Pig breeding screening method and system based on body type database

Abstract

The invention discloses a pig breeding screening method and system based on a body type database, and particularly relates to the technical field of pig breeding screening. The system comprises a multi-modal sign data acquisition and integration module for collecting and standardizing heterogeneous data sources such as phenotypic signs, genomics and environmental parameters of breeding pigs in real time, and a multi-dimensional biological characteristic modeling and analysis module for analyzing a nonlinear action mechanism among genetic markers, phenotypic characteristics and environmental factors, the genetic potential intelligent judgment module is used for quantitatively evaluating spatial expression potential and intergenerational transmission probability of genetic advantages of the breeding pigs through a deep learning algorithm; and the intelligent screening and dynamic monitoring module is used for establishing a multi-dimensional threshold dynamic regulation and control mechanism to realize accurate elimination decision and continuous optimization of a germplasm resource library. According to the method, the screening accuracy is remarkably improved by designing multi-source data fusion and online optimization capability, meanwhile, mistaken elimination is reduced through a redundant superior species temporary storage management module, and collaborative improvement of population genetic diversity and environmental adaptability is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of pig breeding and screening, and in particular to a pig breeding and screening method and system based on a body shape database. Background Art

[0002] Pig breeding and selection is a core technology path for achieving genetic improvement and increasing meat quality and yield in the livestock industry. This requires a dynamic balance between molecular genetics, environmental adaptability, and economic efficiency. Current mainstream breeding systems rely on pedigree profiles and single-dimensional phenotypic data (such as daily weight gain and backfat thickness), combined with BLUP (Best Linear Unbiased Prediction) models to assess genetic value.

[0003] The main technical pain points of existing technologies lie in the lack of spatiotemporal alignment and heterogeneous feature fusion of multi-source data (physiological parameters, population genomes, and microenvironmental factors). Data silos result in marker-assisted selection efficiency below 30%. Screening algorithms rely on static thresholds (such as preset EBV thresholds), making it impossible to dynamically track the adaptive evolution of candidate pig breeders across generations using the LSTM-Attention temporal model. This leads to a 15% probability of inadvertent elimination of high-quality genetic resources. Herd management lacks a redundant superior stock storage mechanism and the NSGA-II multi-objective optimization strategy. This leads to excessive inbreeding coefficient growth (an average annual increase of >0.8%) and an annual loss of genetic diversity exceeding 3%, threatening the herd's disease resistance and environmental adaptability. Therefore, we provide a pig breeding screening method and system based on a body type database. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the existing technology and to propose a pig breeding screening method and system based on a body shape database.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A pig breeding screening method and system based on a body type database includes a multimodal body type data acquisition and integration module for real-time collection and standardized processing of heterogeneous data sources such as phenotypic body types, genomics, and environmental parameters of breeding pigs; a multidimensional biological characteristic modeling and analysis module for analyzing the nonlinear interaction mechanism between genetic markers, phenotypic characteristics, and environmental factors; a genetic potential intelligent identification module for quantitatively evaluating the spatial expression potential and intergenerational transmission probability of breeding pig genetic advantages through deep learning algorithms; and an intelligent screening and dynamic monitoring module for establishing a multidimensional threshold dynamic control mechanism to achieve precise culling decisions and continuous optimization of the germplasm resource bank.

[0007] The multimodal vital sign data acquisition integration module includes a wireless wearable sensing module for collecting dynamic vital sign data such as body temperature, heart rate, and gait trajectory in real time through implantable biosensors and synchronously integrating them into a cloud database; a three-dimensional visual modeling module for using a depth camera array to acquire seven core body parameters such as body length / backfat thickness / muscle density to construct an individual three-dimensional digital twin; a genomic metadata capture module for connecting SNP chips and Hi-C three-dimensional sequencing data to extract structural variation information; an environmental ecology collection node module for monitoring pen temperature and humidity / feed intake / microbial composition to establish a digital archive of the feeding environment; and a historical pedigree tracing module for integrating three generations of genetic pedigrees and trait transmission records of breeding pigs;

[0008] The multi-dimensional biological feature modeling and analysis module includes a multimodal data fusion engine module for associating the dynamic interaction network of phenotype, genome and environment based on knowledge graph technology, a genetic-phenotype association decoding module for locating key SNP sites and haplotype blocks affecting body shape using the GWAS framework, a structural variation deep analysis module for analyzing the regulatory mechanism of CNV / SVs on body shape traits in combination with pig graphic pan-genome data, an epigenetic quantification module for detecting three-dimensional chromatin interactions and identifying long-range enhancement of muscle growth through Hi-ChIP technology, and an environmental fitness calculation module for constructing a dynamic G×E interaction model to evaluate the stability of body shape genotype expression under different feeding conditions.

[0009] The present invention is further configured as follows: the genetic potential intelligent discrimination module includes a candidate gene intelligent recommendation module for performing functional weighted scoring on GWAS significant region genes based on a convolutional neural network-based scoring model, a multidimensional genetic selection index generation module for integrating the Fst genetic differentiation index and the genomic breeding value to generate a composite selection index, a non-reversible defect screening module for establishing a mutation pool of the PDZD2 and MTMR12 core genes to identify lethal mutations that affect body shape stability, a phenotypic plasticity prediction module for screening phenotypically robust individuals by predicting the response amplitude of traits to environmental fluctuations through multi-omics data, and a dynamic reference group optimization module for constructing a multi-layer mixed reference group according to variety characteristics;

[0010] The intelligent screening and dynamic monitoring module includes a real-time abnormal sign detection module for identifying short-term physiological fluctuations based on a time series model to distinguish expected changes from pathological changes, a dynamic prediction module for genetic risk for establishing a linear mixed model to evaluate the ability of candidate breeding pigs to maintain their body shape throughout the breeding cycle, an adaptive filtering threshold module for adjusting the screening criteria in stages according to the genetic progress of the population, a redundant superior breed temporary storage management module for setting an observation period for individuals with temporary negative signs but carrying excellent genotypes and re-evaluating them through a phenotypic correction program, and an elimination feedback self-optimization module for recording misjudgment cases and continuously improving the discrimination algorithm parameters through edge computing nodes.

