Transgenic maize gene drift prediction method based on wind speed dominance

Abstract

The present application belongs to the technical field of agricultural environmental science, and particularly relates to a transgenic corn gene drift prediction method based on wind speed dominance, which collects environmental parameters of wind speed, wind direction, humidity, atmospheric stability, positioning, range and altitude, corn variety specificity parameters and historical gene drift observation data in a target area; constructs a Gaussian plume diffusion model based on the preprocessed data to determine a gene drift frequency atlas; under a Bayesian neural field framework, the gene drift frequency atlas is taken as input, the actually measured gene drift data is taken as model observation values, and the preprocessed environmental parameters and corn variety specificity parameters are taken as covariates to optimize the model, so as to output a predicted gene drift flow threshold distance and give a risk prevention and control report. The present application combines the advantages of physical models and empirical models, overcomes the defects that the prior art does not fully consider the differences in varieties and environmental factors, and realizes accurate prediction of gene drift.

Description

Technical Field

[0001] This invention belongs to the field of agricultural environmental science and technology, specifically relating to a wind speed-driven method for predicting gene drift in transgenic maize. Background Technology

[0002] Gene drift is a key source of ecological risk when genetically modified (GM) and non-GM maize coexist. As a cross-pollinated crop, maize has a high natural cross-pollination rate and long pollen survival time, further increasing the risk of long-distance gene drift. During pollen dispersal, wind speed is the core factor determining the range and intensity of dispersal—high wind speeds (>6 m / s) significantly prolong the horizontal dispersal distance of pollen, while low wind speeds (<2 m / s) cause pollen to accumulate around the donor, greatly increasing the risk of local contamination. Humidity regulates dispersal efficiency by affecting pollen viability, and altitude obstruction exacerbates the risk of local drift by altering airflow trajectories. Wind direction differences lead to significantly different drift rates in different directions (e.g., the drift rate reaches 1.00% at 60m in the prevailing wind direction, while it is only 0.20% in the leeward direction). Currently, there are significant shortcomings in methods for predicting maize gene drift: First, variety-specific pollen parameters are not fully integrated (e.g., the amount of pollen per plant varies by 1.473 times among different varieties), and the use of uniform pollen source intensity leads to prediction bias. Second, calibration of the gene drift threshold distance (MTD) is lacking, and the measured MTD varies significantly in different regions (e.g., MTD 1% fluctuates between 45-60m), and existing models have not established a differentiated calibration mechanism. Third, the regulatory effect of wind direction on drift direction has not been quantified, resulting in insufficient targeting of control measures. Fourth, traditional Gaussian feather models do not incorporate maize flowering period characteristics, and single physical models are insufficient to quantify the prediction uncertainty under scenarios of wind speed fluctuations and direction changes, failing to meet the precise risk control needs after the commercial planting of genetically modified maize. Summary of the Invention

[0003] The purpose of this invention is to provide a wind speed-driven method for predicting gene drift in transgenic maize, in order to solve the problems of inaccurate prediction results caused by existing transgenic maize gene drift prediction methods, which do not fully integrate variety-specific parameters, lack gene flow threshold distance differential calibration mechanisms, and fail to quantify the role of wind direction regulation.

[0004] The present invention achieves the above objectives through the following technical solutions: This invention proposes a wind speed-driven method for predicting gene drift in transgenic maize, the method comprising: S1. Collect environmental parameters, maize variety-specific parameters, and historical gene drift observation data within the target area, and preprocess the acquired data; wherein, the environmental parameters include wind speed, wind direction, humidity, atmospheric stability, location, range, and altitude within the target area; S2. Based on the preprocessed data, a Gaussian plume diffusion model was constructed. The gene drift frequency spectrum was determined with wind speed as the dominant factor. Under the Bayesian neural field framework, the gene drift frequency spectrum was used as input, the measured gene drift data was used as the model observations, and the preprocessed environmental parameters and maize variety-specific parameters were used as covariates. The final weight coefficients in the Bayesian neural field model were determined through optimization training. S3. Input the real-time environmental parameters and variety parameters into the optimized Bayesian neural field model, and output the predicted gene drift threshold distance to guide the setting of the isolation distance between genetically modified and non-genetically modified maize.

[0005] Furthermore, in step S1, the maize variety-specific parameters include pollen quantity, flowering period, pollen viability time, and leaf area index. The environmental parameters include wind speed (hourly wind speed, wind speed standard deviation, and maximum wind speed), humidity (hourly relative humidity, daily average relative humidity, and humidity gradient), and daily temperature difference. The historical gene drift observation data includes measured values ​​of gene drift frequency and measured values ​​of gene drift flow threshold distance under different wind speed and direction conditions.

