A food exogenous chemical genetic toxicity risk screening method based on multi-source evidence fusion

Abstract

The application provides a food exogenous chemical genetic toxicity risk screening method based on multi-source evidence fusion, relates to the technical field of food safety and risk assessment, and comprises the following steps: S1: determining the type of food exogenous chemicals to be analyzed, obtaining food samples to be processed, and obtaining exogenous chemical residue data; S2: obtaining the experimental data score of each chemical by using a genetic toxicity experiment database, and calculating the QSAR prediction score of each chemical by using a QSAR tool; S3: according to the preprocessed experimental data score and the QSAR prediction score, the entropy weight method and the CRITIC method are used for comprehensive weighting; S4: calculating the hazard score of each chemical; S5: combining the hazard score with the residue data, and using a four-quadrant matrix diagram and a comprehensive risk score for safety evaluation; the application can be used for risk priority ranking and safety screening of new food and complex additive ingredients.

Description

Technical Field

[0001] This invention relates to the field of food safety risk assessment technology, and more specifically, to a method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion. Background Technology

[0002] With the rapid development of biotechnology and the food industry, novel foods (such as synthetic biology products, cell-engineered foods, novel processing aids, and complex additives) are constantly emerging. During the production and processing of these novel foods, various exogenous chemical substances (such as metabolic regulators, cross-linking agents, surfactants, residual solvents, and preservatives) are typically used or introduced. These exogenous chemicals may leave trace residues in the final product, posing a potential threat to its genomic integrity and long-term safety for consumption.

[0003] Currently, conducting systematic experimental genotoxicity tests (such as Ames tests and chromosome aberration tests) on all potential exogenous chemicals is limited by high costs, long cycles, and a wide variety of chemicals, making it difficult to meet the rapidly evolving needs of new food research and development and regulation. Quantitative structure-activity relationship (QSAR) prediction technology provides an efficient means of chemical hazard identification, but single QSAR models are limited by algorithmic assumptions, domain of applicability, and molecular descriptor selection, making them prone to prediction bias or numerous "undefined" results. Furthermore, existing assessment methods often focus on single hazard identification, failing to effectively integrate information on both hazard and actual exposure residues.

[0004] Therefore, how to establish a priority screening framework for the genotoxicity risk of exogenous chemicals in food that can integrate multi-source theoretical predictions, objective experimental evidence, and actual residual exposure is an urgent problem to be solved in this field. Summary of the Invention

[0005] In view of this, the present invention provides a method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion. It integrates multi-source QSAR prediction results, experimental toxicology database evidence and objective combination weighting strategies, and combines actual residue levels to carry out four-quadrant risk characterization, so as to achieve rapid and accurate risk screening of complex exogenous chemicals in novel foods.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion includes the following steps: S1: Obtain the food sample to be processed, and combine the sample with the type of exogenous chemical to be analyzed to obtain the corresponding chemical residue data; S2: Obtain experimental data scores for each chemical using the genotoxicity experimental database, and calculate the QSAR prediction score for each chemical using the QSAR tool; S3: Based on the experimental data score and QSAR prediction score of each chemical, the corresponding weighting factor is determined by a combined weighting method that combines entropy weighting and CRITIC weighting. S4: Calculate the hazard score for each exogenous chemical based on the preprocessed experimental data score, QSAR prediction score, and weighting factor; S5: Combining the hazard score and chemical residue data, construct a four-quadrant matrix diagram, and use the four-quadrant matrix diagram and comprehensive risk score to conduct a safety risk assessment.