[0011] The present invention is further configured as follows: the wireless wearable sensing module integrates dynamic vital sign data collected by biosensors and 3D body parameters acquired in depth through spatiotemporal calibration, allowing the three-dimensional visual modeling module to construct a digital twin model with physiological parameters; the three-dimensional visual modeling module cross-correlates seven core body parameters with structural variation data captured by the SNP chip, allowing the genome metadata extraction module to establish a preliminary mapping map of body characteristics and gene mutations; the metadata capture module synchronously annotates epigenetic markers of genetic data with environmental temperature, humidity and feed parameters, allowing the environmental ecology node module to generate a spatiotemporal coupling data set for G×E interaction analysis; the environmental ecology acquisition node module associates the current environmental archive with three-generation genetic pedigree data to the historical pedigree tracing module to construct a longitudinal genetic transmission link including environmental coefficients.

[0012] The present invention is further configured as follows: the multimodal data fusion engine module enables the genetic-phenotype association decoding module to accelerate the causal relationship mining of trait-associated SNP sites through the dynamic network annotated by the knowledge graph; the genetic-phenotype association decoding module enables the structural variation deep analysis module to analyze the physical pathway of CNV / SVs regulating body shape traits through three-dimensional chromatin folding through the significant sites located by GWAS; the structural variation deep analysis module enables the epigenetic quantification module to guide the Hi-ChIP experiment to focus on the enhancer region to verify the long-range regulatory mechanism through the structural variation boundary data; the epigenetic quantification module enables the environmental fitness calculation module to quantify the genotype expression robustness threshold under different feeding conditions through the chromatin conformation data.

[0013] The present invention is further configured as follows: the candidate gene intelligent recommendation module uses the weighted values of functional genes screened by the CNN model to allow the multidimensional genetic selection index generation module to collaboratively generate a composite evaluation vector; the multidimensional genetic selection index generation module uses qualified individuals to allow the irreversible defect screening module to locate PDZD2 / MTMR12 core gene mutations for lethal filtering; the irreversible defect screening module uses surviving candidate individuals to allow the phenotypic plasticity prediction module to establish a response function model to determine the boundary of the environmental stability domain; the phenotypic plasticity prediction module uses robustness prediction results to allow the dynamic reference group optimization module to iteratively update to form a multi-level genetic diversity benchmark;

[0014] The real-time abnormal physical sign detection module allows the genetic risk dynamic prediction module to adjust the assessment confidence interval of the breeding pig's body shape maintenance ability through the physical sign fluctuation classification results; the genetic risk dynamic prediction module allows the adaptive filtering threshold module to drive the phased automatic parameter adjustment mechanism through the risk probability distribution; the adaptive filtering threshold module allows the redundant superior breed temporary storage management module to activate the deviation re-evaluation program through the threshold trigger instruction; the redundant superior breed temporary storage management module allows the elimination feedback self-optimization module to realize the incremental online optimization of the discrimination algorithm through the misjudgment case flow data.

[0015] The present invention is further configured to include the following steps:

[0016] S1, Multi-source vital sign data fusion and synchronous modeling;

[0017] S2. Multi-level analysis of the genetic regulatory mechanisms of traits;

[0018] S3, intelligent evaluation and dynamic optimization of genetic advantages;

[0019] S4. Dynamic decision threshold iteration and continuous optimization of germplasm bank.

[0020] The present invention is further configured as follows: in the step S1, multi-source vital sign data fusion and synchronous modeling:

[0021] S1.1. Deploy an implantable monitoring device (sampling rate ≥ 200 Hz) equipped with a graphene-based tattoo sensor in the breeding pen to capture eight types of physiological time series signals, including the coefficient of variation of body temperature (CV < 3%) and gait dynamics parameters (plantar pressure distribution map), in real time. Simultaneously, an RGB-D depth camera array (NVIDIA Jetson TX2 driver) is used to obtain back fat thickness (unit: mm), lumbar muscle cross-sectional area (cm 2 ) and other seven core body parameters, a personalized 3D digital twin model is constructed based on Blender, and the point cloud registration algorithm (ICP accuracy ±0.5mm) is integrated to achieve spatiotemporal calibration of dynamic vital signs and spatial morphology, generating a multimodal vital sign data stream with time and space stamps;

[0022] S1.2. Whole blood samples were genotyped using the Affymetrix Axiom Porcine SNP60 array (SNP density ≥ 60K), combined with Hi-C sequencing (Illumina NovaSeq 6000, PE150) to identify structural variants (SV resolution ≤ 1 kb). Fecal microbiome analysis was performed using Metatranscriptomics (QIIME2 platform). Environmental parameters were continuously recorded using a LoRa wireless sensor network (temperature and humidity sensor accuracy ±0.5°C) to record pen microclimate. The nutritional composition of individual diets (crude protein content 16%-18%) was recorded in conjunction with an automated feeding system. A Neo4j knowledge graph was constructed to map entity relationships between genomic variation (PDZD2 gene CNV), microbial abundance (Firmicutes / Bacteroidetes ratio), and dietary parameters.

[0023] S1.3. Use TensorFlow framework to build a cross-domain autoencoder, project physical signs (temporal fluctuation of backfat thickness) and genomic features (rs808582 locus genotype) into a unified latent space, and calculate the feature contribution (weight error) between different modalities through the multi-head attention mechanism.

[0024] Pearson correlation analysis (p-value < 0.01) was performed between three-dimensional morphological parameters (body length growth rate) and SNP chip data, and a QTL region (SSC7: 32-35 Mb) significantly associated with body size stability was screened out.

[0025] The present invention is further configured as follows: in the step S2, multi-level analysis of the genetic regulation mechanism of traits:

[0026] S2.1. GWAS analysis was conducted on 1056 Large White pigs using GEMMA software (mixed linear model, Bonferroni correction, p < 5 × 10 -8 ), located a significant SNP (rs3212312, MAF>0.15) affecting backfat thickness, and combined with Hi-C chromatin interaction mapping (Juicer tool) to identify a physical interaction between the upstream enhancer containing rs808582 (1.2Mb from the TSS) and the MYOD1 promoter (contact frequency>10TPM). Using Crispr / Cas9 editing of porcine fibroblasts, they validated the regulatory effect of chromatin looping on myocyte differentiation (Myosin Heavy Chain expression increased by 2.3-fold);

[0027] S2.2. We identified a breed-specific SV (a 15 kb deletion at 52 Mb on chromosome 3 in Duroc pigs) using the GraphGenome construction tool. We analyzed the association strength between the SV and backfat thickness using MatrixEQTL (β = -0.34, se = 0.08). We detected differences in chromatin accessibility in the SV region using single-cell ATAC-seq (10xGenomics platform) (Pielou index difference > 0.2), and used the IBIS algorithm to predict the spatial disruption effect of the SV on a transcription factor binding site (MEF2C) (ΔAffinity > 15%).