[0006] Furthermore, the preprocessing in step S1 includes: Missing wind speed values ​​are filled in and corrected according to altitude; wind directions in different directions are converted into wind direction coefficients with different weights, including the weight of the prevailing wind direction, the weight of the secondary wind direction, and the weight of the leeward direction. Standardize the raw humidity data to a predetermined range; For unknown pollen counts of a variety, the mean of the same variety was used as a substitute, and parameter sensitivity was marked. A correspondence was established between the measured gene drift frequency and the measured gene drift threshold distance in the historical gene drift observation data, which served as the benchmark dataset for model calibration.

[0007] Furthermore, step S2 specifically includes: S201. Construct a Gaussian plume diffusion model, using preprocessed wind speed, wind direction, humidity, atmospheric stability, location, range, and altitude as inputs, and combine them with the pollen source intensity of the variety to determine the gene drift frequency spectrum at different spatiotemporal nodes within the target area. S202. Construct a Bayesian neural field model, using spatiotemporal coordinates and gene drift frequency maps as inputs, the gene drift data as model observations, and preprocessed environmental parameters and maize variety-specific parameters as covariates; wherein, wind speed and variety pollen quantity covariates are assigned an initial weight higher than other environmental parameter covariates; S203. Using a stochastic variational inference method, the lower bound of evidence is maximized to optimize the model parameters of the Bayesian neural field fusion framework, and the final weight coefficients in the Bayesian neural field model are further determined. S204. Generate multiple sets of parameters through Monte Carlo sampling, calculate the predicted values ​​of gene drift frequency and gene drift flow threshold distance, and determine their confidence intervals.

[0008] Furthermore, in step S201, the Gaussian feather diffusion model is used to describe the spatial diffusion process of maize pollen, including the following formula: (1) (2) (3) (4) In the formula, G represents the gene drift frequency, Q represents the pollen source release intensity, and C represents the pollen concentration in the air. These represent the number of pollen grains deposited on the filaments of the donor and recipient flowers, respectively. and These are the genetic competitiveness parameters for genetically modified and non-genetically modified maize, respectively. These are the horizontal and vertical diffusion parameters determined by wind speed, respectively. For the corn ears to be tall, For the sedimentation rate of corn pollen, denoted by , x represents the wind speed at the top of the corn cob; x represents the distance in the downwind direction; y represents the distance perpendicular to the downwind direction; and z represents the vertical height. The cumulative leaf area index, This indicates the percentage of pollen grains that have penetrated the canopy.

[0009] Furthermore, in step S202, the Bayesian neural field model adopts a 3-layer fully connected network structure with 128 nodes in each layer, uses the ELU activation function, and introduces a Dropout layer with a dropout rate of 0.2 in each layer.

[0010] Furthermore, step S3 specifically includes: S301. In the spatial dimension, output the spatial distribution of pollen concentration within the target region and label the spatial boundary of the gene flow threshold distance; S302. In the time dimension, output the hourly pollen diffusion dynamics during the peak pollen shedding period in the flowering period, and trigger an early warning when extreme weather conditions are identified; S303. In the MTD project output results, the gene flow threshold distance and its uncertainty range are output according to different ecological zones, variety types and weather scenarios; S304. Based on the aforementioned spatial dimension, time dimension, and MTD-specific output results, generate a risk prevention and control report that includes recommendations for setting isolation distances between genetically modified and non-genetically modified corn.

[0011] Furthermore, in step S3, when the uncertainty range of the gene flow threshold distance exceeds the preset first range, the area is marked as an area requiring key verification in the risk prevention and control report.

[0012] Furthermore, the environmental parameters are acquired through a data acquisition device, which includes sensors for measuring wind speed, wind direction, and temperature / humidity, as well as a LoRa, 4G / 5G, or Bluetooth transmission module for data transmission.

[0013] The beneficial effects of this invention are as follows: 1. The prediction method proposed in this invention takes wind speed as the core influencing factor, and combines it with maize variety-specific parameters, gene flow threshold distance and other environmental parameters to achieve accurate prediction and uncertainty quantification of maize pollen diffusion in different ecological zones and different varieties, providing a basis for setting isolation distances and risk management for GM and non-GM maize.