[0007] Preferably, S1 includes: S11: Build a basic information database of exogenous chemicals using Comptox online tools or the Pubchem chemical database; S12: Screen the basic information database in conjunction with food safety standards or new food application guidelines to determine the list of exogenous chemicals to be analyzed; S13: Obtain a food sample to be processed, and perform quantitative analysis of the exogenous chemicals in the sample in conjunction with the list of exogenous chemicals to be analyzed, to obtain the corresponding chemical residue data; Preferably, S2 includes: S21: Collect genotoxicity experimental data from the Endpoint module in the QSAR Toolbox and the CompTox Chemicals Dashboard platform; S22: The evidence amount and positive rate of the genotoxicity experimental data are integrated using a logarithmic weighting method, and the adjusted scores are standardized using a min-max scaling method to obtain the experimental data score for each chemical. S23: Select multiple QSAR evaluation tools with different mechanisms or algorithms, input the exogenous chemical structure information into QSARToolbox, Toxtree, TEST, and VEGA tools for QSAR prediction, and classify the prediction results as positive, negative, or undefined according to the classification criteria; S24: The QASR prediction classification results are processed using the minimum-maximum scaling method to obtain the corresponding QSAR prediction scores.

[0008] Preferably, S3 includes: S31: Calculate the initial weights of the QSAR prediction score using the entropy weight method and the CRITIC weight method respectively, and then perform an arithmetic mean on the two weights, using the arithmetic mean as the weight factor of the QSAR prediction score. S32: Calculate the initial weights of the experimental data scores using the entropy weight method and the CRITIC weight method respectively, and then perform an arithmetic mean on the two weights, using the arithmetic mean as the weight factor of the experimental data scores.

[0009] Preferably, S4 includes: Based on the weighting factors of the QSAR prediction score and the experimental data score, the hazard score of each substance is calculated, and the priority of each exogenous substance is divided into three levels: high, medium and low according to the preset threshold range.

[0010] Preferably, S5 includes: S51: Normalize the chemical content data to obtain the residue score; S52: Construct a four-quadrant risk matrix diagram, with the hazard score as the Y-axis and the residual score as the X-axis; S53: Set multi-level risk classification thresholds to divide the four-quadrant matrix into regions: the first quadrant is the area of ​​extremely high concern, the second quadrant is the area of ​​controlled hazards, the third quadrant is the area of ​​process concern, and the fourth quadrant is the area of ​​low risk. S54: For exogenous chemicals whose concentration is higher after processing and cleaning than before cleaning in special food processes, a process rebound coefficient is introduced and its risk level is adjusted upward to obtain the final risk assessment result.

[0011] As can be seen from the above technical solution, compared with the prior art, the present invention provides a method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion. It integrates multi-source QSAR prediction results, experimental toxicology database evidence and objective combination weighting strategies, and combines actual residue levels to carry out four-quadrant risk characterization, so as to achieve rapid and accurate risk screening of complex exogenous chemicals in novel foods. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0013] Figure 1 A flowchart illustrating the overall process of a method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion, as provided in this embodiment of the invention. Figure 2 A schematic diagram illustrating the overall prediction results of four (Q)SAR systems for 43 research chemicals provided in this embodiment of the invention; Figure 3A schematic diagram illustrating the negative, undefined, and overlapping prediction results of four (Q)SAR systems provided in this embodiment of the invention for 43 research chemicals; Figure 4 Data richness and ordination plots of various exogenous SEXP chemicals provided for embodiments of the present invention; Figure 5 A comprehensive ranking chart of the genotoxic hazards of 43 exogenous chemicals provided for embodiments of the present invention; Figure 6 A schematic diagram illustrating the variation of Kendall's correlation coefficient τ with weight α provided in this embodiment of the invention; Figure 7 A schematic diagram illustrating the sensitivity of the Top 8 chemical rankings to the SEXP weight (α) provided in this embodiment of the invention; Figure 8 A hierarchical clustering tree diagram of 80 external verification substances provided in this embodiment of the invention; Figure 9 This invention provides a schematic diagram illustrating the comprehensive risk assessment of 14 exogenous chemicals using a four-quadrant analysis method. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Example 1

[0016] See Figure 1 As shown, Embodiment 1 of the present invention discloses a method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion, comprising the following steps: S1: Obtain the food sample to be processed, and combine the sample with the type of exogenous chemical to be analyzed to obtain the corresponding chemical residue data; S2: Obtain experimental data scores for each chemical using the genotoxicity experimental database, and calculate the QSAR prediction score for each chemical using the QSAR tool; S3: Based on the experimental data score and QSAR prediction score of each chemical, the corresponding weighting factor is determined by a combined weighting method that combines entropy weighting and CRITIC weighting. S4: Based on the preprocessed experimental data score, QSAR prediction score, and weighting factor, calculate the hazard score for each exogenous chemical, as shown in the following expression:

[0017] In the formula, This represents the QSAR prediction score. The weights represent the QSAR prediction scores. Indicates the score of the experimental data. The weights representing the scores of the experimental data; S5: Combining the hazard score and chemical residue data, construct a four-quadrant matrix diagram, and use the four-quadrant matrix diagram and comprehensive risk score to conduct a safety risk assessment.

[0018] In one specific embodiment, S1 includes: S11: Build a basic information database of exogenous chemicals using Comptox online tools or Pubchem chemical database, which may include the CAS number, SMILES and related structural information of each chemical. S12: Combine food safety standards or new food application guidelines (which can be new foods, such as cell-cultured meat, etc.) to screen the basic information database and determine the list of exogenous chemicals to be analyzed. Specifically, the data can be standardized to exclude entries that lack key information (such as CAS number or SMILES structure) or contain mixtures for which structural information cannot be obtained. Finally, the list of exogenous chemicals containing complete structural information is determined. S13: Obtain a food sample to be processed, and perform quantitative analysis of the exogenous chemicals in the sample in conjunction with the list of exogenous chemicals to be analyzed, to obtain the corresponding chemical residue data; In one specific embodiment, S2 includes: S21: Collect genotoxicity experimental data from the Endpoint module in the QSAR Toolbox and the CompTox Chemicals Dashboard platform; S22: The evidence quantity and positive rate of the genotoxicity experimental data were integrated using a logarithmic weighting method, and the adjusted scores were standardized using a min-max scaling method to obtain the experimental data score for each chemical. The specific expression is as follows:

[0019] In the formula, Indicates a positive test result. Indicates the number of positive tests; The adjusted scores can then be normalized using the min-max scaling method to obtain the final experimental data score. The specific expression is as follows:

[0020] In the formula, , These represent the minimum and maximum adjusted scores in the entire dataset, respectively; S23: Select multiple QSAR evaluation tools with different mechanisms or algorithms, input the structural information of exogenous chemicals into QSARToolbox, Toxtree, TEST, and VEGA tools for QSAR prediction, and classify the prediction results as positive, negative, or undefined according to the classification criteria. The specific criteria for classification are detailed in Table 1. S24: The QASR prediction classification results are processed using the minimum-maximum scaling method to obtain the corresponding QSAR prediction scores.

[0021] Table 1. Classification criteria for prediction results

[0022] Specifically, to ensure the scientific rigor and comparability of the evaluation results, this embodiment of the invention employs a standardized method to handle missing data: chemicals missing from all sources are excluded from priority ranking; missing data from specific sources (such as the QSAR Toolbox) are set to zero; and cases lacking clear quantitative structure-activity relationship (Q-SAR) prediction results are categorized as undefined. Based on this framework, standardized scores are calculated to provide a basis for priority ranking under the experimental data dimension. After data processing, 3,2'-dihydroxyflavone, missing from all sources, was removed, and the remaining 43 chemicals were retained for subsequent analysis.

[0023] In one specific embodiment, S2 further includes: S25: Evaluate the QSAR prediction score and the experimental data score respectively, specifically including: (1) To evaluate the robustness of the model framework, this embodiment of the invention employs a successive perturbation method, with the weighting coefficient α within [S EXP S QSAR The weights were adjusted in increments of 0.1 within the range of [0, 1], and the weights were calculated using the formula 1-α. The consistency of rankings among these variants was quantitatively analyzed using the Kendall correlation coefficient τ. (2) To evaluate the predictive performance of this prioritization framework, this embodiment of the invention randomly selected 80 substances with clear Ames test results from publicly available mutagenicity datasets as an external validation set. To avoid class imbalance bias, in accordance with QSAR validation recommendations, a 1:1 ratio of Ames positive to Ames negative substances was maintained. The predictive sensitivity was calculated based on true positive (TP) and false negative (FN) results, using the following formula:

[0024] Assuming independent Bernoulli trials with a constant probability of success, the 95% confidence interval (CI) for observation sensitivity was calculated based on the binomial distribution, and its specific expression is as follows:

[0025] In the formula, Indicates the observed sensitivity. This indicates the number of positive samples in the validation set.