[0028] S2.3. A generalized additive mixed model (GAMM) was constructed to quantify the interaction between feed conversion rate (FCR) and PDZD2 genotype (ΔAIC>6). Control experiments were conducted under different temperature and humidity gradients (10°C-30°C, RH50%-90%) to measure the expression dynamics of the candidate gene (FST) (coefficient of variation of qRT-PCR Ct value <5%) and predict its environmental adaptation threshold (R 2 >0.85), the key parameters are integrated into the environmental stability index (ESI), the formula is:

[0029] ESI=Σ(β_i×Env_j)+ε

[0030] Where β_i represents the genotype effect and Env_j is the standardized environmental variable.

[0031] The present invention is further configured as follows: in the step S3, intelligent evaluation and dynamic optimization of genetic advantages:

[0032] S3.1. Develop the Inception-v4 convolutional neural network and input the LD blocks of the GWAS significant regions (r 2 >0.8) and Hi-C interaction matrix, output gene function score (AUC=0.93), calculate the Fst genetic differentiation index (θ>0.25) for the Shenxian pig population, combine the genomic estimated breeding value (GEBV, estimated by REML algorithm), generate a composite selection index CSI=0.6×GEBV+0.3×Fst+0.1×ESI, use Plink to perform genome-wide association screening, and retain the top 10% individuals in the CSI ranking;

[0033] S3.2. A 50bp probe library was designed for the PDZD2 gene (Chr14: 72Mb). Frameshift mutations and splice site variations (c.1234delG) were detected by Nanopore long-read sequencing (Q20>90%). Combined with CRISPR-Cas9 phenotypic verification (embryo lethality>65%), a lethal mutation blacklist was established, and multi-omics robustness analysis was performed on candidate individuals, including transcriptome PCA distance (principal component contribution rate>85%) and metabolome CV (<15%), to screen individuals with phenotypic fluctuations within ±2σ.

[0034] The present invention is further configured as follows: in the step S4, dynamic decision threshold iteration and germplasm bank continuous optimization:

[0035] S4.1. Deploy an LSTM-Attention temporal model (hidden unit number = 128) to analyze abnormal LF / HF power ratio events (Z-score > 3) in real time in heart rate variability (HRV) to distinguish between stress reactions (reversible) and chronic diseases (irreversible). Build a Cox proportional hazards model that integrates the annual rate of change in body parameters (backfat thickness Δ≤ 0.5 mm / year) and the number of environmental fluctuations (n > 5 / quarter) to predict individual genetic stability (Harrell's C-index = 0.79).

[0036] S4.2. Dynamically balance genetic progress (ΔGEBV ≥ 0.3SD / generation) and inbreeding coefficient (ΔF < 0.5%) using the NSGA-Ⅱ algorithm. Update the screening threshold every breeding cycle. -6 ) The priority upgrade mechanism was initiated (weight increased by 20%), and the impact of threshold adjustment on the effective size of the population (Ne>100) was evaluated through Monte Carlo simulation to ensure the genetic diversity baseline.

[0037] The beneficial effects of the present invention are:

[0038] 1. The present invention uses a multimodal vital sign data acquisition integration module to collect the physiological, genomic and environmental parameters of breeding pigs in real time, and constructs a three-dimensional digital twin model and a spatiotemporal coupling data set. With the help of a multidimensional biological feature modeling and analysis module, it locates key genetic loci, analyzes the mechanism by which structural variations regulate fat deposition through long-range chromatin interactions, and uses the convolutional neural network and composite genetic index of the genetic potential intelligent discrimination module to quantify genotype advantages. Through the LSTM-Attention model and NSGA-Ⅱ algorithm of the intelligent screening and dynamic monitoring module, the system dynamically adjusts the screening threshold, successfully avoids lethal mutations, increases the survival rate by 32%, and achieves an annual genetic progress of 0.4 standard deviations. The system's multi-source data fusion and online optimization capabilities significantly improve screening accuracy. At the same time, the redundant superior seed temporary storage management module reduces erroneous elimination, ensuring the coordinated improvement of population genetic diversity and environmental adaptability.

[0039] 2. The present invention integrates vital sign data collected by graphene sensors with structural variation data from Hi-C sequencing through spatiotemporal calibration fusion technology, establishes a cross-domain autoencoder to map multimodal features, uses GWAS and Hi-C interaction maps to identify regulatory sites, combines CRISPR to verify the regulatory effects of chromatin loops on muscle growth, predicts the environmental stability index through random forest regression and generates gene function scores through the Inception-v4 network, and dynamically optimizes the weights of composite selection indicators. In controlled experiments, this method achieves an annual increase in backfat thickness of ≤0.5 mm and an increase in the inbreeding coefficient of <0.4%. Monte Carlo simulation is used to balance genetic progress and effective population size, ultimately achieving the goals of precise breeding and sustainable genetic improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the system module in the present invention.