[0014] 2. This invention combines a wind-driven Gaussian plume diffusion model to calculate a gene drift frequency map; further, through stochastic variational inference and Monte Carlo sampling under a Bayesian framework, it quantifies the uncertainty of gene flow threshold distance, automatically identifies high-risk areas and generates confidence intervals, and provides differentiated dynamic isolation schemes for different ecological regions (such as the Northeast Plain and the Huang-Huai-Hai region). Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a wind speed-driven transgenic maize gene drift prediction method proposed in this invention. Figure 2 This is a schematic diagram of a process for constructing the optimized fusion prediction model in this invention; Figure 3 This is a Bayesian neural field classification diagram from the present invention; Figure 4 The image shows the simulation results of Gaussian feather pollen concentration in this invention, specifically under constant wind speeds of 1.0 m / s, 3.0 m / s, 5.0 m / s, and 10.0 m / s. Figure 5 The image shows the simulation results of Gaussian feather pollen concentration in this invention, with specific scenarios including prevailing wind speeds of 1.0 m / s, 3.0 m / s, 5.0 m / s, and 10.0 m / s.

[0016] Figure 6The image shows the simulation results of pollen gene drift frequency in this invention, with specific scenarios including constant wind speeds of 1.0 m / s, 3.0 m / s, 5.0 m / s, and 10.0 m / s. Figure 7 The image shows the simulation results of pollen gene drift frequency in this invention, with specific scenarios including prevailing wind speeds of 1.0 m / s, 3.0 m / s, 5.0 m / s, and 10.0 m / s. Figure 8 This is a framework diagram of the wind speed-driven transgenic maize gene drift prediction method proposed in this invention. Detailed Implementation

[0017] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0018] Example 1 Please see Figure 1 , Figure 2 , Figure 3 and Figure 8 A specific embodiment of this invention proposes a wind-speed-dominant method for predicting transgenic maize gene drift. A transgenic maize gene drift MTD model is established, which uses wind speed as the dominant factor and integrates multiple environmental parameters such as humidity, altitude, and wind direction, coupled with a Gaussian feather and Bayesian neural field. This model first acquires environmental parameters in real time through data acquisition devices deployed in the target area, including hourly wind speed, wind direction, humidity, atmospheric stability, location, range, and altitude at the maize ear height. Simultaneously, it collects maize variety-specific parameters and historical gene drift observation data.

[0019] In the data preprocessing stage, missing values ​​were imputed and altitude correction was performed on wind speed data, wind direction was converted into a weighting coefficient, humidity data was standardized, and the mean of similar varieties was used for calibration for varieties with unknown pollen counts. Subsequently, a Gaussian plume diffusion model was constructed, using preprocessed environmental and variety parameters as inputs, to calculate the spatial distribution of gene drift frequencies.

[0020] Building upon this foundation, a Bayesian neural field framework was introduced. Using spatiotemporal coordinates and the aforementioned gene drift frequency map as inputs, gene drift data measured using KASP technology were used as model observations. Preprocessed environmental parameters and maize variety-specific parameters were used as covariates. The model parameters were optimized using stochastic variational inference, and Monte Carlo sampling was employed to quantify prediction uncertainties. The resulting fusion model accurately predicts pollen concentration distribution, gene flow threshold distance, and their confidence intervals, providing a basis for setting isolation distances in scenarios where transgenic (GM) and non-GM maize coexist.

[0021] Specifically, the method includes the following steps: S1. Obtain environmental parameters, maize variety-specific parameters (pollen quantity, flowering time, pollen viability time, leaf area index) and historical gene drift observation data (including measured MTD values) within the target area, and preprocess the acquired data, including data cleaning, standardization and missing value imputation; wherein, the environmental parameters include wind speed, wind direction, humidity, atmospheric stability, location, range and altitude within the target area; Preferably, environmental parameters are acquired through a data acquisition device, which includes sensors for measuring wind speed, wind direction, and temperature / humidity, as well as a LoRa, 4G / 5G, or Bluetooth transmission module for data transmission.

[0022] Specifically, maize variety-specific parameters include pollen quantity, flowering time, pollen viability time, and leaf area index; environmental parameters include wind speed data such as hourly wind speed, wind direction, wind speed standard deviation, and maximum wind speed at the height of the maize ear; humidity data include hourly relative humidity, daily average relative humidity, and humidity gradient; environmental parameters also include daily temperature difference; historical gene drift observation data include measured values ​​of gene drift frequency and measured values ​​of gene drift migration threshold distance under different wind speed and direction conditions.

[0023] The preprocessing in step S1 includes: filling in missing wind speed values ​​and correcting them according to altitude; converting wind directions of different directions into wind direction coefficients with different weights, including the weight of the prevailing wind direction, the weight of the secondary wind direction, and the weight of the leeward wind direction; standardizing the original humidity data to a predetermined range; replacing the pollen count of unknown varieties with the mean of the same type of variety and marking the parameters for sensitivity; and establishing a correspondence between the measured gene drift frequency values ​​and the measured gene drift flow threshold distance values ​​in the historical gene drift observation data as a benchmark dataset for model calibration.