[0026] A genotoxicity hazard prioritization framework was used to calculate weighted multi-source evidence scores for all validating substances. To convert consecutive priority scores into binary predictive results (positive or negative), hierarchical clustering (based on Euclidean distance and full connectivity) was employed to derive data-driven thresholds rather than pre-defined cutoff values. The final clustering results were compared with true mutagenicity labels.

[0027] The validation performance was evaluated using standard classification metrics such as sensitivity, specificity, accuracy, precision, F1 score, ROC-AUC, and Matthews correlation coefficient (MCC) (Sokolova & Lapalme, 2009). The MCC metric was introduced to ensure the model's robustness even with potential class imbalances. The specific formulas for each evaluation metric are as follows:

[0028] In the formula, TN represents a true negative example and FP represents a false positive example.

[0029] In one specific embodiment, S3 includes: S31: Calculate the initial weights of the QSAR prediction score using the entropy weight method and the CRITIC weight method respectively, and then perform an arithmetic mean on the two weights, using the arithmetic mean as the weight factor of the QSAR prediction score. S32: Calculate the initial weights of the experimental data scores using the entropy weight method and the CRITIC weight method respectively, and then perform an arithmetic mean on the two weights, using the arithmetic mean as the weight factor of the experimental data scores.

[0030] In one specific embodiment, S4 includes: Based on the weighting factors of the QSAR prediction score and the experimental data score, the hazard score of each substance is calculated, and the priority of each exogenous substance is divided into three levels: high, medium and low according to the preset threshold range.

[0031] In one specific embodiment, S5 includes: S51: Normalize the chemical content data to obtain the residue score, the specific expression of which is:

[0032] In the formula, This indicates the highest concentration measured in all tested samples. Substances with detection results below the limit of quantitation (BLLOQ) are assigned a value of 0. S52: Construct a four-quadrant risk matrix diagram, with the hazard score as the Y-axis and the residual score as the X-axis; S53: Set multi-level risk classification thresholds to divide the four-quadrant matrix into regions: Quadrant 1 is the area of ​​extremely high concern, Quadrant 2 is the area of ​​controlled hazards, Quadrant 3 is the area of ​​process concern, and Quadrant 4 is the area of ​​low risk. Quadrant 1 (area of ​​extremely high concern) includes substances with both hazard scores and residue scores greater than 0.5, which are listed as the highest priority risk factors due to their high genotoxicity risk and detected high residue levels. Quadrant 2 (area of ​​controlled hazards) includes substances with hazard scores greater than 0.5 but residue scores less than 0.5, indicating that although their inherent genotoxicity risk is high, existing processes can effectively control residue exposure. Quadrant 3 (area of ​​process concern) has hazard scores less than 0.5 but residue scores greater than 0.5; this low genotoxicity risk is accompanied by high residue levels, requiring close attention to its potential impact on product metabolic balance and overall quality. Quadrant 4 (area of ​​low risk) includes substances with both scores less than 0.5, which are assigned the lowest priority in safety assessment. S54: For exogenous chemicals whose concentration is higher after processing and cleaning than before cleaning in special food processes, a process rebound coefficient is introduced and its risk level is adjusted upward to obtain the final risk assessment result.

[0033] Example 2

[0034] The detailed process of the method provided in Specific Application Implementation Example 1 is as follows: 1. In Example 2 of this invention, the cell-cultured porcine biomass was selected from a pilot process for cell-cultured meat production. Pre-cleaned biomass samples (5 grams per tube) and porcine fat biomass samples (5 grams per tube) were obtained, and 14 exogenous chemicals were quantitatively analyzed using liquid chromatography-tandem mass spectrometry (LC-MS / MS). These included four chemicals used in the cell proliferation and differentiation stages (lipoic acid, forscolin, hydrocortisone, and dexamethasone), and ten chemicals used in the cell isolation and cell bank construction stages (sodium penicillin, streptomycin, hypoxanthine, putrescine, thymidine, sodium pyruvate, N-acetyl-L-cysteine, Trolox, Y-27632, and ethanolamine). It should be noted that the specific concentrations of these chemicals have not been publicly disclosed.