[0041] Figure 2 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0043] Example 1

[0044] like Figure 1As shown, a pig breeding and screening system based on a body shape database includes a multimodal body sign data acquisition and integration module for real-time collection and standardized processing of heterogeneous data sources such as phenotypic body signs, genomics, and environmental parameters of breeding pigs; a multidimensional biological feature modeling and analysis module for analyzing the nonlinear interaction mechanism between genetic markers, phenotypic characteristics, and environmental factors; a genetic potential intelligent discrimination module for quantitatively evaluating the spatial expression potential and intergenerational transmission probability of breeding pig genetic advantages through deep learning algorithms; and an intelligent screening and dynamic monitoring module for establishing a multidimensional threshold dynamic control mechanism to achieve precise elimination decisions and continuous optimization of germplasm resource banks;

[0045] The multimodal vital sign data acquisition integration module includes a wireless wearable sensing module for collecting dynamic vital sign data such as body temperature, heart rate, and gait trajectory in real time through implantable biosensors and synchronously integrating them into a cloud database; a three-dimensional visual modeling module for using a depth camera array to acquire seven core body parameters such as body length / backfat thickness / muscle density to construct an individual three-dimensional digital twin; a genomic metadata capture module for connecting SNP chips and Hi-C three-dimensional sequencing data to extract structural variation information; an environmental ecology collection node module for monitoring pen temperature and humidity / feed intake / microbial composition to establish a digital archive of the feeding environment; and a historical pedigree tracing module for integrating three generations of genetic pedigrees and trait transmission records of breeding pigs;

[0046] The multi-dimensional biological feature modeling and analysis module includes a multimodal data fusion engine module for associating the dynamic interaction network of phenotype, genome and environment based on knowledge graph technology, a genetic-phenotype association decoding module for locating key SNP sites and haplotype blocks affecting body shape using the GWAS framework, a structural variation deep analysis module for analyzing the regulatory mechanism of CNV / SVs on body shape traits by combining pig graphic pan-genome data, an epigenetic quantification module for detecting three-dimensional chromatin interactions and identifying long-range enhancements regulating muscle growth through Hi-ChIP technology, and an environmental fitness calculation module for constructing a dynamic G×E interaction model to evaluate the stability of body shape genotype expression under different feeding conditions;

[0047] The genetic potential intelligent discrimination module includes a candidate gene intelligent recommendation module for performing functional weighted scoring on GWAS significant region genes based on a convolutional neural network-based scoring model, a multidimensional genetic selection index generation module for integrating the Fst genetic differentiation index and genomic breeding values to generate a composite selection index, a non-reversible defect screening module for establishing a mutation pool of the PDZD2 and MTMR12 core genes to identify lethal mutations that affect body shape stability, a phenotypic plasticity prediction module for predicting the response amplitude of traits to environmental fluctuations through multi-omics data to screen phenotypically robust individuals, and a dynamic reference group optimization module for constructing a multi-layer mixed reference group according to variety characteristics.

[0048] The intelligent screening and dynamic monitoring module includes a real-time abnormal sign detection module for identifying short-term physiological fluctuations based on a time series model to distinguish expected changes from pathological changes; a dynamic genetic risk prediction module for establishing a linear mixed model to evaluate the ability of candidate breeding pigs to maintain their body shape throughout the reproductive cycle; an adaptive filtering threshold module for adjusting the screening criteria in stages according to the genetic progress of the population; a redundant superior breed temporary management module for setting an observation period for individuals with temporary negative physical signs but carrying superior genotypes and re-evaluating them through a phenotypic correction program; and an elimination feedback self-optimization module for recording misjudgment cases and continuously improving the discrimination algorithm parameters through edge computing nodes;

[0049] The wireless wearable sensing module integrates dynamic vital sign data collected by biosensors with 3D body parameters acquired in depth through spatiotemporal calibration, allowing the 3D visual modeling module to construct a digital twin model with physiological parameters. The 3D visual modeling module cross-correlates seven core body parameters with structural variation data captured by the SNP chip, allowing the genomic metadata extraction module to establish a preliminary mapping map of body characteristics and gene mutations. The metadata capture module synchronously annotates epigenetic markers of gene data with environmental temperature, humidity and feed parameters, allowing the environmental ecology node module to generate a spatiotemporal coupling data set for G×E interaction analysis. The environmental ecology acquisition node module associates the current environmental archive with three-generation genetic pedigree data to construct a longitudinal genetic transmission link containing environmental coefficients through the historical pedigree tracing module.

[0050] The multimodal data fusion engine module uses a dynamic network annotated with a knowledge graph to enable the genetic-phenotype association decoding module to accelerate the causal relationship mining of trait-associated SNP sites; the genetic-phenotype association decoding module uses significant sites located by GWAS to enable the structural variation deep analysis module to analyze the physical pathways by which CNVs / SVs regulate body shape traits through three-dimensional chromatin folding; the structural variation deep analysis module uses structural variation boundary data to enable the epigenetic quantification module to guide Hi-ChIP experiments to focus on enhancer regions to verify long-range regulatory mechanisms; the epigenetic quantification module uses chromatin conformation data to enable the environmental fitness calculation module to quantify the robustness threshold of genotype expression under different feeding conditions;

[0051] The candidate gene intelligent recommendation module uses the weighted values of functional genes screened by the CNN model to enable the multidimensional genetic selection index generation module to collaboratively generate a composite evaluation vector; the multidimensional genetic selection index generation module uses qualified individuals to enable the irreversible defect screening module to locate PDZD2 / MTMR12 core gene mutations for lethal filtering; the irreversible defect screening module uses surviving candidate individuals to enable the phenotypic plasticity prediction module to establish a response function model to determine the boundary of the environmental stability domain; the phenotypic plasticity prediction module uses robustness prediction results to enable the dynamic reference group optimization module to iteratively update to form a multi-level genetic diversity benchmark;

[0052] The real-time abnormal physical sign detection module allows the genetic risk dynamic prediction module to adjust the assessment confidence interval of the breeding pig's body shape maintenance ability through the physical sign fluctuation classification results; the genetic risk dynamic prediction module allows the adaptive filtering threshold module to drive the phased automatic parameter adjustment mechanism through the risk probability distribution; the adaptive filtering threshold module allows the redundant superior breed temporary storage management module to activate the deviation re-evaluation program through the threshold trigger instruction; the redundant superior breed temporary storage management module allows the elimination feedback self-optimization module to realize the incremental online optimization of the discrimination algorithm through the misjudgment case flow data.

[0053] In the above embodiment, the multimodal vital sign data acquisition integration module uses wireless wearable sensors to collect dynamic vital signs such as the heart rate and gait of the breeding pig in real time, and combines it with a depth camera array to generate a three-dimensional digital twin containing seven parameters such as body length and back fat thickness, and synchronously integrates SNP chip and Hi-C sequencing data through the genome metadata capture module.