[0024] The data and preprocessing procedures obtained above, for example, include the following: (1) Wind speed and direction data: wind speed (unit: m / s), wind speed standard deviation, maximum wind speed (e.g., extreme wind speed of 8 m / s in Gongzhuling area), and hourly wind direction (recorded in 16 directions, such as 0° due north and 90° due east), with a time resolution of not less than 1 hour, and the data duration must cover at least 3 corn growth cycles, and include complete data on flowering period (e.g., July 21-27 in Jilin); (2) Humidity and temperature data: daily average relative humidity (unit: %), humidity gradient, and daily average temperature (unit: ℃), with a focus on collecting matching data of 25-30℃ (suitable temperature for corn flowering) and 60-80% humidity (common range for flowering in Jilin region); (3) Altitude data: used to obtain altitude and terrain slope information in the target area (slope > 5° is weak occlusion, > 15° is strong occlusion). (4) Maize variety-specific parameters: donor / recipient variety name (e.g., Zinuo 18, Jidan 35, GA21), pollen yield per plant (Zinuo 18: 3.2515g, Jidan 35: 2.2073g, GA21: 2.8g, referencing similar varieties), flowering time (accurate to the day, e.g., flowering begins on July 21 and ends on July 27 in Jilin), pollen viability time (accurate to the hour, e.g., usually 2-8 hours under normal temperature conditions, and viability is maintained for about 4-6 hours at 25℃); (5) Historical gene drift observation data: measured gene drift frequency under different wind speed and direction conditions (e.g., 45.10% at 1m, 1.00% at 60m, 0.05% at 150m, 0% at 200m), measured MTD values ​​(MTD1%=45m, MTD0.1%=300m), and the data should cover typical areas such as the Northeast spring maize region (e.g., Gongzhuling, Jilin) ​​and the Huang-Huai-Hai summer maize region.

[0025] Regarding the preprocessing process: (1) Wind speed and direction data processing: The inverse distance weighted interpolation method of wind speed data from neighboring meteorological stations was used to fill the missing values, combined with altitude difference correction (the wind speed correction coefficient is 1.05 for every 100m increase in altitude); the 16 directional wind directions were converted into wind direction weight coefficients, with the prevailing wind direction (such as northwest wind in Gongzhuling) weighted at 1.2, the secondary wind direction at 1.0, and the leeward direction at 0.8, for subsequent model direction correction; (2) Standardization of humidity and temperature data: Humidity is calculated using the following formula: Standardized to the [0,1] range, for every 0.1 increase in the standardized humidity value, pollen survival time is extended by 15%; a temperature correction factor is introduced to reflect the inhibitory effect of temperature on pollen viability; (3) Calibration of pollen count data for different varieties: If the pollen count of the donor variety is unknown, the average value of the same type of variety is used as a substitute (e.g., for waxy corn, refer to Zinuo 18 and take 3.25g, for common corn, refer to Jidan 35 and take 2.21g), and it is marked as a medium-sensitive parameter. In the uncertainty analysis, the parameter fluctuation range is increased by 5%. (4) Matching historical observation data: The measured values ​​of gene drift frequency and MTD are used as the benchmark data for model calibration to ensure that the preprocessed data is consistent with the actual measurement scenario.

[0026] S2. Based on the preprocessed data, a Gaussian plume diffusion model was constructed. The gene drift frequency spectrum was calculated with wind speed as the dominant factor. Under the Bayesian neural field framework, the gene drift frequency spectrum was used as input, the gene drift data measured by KASP technology was used as the model observations, and the preprocessed environmental parameters and maize variety-specific parameters were used as covariates. The model parameters corresponding to each covariate were determined by the optimization algorithm.

[0027] Preferably, step S2 integrates the Gaussian feather physics model with the Bayesian neural field to establish a gene drift prediction model with wind speed as the dominant factor. This model fully considers the multi-factor coupling effect of environmental parameters, and particularly highlights the dominant role of wind speed in the pollen dispersal process, specifically including: S201. Based on Gaussian diffusion theory, a Gaussian plume diffusion model was constructed. Using preprocessed wind speed, humidity, altitude and wind direction data as input, and combined with the pollen source intensity of the variety, the gene drift frequency at different spatiotemporal nodes in the target area was calculated. The Gaussian feather diffusion model is used to describe the spatial diffusion process of maize pollen, and includes the following formulas: (1) (2) (3) (4) In the formula, G represents the gene drift frequency, Q represents the pollen source release intensity, and C represents the pollen concentration in the air. These represent the number of pollen grains deposited on the filaments of the donor and recipient flowers, respectively. These are the genetic competitiveness parameters for genetically modified and non-genetically modified maize, respectively. These are the horizontal and vertical diffusion parameters determined by wind speed, respectively. For the corn ears to be tall, For corn pollen settling velocity, denoted by , x represents the wind speed at the top of the corn cob; x represents the distance in the downwind direction; y represents the distance perpendicular to the downwind direction; and z represents the vertical height. The cumulative leaf area index, This indicates the percentage of pollen grains that have penetrated the canopy.