[0035] Following initial screening, 43 exogenous chemicals used in cell-cultured meat production were selected for evaluation. These chemicals can be broadly categorized according to their function as follows: antibiotics, non-enzymatic dissociation agents and processing chemicals, culture medium nutrients and metabolic regulators, hormones and growth promoters, cell regulation reagents, and trace elements or vitamin-like metabolic factors.

[0036] The list of items is shown in Table 2.

[0037] Table 2 List of Chemicals

[0038] 2. Analysis of QSAR prediction results and distribution characteristics: Figure 2This paper presents the overall prediction results of four QSAR models for 43 research chemicals. Overall, there are significant differences among the tools in mutagenicity prediction: the positive prediction rates vary significantly among the tools. Toxtree had the highest rate (18.6%, 8 / 43), followed by QSAR Toolbox (11.6%, 5 / 43), TEST (7.0%, 3 / 43), and VEGA (4.7%, 2 / 43). It should be noted that within the recoding framework, a negative result indicates "no positive reaction detected," rather than a definitive absence of mutagenicity (VanBossuyt et al., 2017). QSAR Toolbox had the highest negative prediction rate (62.8%, 27 / 43), followed by Toxtree (48.8%, 21 / 43), VEGA (39.5%, 17 / 43), and TEST (37.2%, 16 / 43). The prediction results for single compounds also showed significant differences among different tools: Trolox was predicted positive by TEST, but negative by the other three tools; Estradiol-17β, dexamethasone, testosterone, progesterone, and folic acid were predicted positive by Toxtree, while negative by QSAR Toolbox, VEGA, and TEST; conversely, ethanolamine was predicted positive by QSAR Toolbox, but negative by the other tools.

[0039] The proportion of predictions classified as "undefined" (i.e., results that are outside the scope of applicability or do not meet the model's internal criteria) varies significantly across different tools (see [link to tool name]). Figure 3 (b) In VEGA and TEST, the percentage of undefined predictions was 55.8% (24 / 43); in Toxtree it was 32.6% (14 / 43); and in QSAR Toolbox it was 25.6% (11 / 43). Figure 3 The 'c' in the figure further demonstrates the overlap of positive prediction results, indicating that only one compound (1,2-dichloroethane) was consistently identified as positive by all four tools.

[0040] 3. Analysis of the distribution of data richness and experimental evidence scores: To supplement the prediction comparisons between different models, this embodiment of the invention further evaluated the experimental evidence for each chemical using SEXP and data richness indices. Significant differences were found in the distributions of SEXP and data richness among the 43 exogenous chemicals (…). Figure 4(As shown). Nickel chloride, formaldehyde, and glutaraldehyde showed the highest levels of experimental evidence (SEXP ≥ 0.59; data richness ≥ 0.90), indicating well-documented genotoxicity. Quercetin and 1,2-dichloroethane ranked second (SEXP range: 0.51–0.84), with moderate breadth of database evidence. In contrast, commonly used culture medium components such as sodium benzoate, sodium pyruvate, and folic acid showed relatively low experimental support (≤ 0.34). Similarly, hormones and antibiotics had lower overall scores (≤ 0.18), and their data richness values ​​were all below 0.15.

[0041] 4. Priority assessment of genotoxic hazards of exogenous chemicals

[0042] Based on the above (Q)SAR prediction results and experimental data distribution, it is necessary to assess the impact of different evidence sources on the overall prioritization results. To this end, two objective methods, entropy weighting and CRITIC method, are used to estimate S. QSAR With S EXP The relative contribution. The entropy weight method will S QSAR Assigning a weight of 0.3941, S EXP The first method assigns a weight of 0.6059; while the CRITIC method assigns weights of 0.4246 and 0.5754 respectively. The arithmetic mean of the two weights yields a combined weight of 0.4093 (S). QSAR ) and 0.5907 (S EXP Then, these weights are used to calculate a comprehensive score and to establish a priority ranking for all chemicals.