[0054] After applying this system in the breeding farm, it was found that the back fat thickness of breeding pigs carrying a specific structural variation (CNV) was significantly lower than the group average, and the structural variation deep analysis module was used to verify that the variation regulates fat deposition through long-range chromatin interactions.

[0055] Subsequently, the genetic potential intelligent discrimination module used convolutional neural networks to recommend candidate genes (PDZD2) and combined them with the environmental fitness model to evaluate their expression stability under different feeding conditions.

[0056] During the dynamic monitoring phase, if individuals with superior genotypes experience abnormal short-term backfat thickening, the redundant superior breed temporary management module will initiate an observation period. Using a phenotypic correction program, it will determine whether these individuals exhibit pathological variations, thus preventing inadvertent elimination. Ultimately, the system continuously optimizes screening thresholds and algorithm parameters to help improve both the genetic progress and phenotypic robustness of the breeding pig population.

[0057] Example 2

[0058] like Figure 1-2 As shown, a pig breeding screening method based on a body shape database comprises the following steps:

[0059] S1, Multi-source vital sign data fusion and synchronous modeling;

[0060] S2. Multi-level analysis of the genetic regulatory mechanisms of traits;

[0061] S3, intelligent evaluation and dynamic optimization of genetic advantages;

[0062] S4, dynamic decision threshold iteration and continuous optimization of germplasm bank;

[0063] In the step S1, multi-source vital sign data fusion and synchronous modeling:

[0064] S1.1. Deploy an implantable monitoring device (sampling rate ≥ 200 Hz) equipped with a graphene-based tattoo sensor in the breeding pen to capture eight types of physiological time series signals, including the coefficient of variation of body temperature (CV < 3%) and gait dynamics parameters (plantar pressure distribution map), in real time. Simultaneously, an RGB-D depth camera array (NVIDIA Jetson TX2 driver) is used to obtain back fat thickness (unit: mm), lumbar muscle cross-sectional area (cm 2 ) and other seven core body parameters, a personalized 3D digital twin model is constructed based on Blender, and the point cloud registration algorithm (ICP accuracy ±0.5mm) is integrated to achieve spatiotemporal calibration of dynamic vital signs and spatial morphology, generating a multimodal vital sign data stream with time and space stamps;

[0065] S1.2. Whole blood samples were genotyped using the Affymetrix Axiom Porcine SNP60 array (SNP density ≥ 60K), combined with Hi-C sequencing (Illumina NovaSeq 6000, PE150) to identify structural variants (SV resolution ≤ 1 kb). Fecal microbiome analysis was performed using Metatranscriptomics (QIIME2 platform). Environmental parameters were continuously recorded using a LoRa wireless sensor network (temperature and humidity sensor accuracy ±0.5°C) to record pen microclimate. The nutritional composition of individual diets (crude protein content 16%-18%) was recorded in conjunction with an automated feeding system. A Neo4j knowledge graph was constructed to map entity relationships between genomic variation (PDZD2 gene CNV), microbial abundance (Firmicutes / Bacteroidetes ratio), and dietary parameters.

[0066] S1.3. A cross-domain autoencoder was constructed using the TensorFlow framework. Physical characteristics (temporal fluctuations in backfat thickness) and genomic features (rs808582 genotype) were projected into a unified latent space. A multi-head attention mechanism was used to calculate the contribution of features between different modalities (weight error < 5%). Pearson correlation analysis (p-value < 0.01) was performed on three-dimensional morphological parameters (body length growth rate) and SNP chip data, identifying a QTL region (SSC7: 32-35 Mb) significantly associated with body shape stability.

[0067] In the step S2, multi-level analysis of the genetic regulation mechanism of traits:

[0068] S2.1. GWAS analysis was conducted on 1056 Large White pigs using GEMMA software (mixed linear model, Bonferroni correction, p < 5 × 10 -8 ), located a significant SNP (rs3212312, MAF>0.15) affecting backfat thickness, and combined with Hi-C chromatin interaction mapping (Juicer tool) to identify a physical interaction between the upstream enhancer containing rs808582 (1.2Mb from the TSS) and the MYOD1 promoter (contact frequency>10TPM). Using Crispr / Cas9 editing of porcine fibroblasts, they validated the regulatory effect of chromatin looping on myocyte differentiation (Myosin Heavy Chain expression increased by 2.3-fold);

[0069] S2.2. We identified a breed-specific SV (a 15 kb deletion at 52 Mb on chromosome 3 in Duroc pigs) using the GraphGenome construction tool. We analyzed the association strength between the SV and backfat thickness using MatrixEQTL (β = -0.34, se = 0.08). We detected differences in chromatin accessibility in the SV region using single-cell ATAC-seq (10xGenomics platform) (Pielou index difference > 0.2), and used the IBIS algorithm to predict the spatial disruption effect of the SV on a transcription factor binding site (MEF2C) (ΔAffinity > 15%).

[0070] S2.3. A generalized additive mixed model (GAMM) was constructed to quantify the interaction between feed conversion rate (FCR) and PDZD2 genotype (ΔAIC>6). Control experiments were conducted under different temperature and humidity gradients (10°C-30°C, RH50%-90%) to measure the expression dynamics of the candidate gene (FST) (coefficient of variation of qRT-PCR Ct value <5%) and predict its environmental adaptation threshold (R 2>0.85), and the key parameters were integrated into the environmental stability index (ESI), with the formula: ESI = Σ(β_i×Env_j)+ε, where β_i represents the genotype effect and Env_j is the standardized environmental variable;

[0071] In the step S3, intelligent evaluation and dynamic optimization of genetic advantages:

[0072] S3.1. Develop the Inception-v4 convolutional neural network and input the LD blocks of the GWAS significant regions (r 2 >0.8) and Hi-C interaction matrix, output gene function score (AUC=0.93), calculate the Fst genetic differentiation index (θ>0.25) for the Shenxian pig population, combine the genomic estimated breeding value (GEBV, estimated by REML algorithm), generate a composite selection index CSI=0.6×GEBV+0.3×Fst+0.1×ESI, use Plink to perform genome-wide association screening, and retain the top 10% individuals in the CSI ranking;

[0073] S3.2. Design a 50bp probe library for the PDZD2 gene (Chr14: 72Mb). Detect frameshift mutations and splice site variants (c.1234delG) using Nanopore long-read sequencing (Q20 > 90%). Combined with CRISPR-Cas9 phenotypic verification (embryo lethality > 65%), establish a blacklist of lethal mutations. Perform multi-omics robustness analysis on candidate individuals, including transcriptome PCA distance (principal component contribution > 85%) and metabolome CV (< 15%), to screen individuals with phenotypic fluctuations within ± 2σ.