[0028] S202. Construct a Bayesian neural field model. This model uses spatiotemporal coordinates and the aforementioned gene drift frequency map as inputs, gene drift data measured by KASP technology as model observations, and preprocessed environmental parameters and maize variety-specific parameters as covariates. Specifically, wind speed and variety pollen quantity are assigned higher initial weights than other environmental parameter covariates.

[0029] S203. Using a stochastic variational inference method, the lower bound of evidence is maximized to optimize the model parameters of the Bayesian neural field fusion framework, and the final weight coefficients in the Bayesian neural field model are further determined. S204. Generate multiple sets of parameters through Monte Carlo sampling, calculate the predicted values ​​of gene drift frequency and gene drift flow threshold distance, and determine their confidence intervals.

[0030] Preferably, in step S202, the Bayesian neural field model adopts a 3-layer fully connected network structure with 128 nodes in each layer, uses the ELU activation function, and introduces a Dropout layer with a dropout rate of 0.2 in each layer.

[0031] In this embodiment, the construction of the Bayesian neural field fusion framework can be combined with... Figure 3 For example, it includes the following: (1) Input layer design: includes spatiotemporal coordinates of gene drift frequency spectrum calculated by Gaussian feather model (s is spatial coordinate, accurate to 10m×10m grid; t is time coordinate, accurate to hour level, covering the peak pollen shedding period of 9:00-16:00 every day during the flowering period of maize), core covariates (wind speed v, unit m / s; humidity correction coefficient RHn, value 0-1; altitude shading coefficient H, H=1.0 when slope ≤5°, H=1.2 when 5°<slope≤15°, H=1.5 when slope>15°; wind direction weight W, 1.2 for the prevailing wind direction, 1.0 for the secondary wind direction, and 0.8 for the leeward direction; pollen amount of variety P, unit g / plant). The spatiotemporal coordinates, core covariates, humidity correction coefficient, altitude shading coefficient, wind direction weight, and pollen quantity of each variety are used as inputs. Among them, wind speed and pollen quantity of each variety are assigned weights separately as key covariates (the weight coefficients are 1.5 times and 1.2 times that of other covariates, respectively), to ensure that the core factors play a dominant role in the prediction results.

[0032] (2) Hidden layer design: Three fully connected networks can be set as hidden layers, with 128 nodes in each layer. The ELU activation function is used to introduce a Dropout layer (dropout rate 0.2) to prevent the model from overfitting. Each layer input contains separate feature mappings for wind speed and pollen content of the species to avoid the loss of key covariate information in feature extraction. (3) Observation layer design: Using gene drift data measured by KASP technology as observation values, a Gaussian observation model is constructed. ,in Output of the hidden layer of the Bayesian neural field (corrected gene flow frequency prediction). The observed variance (dynamically adjusted for the degree of wind speed fluctuation and the sensitivity of pollen quantity of the variety: for every 0.5 m / s increase in the standard deviation of wind speed or every 10% deviation of the pollen quantity of the variety from the mean) (Increase of 8%).

[0033] For example, the model parameter optimization in step S203 includes the following: The stochastic variational inference method (ADVI algorithm) is used to approximate the posterior distribution of the model parameters, and parameter optimization is achieved by maximizing the lower bound of evidence (ELBO). The specific process is as follows: (1) Variational distribution setting: The mean field normal distribution is adopted as the variational distribution. ,in These are the model parameters (including hidden layer weights, biases, and covariate weights). These are variational parameters (mean and variance); (2) ELBO calculation: ELBO formula ,in: The likelihood function is calculated based on the Gaussian model of the observation layer, reflecting the degree of fit between the model parameters and the observed data. The KL divergence measures the difference between the variational distribution and the prior distribution π(Θ) (prior distribution settings: weight prior is N(0,0.01), and key covariate weight prior is N(0,0.02); (3) Optimization execution: The Adam optimizer is used to minimize the negative value of ELBO (i.e. maximize ELBO). The learning rate is set to 0.001 and the number of iterations is 5000. Training is stopped when the change of ELBO is less than 1e-6 for 200 consecutive iterations to ensure model convergence. For example, the uncertainty quantification in step S204 includes the following: Uncertainty quantification is achieved through Monte Carlo sampling. The specific steps are as follows: (1) Parameter sampling: 1000 sets of parameters are randomly generated from the optimized variational distribution; (2) Prediction calculation: Each set of parameters is substituted into the Bayesian neural field model to calculate the corresponding gene flow frequency prediction value and MTD prediction value; (3) Calculation of uncertainty index: Confidence interval: The 2.5% quantile and 97.5% quantile of 1000 predicted values ​​are used as the 95% confidence interval. The greater the wind speed fluctuation (wind speed standard deviation > 2 m / s), the wider the confidence interval should be increased by 15%-20%. Risk probability standard deviation: Calculate the standard deviation of gene drift probability and mark the uncertainty area, where monitoring points need to be increased.