[0043] By employing a comprehensive scoring system, embodiments of the present invention prioritize and classify these exogenous chemicals based on their potential impact on genotoxicity (see [link to relevant documentation]). Figure 5 (As shown). Among them, 1,2-dichloroethane was classified as having a high genotoxicity hazard level (high level); formaldehyde, nickel chloride, glutaraldehyde, quercetin, sodium pyruvate, sodium benzoate, and streptomycin were classified as having a medium genotoxicity hazard level (medium level); and the remaining 35 chemicals were classified as having a low genotoxicity hazard level (low level).

[0044] Table 3 summarizes the classification and measured concentrations of the medium- and high-risk chemicals identified in the embodiments of this invention, including their CAS numbers, functions / uses, priorities, and regulatory status or control measures. High- and medium-risk chemicals (such as 1,2-dichloroethane, formaldehyde, and glutaraldehyde) can be used as cross-linking agents in cell scaffolds (Ong, etc.). Medium-priority chemicals mainly refer to those with metabolic regulation or cell culture-related preservative functions. Nickel chloride (a nutrient supporting primary cell metabolism) was detected at 0.06 ppm in the collected cell samples (Mission Barns), equivalent to a daily dietary intake of approximately 1.7 × 10⁻⁶ ppm. -5 The concentration of sodium pyruvate (mg / kg body weight / day) is significantly lower than the EFSA reference dose of 0.02 mg / kg body weight / day (FDA, 2025b). Quercetin and other flavonoids, as functional chemicals or growth regulators, have been shown to promote the proliferation and differentiation of muscle stem cells (Fageeh et al., 2021). Sodium pyruvate (as a component of the culture medium supporting cell metabolism) was detected in chicken cell culture samples at concentrations of 1.04–1.59 mg per serving, which is below the EFSA safety threshold and comparable to levels found in natural foods. Sodium benzoate and streptomycin were also detected at concentrations suitable for cell culture applications.

[0045]

[0046] 5. Robustness assessment and performance verification

[0047] (1) Weight sensitivity analysis

[0048] The robustness of the overall ranking was assessed by performing a sensitivity analysis of the relative contribution of gender (α) from 0 to 1. Figure 6 As shown, the Kendall coefficient τ (which measures ranking consistency relative to the final overall ranking) exhibits a non-linear relationship with α. Ranking consistency is lowest when relying on a single data source (α=0, τ=0.615); it reaches its maximum at α=0.6 (τ=1.000); and when α=1.0, ranking consistency drops to 0.681, indicating that relying on a single data source reduces the robustness of the framework.

[0049] Analysis of the top 8 chemicals further illustrates the impact of the weighting. Within the α range [0.4, 0.7], i.e., when both QSAR and EXP evidence are of moderate weight, the top 8 list remains stable, supporting the final selection of α ≈ 0.5849. When the α value is low (α ≤ 0.3), QSAR-dominated chemicals tend to occupy the top positions; while when the α value is high (α ≥ 0.8), EXP-dominated chemicals gradually replace those predicted by QSAR.

[0050] Figure 7 The stability of each chemical under weight perturbations was further demonstrated. 1,2-Dichloroethane and formaldehyde consistently ranked in the top 8 across all α values, indicating the highest ranking stability. Streptomycin, sodium pyruvate, nickel chloride, glutaraldehyde, and quercetin also appeared frequently, showing relatively robust rankings. As the SEXP weight increased (α∈0–1), the ranking of the top 8 fluctuated. When α≤0.1, the top 8 list remained completely consistent. When α was between 0.2 and 0.5, chemicals such as procaine hydrochloride, streptomycin sulfate, glutaraldehyde, and nickel chloride entered the top 8, indicating that the framework became more sensitive to changes in SEXP weights. When α≥0.6, QSAR-dominated additives were gradually replaced by SEXP-dominated chemicals (such as sodium benzoate and folic acid), and the rankings tended to stabilize. It is worth noting that folic acid only made it into the top 8 under high SEXP weighting, while dimethyl sulfoxide and SB-20358 only appeared under low α values, indicating that their rankings are highly dependent on a single indicator.