[0074] In the step S4, dynamic decision threshold iteration and continuous optimization of germplasm bank:

[0075] S4.1. Deploy an LSTM-Attention temporal model (hidden layer units = 128) to analyze abnormal LF / HF power ratio events (Z-score > 3) in real time in heart rate variability (HRV) to distinguish between stress reactions (reversible) and chronic diseases (irreversible). A Cox proportional hazards model was constructed, integrating the annual rate of change in body parameters (backfat thickness Δ≤ 0.5 mm / year) and the number of environmental fluctuations (n > 5 / quarter) to predict individual genetic stability risk (Harrell's C-index = 0.79).

[0076] S4.2. Dynamically balance genetic progress (ΔGEBV ≥ 0.3SD / generation) and inbreeding coefficient (ΔF < 0.5%) using the NSGA-Ⅱ algorithm. Update the screening threshold every breeding cycle. -6) The priority upgrade mechanism was initiated (weight increased by 20%), and the impact of threshold adjustment on the effective size of the population (Ne>100) was evaluated through Monte Carlo simulation to ensure the genetic diversity baseline.

[0077] In the above embodiment, the system deploys graphene sensors to collect real-time vital sign data such as the coefficient of variation of pig body temperature (CV=2.1%) and gait pressure distribution, and combines with RGB-D cameras to obtain dynamic data of backfat thickness (annual average growth of 0.3mm).

[0078] When an individual's rs808582 genotype was detected to have an enhancer interaction with the MYOD1 gene (contact frequency 12.5 TPM), the system automatically increased its genomic score. CRISPR verification found that the efficiency of muscle cell differentiation increased by 2.1 times after editing this site, thereby increasing its CSI index from 78 to 91, and selecting it for inclusion in the core germplasm library. The screening threshold is dynamically adjusted every quarter using the NSGA-Ⅱ algorithm. When the environmental sensor detects a temperature fluctuation of >8°C for three consecutive days, the system increases the weight of the PDZD2 gene from 0.15 to 0.18 and triggers a lethal mutation screening, successfully avoiding the embryonic lethality risk of the c.1234delG mutation (survival rate increased by 32%), ultimately achieving an annual breeding progress of 0.4 genetic standard deviations while controlling the annual increase in the inbreeding coefficient to <0.4%.

[0079] Working Principle: When in use, the system uses implanted biosensors to collect dynamic vital signs of sows, such as body temperature, heart rate, and gait. It also uses a depth camera array to capture core body parameters, such as backfat thickness and body length, and constructs a digital twin model to accurately record individual three-dimensional body shape changes. This data is integrated with genomic data (e.g., SNP chip data) and environmental information (e.g., temperature, humidity, and feed composition) to establish a spatiotemporally synchronized database of physical, genomic, and environmental information, laying the foundation for subsequent analysis.

[0080] Next, through the multi-dimensional biological feature modeling and analysis module, the system conducts an in-depth analysis of the genetic mechanism of the breeding pigs. This module uses GWAS analysis to determine the key genetic sites that affect body shape traits, and combines Hi-C three-dimensional sequencing data to identify chromatin interactions to reveal the association between genes and phenotypes. For example, by analyzing specific genes (such as PDZD2 and MYOD1) and their structural variations, the system can determine how these genes affect the body shape stability or growth potential of breeding pigs. At the same time, using deep learning technology, the system evaluates the performance of genes under different environmental conditions to predict the adaptability and robustness of individuals to environmental fluctuations.

[0081] Based on the above analysis, the system uses the genetic potential intelligent discrimination module to quantitatively assess the genetic advantages of breeding pigs. Through deep learning methods such as convolutional neural networks (CNN), the advantages and disadvantages of genotypes are evaluated and a comprehensive selection index (CSI) is generated. On this basis, the system uses a multidimensional genetic selection index, combined with factors such as the genetic differentiation index (Fst) and phenotypic stability, to dynamically adjust the screening criteria for precise selection. In addition, the system also uses time series models to monitor the physical signs of breeding pigs in real time, analyze short-term physiological fluctuations, distinguish between normal and pathological changes, and avoid inadvertent elimination. During the dynamic decision-making stage, the system will adjust the elimination threshold in real time based on factors such as environmental changes, genetic expression, and breeding progress, optimize the germplasm pool, ensure genetic progress, avoid excessive inbreeding coefficients, and ensure genetic diversity. By continuously optimizing algorithms and thresholds, the system can adjust the genetic structure of the breeding pig population after each breeding cycle and improve breeding efficiency.

[0082] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A pig breeding screening system based on body shape database, characterized in that: It includes a multimodal physical sign data acquisition and integration module for real-time collection and standardized processing of heterogeneous data sources such as phenotypic signs, genomics, and environmental parameters of breeding pigs; a multidimensional biological characteristic modeling and analysis module for analyzing the nonlinear interaction mechanism between genetic markers, phenotypic characteristics, and environmental factors; a genetic potential intelligent identification module for quantitatively evaluating the spatial expression potential and intergenerational transmission probability of breeding pig genetic advantages through deep learning algorithms; and an intelligent screening and dynamic monitoring module for establishing a multidimensional threshold dynamic control mechanism to achieve precise elimination decisions and continuous optimization of germplasm resource banks; The multimodal vital sign data acquisition integration module includes a wireless wearable sensing module for collecting dynamic vital sign data such as body temperature, heart rate, and gait trajectory in real time through implantable biosensors and synchronously integrating them into a cloud database; a three-dimensional visual modeling module for using a depth camera array to acquire seven core body parameters such as body length / backfat thickness / muscle density to construct an individual three-dimensional digital twin; a genomic metadata capture module for connecting SNP chips and Hi-C three-dimensional sequencing data to extract structural variation information; an environmental ecology collection node module for monitoring pen temperature and humidity / feed intake / microbial composition to establish a digital archive of the feeding environment; and a historical pedigree tracing module for integrating three generations of genetic pedigrees and trait transmission records of breeding pigs; The multi-dimensional biological feature modeling and analysis module includes a multimodal data fusion engine module for associating the dynamic interaction network of phenotype, genome and environment based on knowledge graph technology, a genetic-phenotype association decoding module for locating key SNP sites and haplotype blocks affecting body shape using the GWAS framework, a structural variation deep analysis module for analyzing the regulatory mechanism of CNV / SVs on body shape traits in combination with pig graphic pan-genome data, an epigenetic quantification module for detecting three-dimensional chromatin interactions and identifying long-range enhancement of muscle growth through Hi-ChIP technology, and an environmental fitness calculation module for constructing a dynamic G×E interaction model to evaluate the stability of body shape genotype expression under different feeding conditions.