[0034] It should be noted that this invention establishes the dominant role of wind speed in gene drift prediction. Firstly, in the Gaussian plume model, wind speed is considered the primary driving force for pollen dispersal, directly influencing the dispersal range and intensity. Secondly, in the Bayesian neural field framework, wind speed covariates are assigned a higher weight coefficient than other environmental parameters (e.g., 1.5 times that of other variables), ensuring its dominant role in the prediction results. This invention's multi-factor coupled modeling with wind speed at its core not only conforms to the physical nature of pollen dispersal but also effectively quantifies prediction uncertainty.

[0035] S3. Input real-time environmental parameters (wind speed, wind direction, humidity, temperature, altitude and slope) and variety parameters (pollen quantity, flowering time, pollen viability time, leaf area index) into the optimized model to determine pollen concentration distribution, gene flow threshold distance and risk control report, so as to guide the setting of isolation distance between genetically modified and non-genetically modified maize.

[0036] Preferably, step S3 specifically includes: S301. In the spatial dimension, output the spatial distribution of pollen concentration within the target region and label the spatial boundary of the gene flow threshold distance.

[0037] Regarding the spatial distribution of MTD: Marking MTD1% and MTD 0-1 The spatial boundary is defined by % to specify the minimum isolation distance that different areas must meet (e.g., MTD1%=50m in the prevailing wind direction and MTD1%=40m in the leeward direction). High-risk region markers: Regions with a gene flow frequency >1% (corresponding to MTD within 1%) and high uncertainty are highlighted with red shading. S302. In the time dimension, output the hourly pollen diffusion dynamics during the peak pollen shedding period within the flowering period, and trigger an early warning when extreme weather conditions are identified.

[0038] Daily pollen dispersal prediction during flowering period: Output the peak pollen concentration, occurrence time, and corresponding gene flow frequency from 9:00 to 16:00 daily (peak pollen dispersal). Extreme weather warning: When the predicted wind speed is >8m / s (extreme high wind speed) or the humidity is <30% (low humidity), a real-time warning will be triggered, indicating that monitoring needs to be strengthened during this period; Flowering synchronicity analysis: output the percentage of overlapping flowering periods between donors and recipients, and the average gene drift probability within the overlapping period. S303. In the MTD project output results, the gene flow threshold distance and its uncertainty range are output according to different ecological zones, variety types and weather scenarios.

[0039] For example, output MTD1% and MTD by "Ecological Zone + Variety + Weather Type" 0-1 For example, in the spring corn region of Enyang, Sichuan (altitude 500-800m), when planting Zinuo 18 (pollen content per plant 3.25g), under normal weather conditions (wind speed 2-4m / s, humidity 60-80%), MTD1% = 45m, MTD 0-1 %=280m, under extreme high wind speed weather (wind speed > 8m / s), MTD1%=65m, MTD 0-1 %=350m.

[0040] MTD Uncertainty Analysis: Output the 95% confidence interval of MTD under different scenarios. For example, the 1% confidence interval of MTD in the Northeast spring corn area is [42m, 48m], and mark the width of the confidence interval and the influencing factors (such as the confidence interval width increases by 5%-8% for every 0.5m / s increase in wind speed standard deviation). Scenarios with a confidence interval width > 10m are marked as "scenarios that need to be verified".

[0041] S304. Based on spatial dimensions, temporal dimensions, and MTD-specific output results, generate a risk prevention and control report that includes recommendations for setting isolation distances between genetically modified and non-genetically modified corn.

[0042] For example, the MTD value can be correlated with the recommended actual planting isolation distance. For instance, when MTD1% = 50m, the recommended isolation distance is 1.2 times the MTD value (i.e., 60m) to allow for a safety margin; when MTD... 0.1 When %=300m, the recommended isolation distance for seed production fields is 300m, while for commercial corn planting fields, it can be adjusted to 250-300m according to purity requirements.

[0043] like Figure 4 and Figure 5 The simulation results of Gaussian feather pollen concentration shown in the figure intuitively demonstrate the simulation effect of the physical model of this invention on pollen diffusion. The following is an explanation and conclusion analysis of the scheme in conjunction with the figure: (1) Figure 4 (Constant wind) and Figure 5 The prevailing winds demonstrate that the Gaussian plume diffusion model used in this invention can adapt to different types of wind field conditions. Under constant wind conditions, pollen forms a regular plume distribution along a single direction; under prevailing wind conditions, pollen forms a wider diffusion range due to wind direction fluctuations. These two scenarios cover typical wind field patterns in real-world environments, reflecting the model's applicability under complex meteorological conditions.