[0051] (2) Performance validation using external datasets

[0052] To evaluate the reliability of the model, this embodiment of the invention uses a validation set consisting of 80 external substances with known Ames test results. Assuming an observational sensitivity of 0.80, the standard error (SE) of the sensitivity is calculated using a binomial distribution, yielding SE ≈ 0.063. Based on this standard error, the 95% confidence interval corresponding to the sensitivity is calculated to be 0.676–0.924. These results indicate that the validation set is sufficiently large to provide statistically reliable estimates of the model's predictive performance.

[0053] Hierarchical clustering results (see) Figure 9 As shown in the figure, this framework can accurately distinguish between negative and positive substances and is highly consistent with experimental labels. Performance metrics are as follows: Sensitivity = 1.000, Specificity = 0.925, Accuracy = 0.963, Precision = 0.930, F1 score = 0.964; ROC analysis shows an area under the curve (AUC) of 0.979, demonstrating its strong discriminative ability at different decision thresholds; the Matthews correlation coefficient (MCC) is 0.928.

[0054] 6. Framework Application and Comprehensive Risk Characteristic Analysis

[0055] To evaluate the practical applicability of the comprehensive priority framework, the embodiments of the present invention conduct a case analysis using actual cultivated meat samples (see Table 4). According to the combined scoring results of QSAR and experimental data, sodium pyruvate is determined to have a high genotoxicity risk, while streptomycin, N-acetyl-L-cysteine, hydrocortisone and ethanolamine are classified as the medium risk level. In addition, the actual concentrations of these chemicals in the biomass are measured before and after the washing treatment. 14 exogenous chemicals are quantitatively analyzed by liquid chromatography-tandem mass spectrometry (LC-MS / MS). The concentrations of most chemicals are below the limit of quantitation (BLLOQ). The detectable chemicals include hypoxanthine (524.724 mg / kg), 1,4-diaminobutane dihydrochloride (18.363 mg / kg), thymidine (1.624 mg / kg), sodium pyruvate (31.229 mg / kg), N-acetyl-L-cysteine (0.035 mg / kg), forskolin (0.223 mg / kg) and ethanolamine (0.339 mg / kg). It is worth noting that the concentrations of 1,4-diaminobutane dihydrochloride and sodium pyruvate increase during the processing, which are 12.160 mg / kg and 16.342 mg / kg respectively before washing, and rise to 18.363 mg / kg and 31.229 mg / kg respectively in the final product.

[0056] By integrating the hazard score and the residual amount score, a four-quadrant analysis method is adopted to evaluate the comprehensive risk of the 14 exogenous chemicals (see Figure 9 shown). It is worth noting that sodium pyruvate is classified into the first quadrant (high hazard - high residue), with a hazard score of 0.795 and a residual amount score of 0.554, indicating that this substance not only has significant genotoxic hazard, but also has limited removal effect (the concentration rebound value after elution is 31.229 mg / kg, while that of the control group is 16.342 mg / kg). Streptomycin is located in the second quadrant: although its hazard score is relatively high (0.545), its residual amount cannot be quantitatively detected (<BLLOQ), indicating that the process control is effective. On the contrary, hypoxanthine presents the characteristics of low hazard but high residue (the third quadrant), with a hazard score of 0.153, and reaches the maximum residue limit (524.724 mg / kg, residual amount score of 1). The remaining substances (such as ethanolamine, N-acetyl-L-cysteine, thymidine and forskolin) are mainly concentrated in the low-risk fourth quadrant. It is particularly worth noting that although putrescine has a slight process enrichment phenomenon (the concentration after elution increases from 12.160 mg / kg to 18.363 mg / kg), it still remains safely in this low-priority region due to its low hazard score (0.256).