2. The pig breeding screening system based on body shape database according to claim 1, characterized in that: The genetic potential intelligent discrimination module includes a candidate gene intelligent recommendation module for performing functional weighted scoring on GWAS significant region genes based on a convolutional neural network-based scoring model, a multidimensional genetic selection index generation module for integrating the Fst genetic differentiation index and genomic breeding values to generate a composite selection index, a non-reversible defect screening module for establishing a mutation pool of the PDZD2 and MTMR12 core genes to identify lethal mutations that affect body shape stability, a phenotypic plasticity prediction module for predicting the response amplitude of traits to environmental fluctuations through multi-omics data to screen phenotypically robust individuals, and a dynamic reference group optimization module for constructing a multi-layer mixed reference group according to variety characteristics. The intelligent screening and dynamic monitoring module includes a real-time abnormal sign detection module for identifying short-term physiological fluctuations based on a time series model to distinguish expected changes from pathological changes, a dynamic prediction module for genetic risk for establishing a linear mixed model to evaluate the ability of candidate breeding pigs to maintain their body shape throughout the breeding cycle, an adaptive filtering threshold module for adjusting the screening criteria in stages according to the genetic progress of the population, a redundant superior breed temporary storage management module for setting an observation period for individuals with temporary negative signs but carrying excellent genotypes and re-evaluating them through a phenotypic correction program, and an elimination feedback self-optimization module for recording misjudgment cases and continuously improving the discrimination algorithm parameters through edge computing nodes.

3. The pig breeding screening system based on body shape database according to claim 1, characterized in that: The wireless wearable sensing module uses spatiotemporal calibration to fuse the dynamic vital sign data collected by biosensors with the 3D body parameters acquired in depth, allowing the 3D visual modeling module to construct a digital twin model with physiological parameters; the 3D visual modeling module cross-correlates seven core body parameters with the structural variation data captured by the SNP chip, allowing the genome metadata extraction module to establish a preliminary mapping map of body characteristics and gene mutations; the metadata capture module synchronously annotates the epigenetic markers of genetic data with environmental temperature, humidity and feed parameters, allowing the environmental ecology node module to generate a spatiotemporal coupling data set for G×E interaction analysis; the environmental ecology acquisition node module associates the current environmental archive with three-generation genetic pedigree data to the historical pedigree tracing module to construct a longitudinal genetic transmission link including environmental coefficients.

4. The pig breeding screening system based on body shape database according to claim 1, characterized in that: The multimodal data fusion engine module uses a dynamic network annotated by a knowledge graph to enable the genetic-phenotype association decoding module to accelerate the causal relationship mining of trait-associated SNP sites; the genetic-phenotype association decoding module uses significant sites located by GWAS to enable the structural variation deep analysis module to analyze the physical pathway by which CNV / SVs regulate body shape traits through three-dimensional chromatin folding; the structural variation deep analysis module uses structural variation boundary data to enable the epigenetic quantification module to guide Hi-ChIP experiments to focus on enhancer regions to verify remote regulatory mechanisms; the epigenetic quantification module uses chromatin conformation data to enable the environmental fitness calculation module to quantify the robustness threshold of genotype expression under different feeding conditions.

5. The pig breeding screening system based on body shape database according to claim 2, characterized in that: The candidate gene intelligent recommendation module uses the weighted values of functional genes screened by the CNN model to enable the multidimensional genetic selection index generation module to collaboratively generate a composite evaluation vector; the multidimensional genetic selection index generation module uses qualified individuals to enable the irreversible defect screening module to locate PDZD2 / MTMR12 core gene mutations for lethal filtering; the irreversible defect screening module uses surviving candidate individuals to enable the phenotypic plasticity prediction module to establish a response function model to determine the boundary of the environmental stability domain; the phenotypic plasticity prediction module uses robustness prediction results to enable the dynamic reference group optimization module to iteratively update to form a multi-level genetic diversity benchmark; The real-time abnormal physical sign detection module allows the genetic risk dynamic prediction module to adjust the assessment confidence interval of the breeding pig's body shape maintenance ability through the physical sign fluctuation classification results; the genetic risk dynamic prediction module allows the adaptive filtering threshold module to drive the phased automatic parameter adjustment mechanism through the risk probability distribution; the adaptive filtering threshold module allows the redundant superior breed temporary storage management module to activate the deviation re-evaluation program through the threshold trigger instruction; the redundant superior breed temporary storage management module allows the elimination feedback self-optimization module to realize the incremental online optimization of the discrimination algorithm through the misjudgment case flow data.

6. A pig breeding screening method based on a body shape database, characterized in that: The following steps are involved: S1, Multi-source vital sign data fusion and synchronous modeling; S2. Multi-level analysis of the genetic regulatory mechanisms of traits; S3, intelligent evaluation and dynamic optimization of genetic advantages; S4. Dynamic decision threshold iteration and continuous optimization of germplasm bank.

7. The pig breeding screening method based on body shape database according to claim 6, characterized in that: In the step S1, multi-source vital sign data fusion and synchronous modeling: S1.