[0044] (2) Both figures demonstrate the significant impact of wind speed on pollen dispersal. Under the same wind field type, higher wind speeds (e.g., 10 m / s) lead to a downward shift in the peak pollen concentration and an increase in the dispersal distance, while lower wind speeds (e.g., 1 m / s) cause the concentration to be more concentrated near the source region. This directly verifies the importance of wind speed data as the dominant parameter in this invention and provides a physical basis for the subsequent calculation of gene drift frequency.

[0045] (3) Wind speed is the dominant factor affecting pollen dispersal distance: regardless of constant or prevailing wind conditions, higher wind speeds lead to pollen spreading over greater distances. For example, at a constant wind speed of 10 m / s, significant concentrations may extend beyond 200 meters, while at 1 m / s, the main area of ​​influence may be limited to within 100 meters. This provides a scientific basis for determining differentiated isolation distances.

[0046] (4) Wind field type significantly alters the spatial pattern of diffusion: Figure 4 (Constant wind) shows that pollen forms a symmetrical, narrow distribution along the downwind side, while Figure 5 The prevailing winds exhibit an irregular fan-shaped diffusion pattern, with a significant increase in lateral diffusion range. This indicates that in regions with variable wind direction, considering only the average wind direction will underestimate the risk of gene flow, and the wind direction variability parameter must be introduced.

[0047] like Figure 6 and Figure 7 The simulated pollen gene drift frequency diagram shown directly illustrates the core output of this invention. The following is an explanation and conclusion analysis of the scheme in conjunction with the diagram: (1) Figure 6 (Constant wind) and Figure 7 The gene drift frequency heatmap shown in (prevailing wind) is a direct result of the cascaded operation of the Gaussian feather diffusion model and the gene drift frequency calculation model of this invention. They represent the physical diffusion process of pollen ( Figure 4 , Figure 5 Ultimately, this is transformed into biological indicators that can be directly used for risk assessment.

[0048] (2) The two figures clearly demonstrate the quantitative impact of wind speed and wind field type, the two dominant environmental parameters, on the risk of gene drift. This proves the rationality and necessity of using these key factors as core inputs in the design phase of this invention, and the model's output results show a high degree of sensitivity to them.

[0049] (3) Wind speed is a key factor determining the threshold distance (MTD): Figure 6 and Figure 7It can be clearly observed that, under the same wind field type, the higher the wind speed, the slower the decay of gene drift frequency with distance, resulting in a longer maximum threshold distance (MTD) required to reach the same threshold (e.g., 1% or 0.1%). For example, under constant wind conditions, a wind speed of 5 m / s corresponds to... This could be several times greater than at a wind speed of 1 m / s. This provides the most direct quantitative basis for determining differentiated isolation distances.

[0050] (4) Wind field type significantly affects the pattern of risk distribution: Figure 6 (Constant wind) suggests that in areas with stable wind direction, the effectiveness of isolation measures is more predictable and guaranteed; Figure 7 (Prevailing wind) means that, at the same downwind distance, the risk of gene drift is higher under prevailing wind conditions than under constant wind conditions. Therefore, in areas dominated by prevailing winds (such as coastal areas and plateaus), it is necessary to establish longer isolation distances or wider buffer zones.

[0051] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated.

[0052] The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0053] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0054] In addition, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0055] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A wind-speed-dominated method for predicting gene drift in transgenic maize, characterized in that, The method includes: S1. Collect environmental parameters, maize variety-specific parameters, and historical gene drift observation data within the target area, and preprocess the acquired data; wherein, the environmental parameters include wind speed, wind direction, humidity, atmospheric stability, location, range, and altitude within the target area; S2. Based on the preprocessed data, a Gaussian plume diffusion model was constructed. The gene drift frequency spectrum was determined with wind speed as the dominant factor. Under the Bayesian neural field framework, the gene drift frequency spectrum was used as input, the measured gene drift data was used as the model observations, and the preprocessed environmental parameters and maize variety-specific parameters were used as covariates. The final weight coefficients in the Bayesian neural field model were determined through optimization training. S3. Input the real-time environmental parameters and variety parameters into the optimized Bayesian neural field model, and output the predicted gene drift threshold distance to guide the setting of the isolation distance between genetically modified and non-genetically modified maize.

