen activation genes, and toxin-related genes; align the genome sequence to the harmful metabolic gene database to obtain harmful metabolic gene annotation information; Phenotypic prediction steps: Protein family annotation is performed on the genome coding sequence, and phenotypic prediction is performed based on the phylogenetic distribution pattern of the protein family to obtain phenotypic prediction information including pathogenicity-related phenotypes, growth condition phenotypes, enzyme activity phenotypes, and sugar fermentation phenotypes. Risk annotation result summary steps: Summarize the virulence factor annotation information, drug resistance gene annotation information, mobile element annotation information, harmful metabolic gene annotation information, and phenotypic prediction information to obtain risk annotation results.

10. A genome-based bacterial risk annotation device, characterized in that, The device includes a memory and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code to implement the method of claim 9.

Images