1. Deploy an implantable monitoring device (sampling rate ≥ 200 Hz) equipped with a graphene-based tattoo sensor in the breeding pen to capture eight types of physiological time series signals, including the coefficient of variation of body temperature (CV < 3%) and gait dynamics parameters (plantar pressure distribution map), in real time. Simultaneously, an RGB-D depth camera array (NVIDIA Jetson TX2 driver) is used to obtain back fat thickness (unit: mm), lumbar muscle cross-sectional area (cm 2 ) and other seven core body parameters, a personalized 3D digital twin model is constructed based on Blender, and the point cloud registration algorithm (ICP accuracy ±0.5mm) is integrated to achieve spatiotemporal calibration of dynamic vital signs and spatial morphology, generating a multimodal vital sign data stream with time and space stamps; S1.

2. Whole blood samples were genotyped using the Affymetrix Axiom Porcine SNP60 array (SNP density ≥ 60K), combined with Hi-C sequencing (Illumina NovaSeq 6000, PE150) to identify structural variants (SV resolution ≤ 1 kb). Fecal microbiome analysis was performed using Metatranscriptomics (QIIME2 platform). Environmental parameters were continuously recorded using a LoRa wireless sensor network (temperature and humidity sensor accuracy ±0.5°C) to record pen microclimate. The nutritional composition of individual diets (crude protein content 16%-18%) was recorded in conjunction with an automated feeding system. A Neo4j knowledge graph was constructed to map entity relationships between genomic variation (PDZD2 gene CNV), microbial abundance (Firmicutes / Bacteroidetes ratio), and dietary parameters. S1.

3. A cross-domain autoencoder was constructed using the TensorFlow framework. Physical signs (temporal fluctuations in back fat thickness) and genomic features (rs808582 locus genotype) were projected into a unified latent space. The feature contribution between different modalities was calculated through the multi-head attention mechanism (weight error <5%). Pearson correlation analysis was performed on three-dimensional morphological parameters (body length growth rate) and SNP chip data (p-value < 0.01), and a QTL region (SSC7: 32-35Mb) significantly associated with body shape stability was screened out.

8. The pig breeding screening method based on body shape database according to claim 6, characterized in that: In the step S2, multi-level analysis of the genetic regulation mechanism of traits: S2.

1. GWAS analysis was conducted on 1056 Large White pigs using GEMMA software (mixed linear model, Bonferroni correction, p < 5 × 10 -8 ), located a significant SNP (rs3212312, MAF>0.15) affecting backfat thickness, and combined with Hi-C chromatin interaction mapping (Juicer tool) to identify a physical interaction between the upstream enhancer containing rs808582 (1.2Mb from the TSS) and the MYOD1 promoter (contact frequency>10TPM). Using Crispr / Cas9 editing of porcine fibroblasts, they validated the regulatory effect of chromatin looping on myocyte differentiation (Myosin Heavy Chain expression increased by 2.3-fold); S2.

2. We identified a breed-specific SV (a 15 kb deletion at 52 Mb on chromosome 3 in Duroc pigs) using the GraphGenome construction tool. We analyzed the association strength between the SV and backfat thickness using MatrixEQTL (β = -0.34, se = 0.08). We detected differences in chromatin accessibility in the SV region using single-cell ATAC-seq (10xGenomics platform) (Pielou index difference > 0.2), and used the IBIS algorithm to predict the spatial disruption effect of the SV on a transcription factor binding site (MEF2C) (ΔAffinity > 15%). S2.

3. A generalized additive mixed model (GAMM) was constructed to quantify the interaction between feed conversion rate (FCR) and PDZD2 genotype (ΔAIC>6). Control experiments were conducted under different temperature and humidity gradients (10°C-30°C, RH50%-90%) to measure the expression dynamics of the candidate gene (FST) (coefficient of variation of qRT-PCR Ct value <5%) and predict its environmental adaptation threshold (R 2 >0.85), key parameters are integrated into the environmental stability index (ESI), formula: ESI=Σ(β_i×Env_j)+ε Where β_i represents the genotype effect and Env_j is the standardized environmental variable.

9. The pig breeding screening method based on body shape database according to claim 6, characterized in that: In the step S3, intelligent evaluation and dynamic optimization of genetic advantages: S3.

1. Develop the Inception-v4 convolutional neural network and input the LD blocks of the GWAS significant regions (r 2 >0.8) and Hi-C interaction matrix, output gene function score (AUC=0.93), calculate the Fst genetic differentiation index (θ>0.25) for the Shenxian pig population, combine the genomic estimated breeding value (GEBV, estimated by REML algorithm), generate a composite selection index CSI=0.6×GEBV+0.3×Fst+0.1×ESI, use Plink to perform genome-wide association screening, and retain the top 10% individuals in the CSI ranking; S3.

2. A 50bp probe library was designed for the PDZD2 gene (Chr14: 72Mb). Frameshift mutations and splice site variations (c.1234delG) were detected using Nanopore long-read sequencing (Q20>90%). Combined with CRISPR-Cas9 phenotypic verification (embryo lethality>65%), a lethal mutation blacklist was established, and multi-omics robustness analysis was performed on candidate individuals, including transcriptome PCA distance (principal component contribution rate>85%) and metabolome CV (<15%), to screen individuals with phenotypic fluctuations within ±2σ.

10. The pig breeding screening method based on body shape database according to claim 6, characterized in that: In the step S4, dynamic decision threshold iteration and continuous optimization of germplasm bank: S4.

1. Deploy an LSTM-Attention temporal model (hidden unit number = 128) to analyze abnormal LF / HF power ratio events (Z-score > 3) in real time in heart rate variability (HRV) to distinguish between stress reactions (reversible) and chronic diseases (irreversible). Build a Cox proportional hazards model that integrates the annual rate of change in body parameters (backfat thickness Δ≤ 0.5 mm / year) and the number of environmental fluctuations (n > 5 / quarter) to predict individual genetic stability (Harrell's C-index = 0.79). S4.

2. Dynamically balance genetic progress (ΔGEBV ≥ 0.3SD / generation) and inbreeding coefficient (ΔF < 0.5%) using the NSGA-Ⅱ algorithm. Update the screening threshold every breeding cycle. -6 ) The priority upgrade mechanism was initiated (weight increased by 20%), and the impact of threshold adjustment on the effective size of the population (Ne>100) was evaluated through Monte Carlo simulation to ensure the genetic diversity baseline.

Images