2. The wind speed-dominated transgenic maize gene drift prediction method according to claim 1, characterized in that, In step S1, the maize variety-specific parameters include pollen quantity, flowering time, pollen viability time, and leaf area index. The environmental parameters include wind speed (hourly wind speed, wind speed standard deviation, and maximum wind speed), humidity (hourly relative humidity, daily average relative humidity, and humidity gradient), and daily temperature difference. The historical gene drift observation data includes measured values ​​of gene drift frequency and measured values ​​of gene drift flow threshold distance under different wind speed and direction conditions.

3. The wind speed-driven transgenic maize gene drift prediction method according to claim 2, characterized in that, The preprocessing in step S1 includes: Missing wind speed values ​​are filled in and corrected according to altitude; wind directions in different directions are converted into wind direction coefficients with different weights, including the weight of the prevailing wind direction, the weight of the secondary wind direction, and the weight of the leeward direction. Standardize the raw humidity data to a predetermined range; For unknown pollen counts of a variety, the mean of the same variety was used as a substitute, and parameter sensitivity was marked. A correspondence was established between the measured gene drift frequency and the measured gene drift threshold distance in the historical gene drift observation data, which served as the benchmark dataset for model calibration.

4. The wind speed-driven transgenic maize gene drift prediction method according to claim 1, characterized in that, Step S2 specifically includes: S201. Construct a Gaussian plume diffusion model, using preprocessed wind speed, wind direction, humidity, atmospheric stability, location, range, and altitude as inputs, and combine them with the pollen source intensity of the variety to determine the gene drift frequency spectrum at different spatiotemporal nodes within the target area. S202. Construct a Bayesian neural field model, using spatiotemporal coordinates and gene drift frequency maps as inputs, the gene drift data as model observations, and preprocessed environmental parameters and maize variety-specific parameters as covariates; wherein, wind speed and variety pollen quantity covariates are assigned an initial weight higher than other environmental parameter covariates; S203. Using a stochastic variational inference method, the lower bound of evidence is maximized to optimize the model parameters of the Bayesian neural field fusion framework, and the final weight coefficients in the Bayesian neural field model are further determined. S204. Generate multiple sets of parameters through Monte Carlo sampling, calculate the predicted values ​​of gene drift frequency and gene drift flow threshold distance, and determine their confidence intervals.

5. The wind speed-driven transgenic maize gene drift prediction method according to claim 4, characterized in that, In step S201, the Gaussian feather diffusion model is used to describe the spatial diffusion process of maize pollen, including the following formula: （1） （2） （3） （4） In the formula, G represents the gene drift frequency, Q represents the pollen source release intensity, and C represents the pollen concentration in the air. These represent the number of pollen grains deposited on the filaments of the donor and recipient flowers, respectively. and These are the genetic competitiveness parameters for genetically modified and non-genetically modified maize, respectively. These are the horizontal and vertical diffusion parameters determined by wind speed, respectively. For the corn ears to be tall, For the sedimentation rate of corn pollen, denoted by , x represents the wind speed at the top of the corn cob; x represents the distance in the downwind direction; y represents the distance perpendicular to the downwind direction; and z represents the vertical height. The cumulative leaf area index, This indicates the percentage of pollen grains that have penetrated the canopy.

6. The wind speed-dominated transgenic maize gene drift prediction method according to claim 4, characterized in that, In step S202, the Bayesian neural field model adopts a 3-layer fully connected network structure with 128 nodes in each layer, uses the ELU activation function, and introduces a Dropout layer with a dropout rate of 0.2 in each layer.

7. The wind speed-driven transgenic maize gene drift prediction method according to claim 1, characterized in that, Step S3 specifically includes: S301. In the spatial dimension, output the spatial distribution of pollen concentration within the target region and label the spatial boundary of the gene flow threshold distance; S302. In the time dimension, output the hourly pollen diffusion dynamics during the peak pollen shedding period in the flowering period, and trigger an early warning when extreme weather conditions are identified; S303. In the MTD project output results, the gene flow threshold distance and its uncertainty range are output according to different ecological zones, variety types and weather scenarios; S304. Based on the aforementioned spatial dimension, time dimension, and MTD-specific output results, generate a risk prevention and control report that includes recommendations for setting isolation distances between genetically modified and non-genetically modified corn.

8. The method for predicting transgenic maize gene drift based on wind speed as described in claim 7, characterized in that, In step S3, when the uncertainty range of the gene flow threshold distance exceeds the preset first range, the area is marked as an area requiring key verification in the risk prevention and control report.

9. The wind speed-dominated transgenic maize gene drift prediction method according to claim 1, characterized in that, The environmental parameters are acquired through a data acquisition device, which includes sensors for measuring wind speed, wind direction, and temperature / humidity, as well as a LoRa, 4G / 5G, or Bluetooth transmission module for data transmission.

Images