[0057]

[0058] In summary, the method provided in this embodiment of the invention uses an entropy-critique combined weighting method to integrate mutagenicity prediction results based on quantitative structure-activity relationship (QSAR) with genotoxicity experimental database data, thereby realizing the classification of the hazard level of exogenous chemicals used in the production of cell-cultured meat and the graded assessment of their genotoxicity intensity.

[0059] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0060] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion, characterized in that, Includes the following steps: S1: Obtain the food sample to be processed, and combine the sample with the type of exogenous chemical to be analyzed to obtain the corresponding chemical residue data; S2: Obtain experimental data scores for each chemical using the genotoxicity experimental database, and calculate the QSAR prediction score for each chemical using the QSAR tool; S3: Based on the experimental data score and QSAR prediction score of each chemical, the corresponding weighting factor is determined by a combined weighting method that combines entropy weighting and CRITIC weighting. S4: Calculate the hazard score for each exogenous chemical based on the preprocessed experimental data score, QSAR prediction score, and weighting factor; S5: Combining the hazard score and chemical residue data, construct a four-quadrant matrix diagram, and use the four-quadrant matrix diagram and comprehensive risk score to conduct a safety risk assessment.

2. The method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion according to claim 1, characterized in that, S1 includes: S11: Build a basic information database of exogenous chemicals using Comptox online tools or the Pubchem chemical database; S12: Screen the basic information database in conjunction with food safety standards or new food application guidelines to determine the list of exogenous chemicals to be analyzed; S13: Obtain the food sample to be processed, and perform quantitative analysis of the exogenous chemicals in the sample in conjunction with the list of exogenous chemicals to be analyzed, to obtain the corresponding chemical residue data.

3. The method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion according to claim 2, characterized in that, S2 includes: S21: Collect genotoxicity experimental data from the Endpoint module in the QSAR Toolbox and the CompTox Chemicals Dashboard platform; S22: The evidence amount and positive rate of the genotoxicity experimental data are integrated using a logarithmic weighting method, and the adjusted scores are standardized using a min-max scaling method to obtain the experimental data score for each chemical. S23: Select multiple QSAR evaluation tools with different mechanisms or algorithms, input the exogenous chemical structure information into QSARToolbox, Toxtree, TEST, and VEGA tools for QSAR prediction, and classify the prediction results as positive, negative, or undefined according to the classification criteria; S24: The QASR prediction classification results are processed using the minimum-maximum scaling method to obtain the corresponding QSAR prediction scores.

4. The method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion according to claim 3, characterized in that, S3 includes: S31: Calculate the initial weights of the QSAR prediction score using the entropy weight method and the CRITIC weight method respectively, and then perform an arithmetic mean on the two weights, using the arithmetic mean as the weight factor of the QSAR prediction score. S32: Calculate the initial weights of the experimental data scores using the entropy weight method and the CRITIC weight method respectively, and then perform an arithmetic mean on the two weights, using the arithmetic mean as the weight factor of the experimental data scores.

5. The method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion according to claim 4, characterized in that, S4 includes: Based on the weighting factors of the QSAR prediction score and the experimental data score, the hazard score of each substance is calculated, and the priority of each exogenous substance is divided into three levels: high, medium and low according to the preset threshold range.

6. The method for screening the genotoxicity risk of exogenous chemicals in food based on multi-source evidence fusion according to claim 1, characterized in that, S5 includes: S51: Normalize the chemical content data to obtain the residue score; S52: Construct a four-quadrant risk matrix diagram, with the hazard score as the Y-axis and the residual score as the X-axis; S53: Set multi-level risk classification thresholds to divide the four-quadrant matrix into regions: the first quadrant is the area of ​​extremely high concern, the second quadrant is the area of ​​controlled hazards, the third quadrant is the area of ​​process concern, and the fourth quadrant is the area of ​​low risk. S54: For exogenous chemicals whose concentration is higher after processing and cleaning than before cleaning in special food processes, a process rebound coefficient is introduced and its risk level is adjusted upward to obtain the final risk assessment result.

Images