Geographic tracing method for endangered wild animals and products thereof based on machine learning and DNA imprinting and application thereof

Abstract

The invention discloses a geographical tracing method for endangered wild animals and products thereof based on machine learning and DNA imprinting and application thereof, and relates to the technical field of geographical tracing of wild animals and products thereof. The method comprises the following steps: deducing the geographic genetic structure of a sample with known geographic coordinates by applying a machine learning method, establishing a linear regression model, and tracing the sample with unknown geographic coordinates according to the linear regression model. According to the machine learning and DNA imprinting-based endangered wild animal and product geographical tracing method and the application thereof, a large number of genetic markers obtained by original sequencing are simplified into a group of simplified features which cannot be directly compared through dimension reduction, and each feature contains individual geographical source information in a sample; on the basis, a linear regression model is established through simplified feature training, so that the training cost is greatly reduced; the principal component with the maximum variance is mapped to the geographic coordinates of the individuals, and the effect of tracing the geographic sources of the individuals is achieved through a linear regression model.

Description

Technical Field

[0001] The present invention relates to the technical field of geographical traceability of wild animals and their products, and in particular to a geographical traceability method for endangered wild animals and their products based on machine learning and DNA imprinting and its application. Background Art

[0002] Illegal wildlife trade not only disrupts the international trade market, but is also an important cause of wildlife endangerment and extinction (Schneider, 2012). In addition, wild animals may carry pathogenic bacteria, viruses, and parasites, causing animal diseases. If the diseases carried are zoonotic, they will seriously threaten people's health (White & Razgour, 2020). Therefore, tracing the source of wild animals and their products, especially the geographical source, can not only monitor the source of wild animals and their products, but also reveal the illegal trade routes of wild animals, providing technical support for the traceability and protection of wild animals.

[0003] Previous researchers have used methods such as mitochondrial haplotype network diagrams or genetic geographic pedigrees to trace the origin of wild animals and their products (Ogden & Linacre, 2015). However, this type of tracing is based on obtaining samples from background areas to identify closely related groups of samples. It can only distinguish the ownership of samples, but cannot accurately estimate the geographical coordinates of samples. Its accuracy and precision depend on the background information of known samples. For wild animals and their products with a wide distribution range and a large number of circulation areas, the accuracy and precision of these methods cannot meet the current needs of geographical tracing.

[0004] In recent years, the development of molecular provenance techniques based on machine learning and population genetics has provided more scientific technical support for such studies (Qin et al., 2022a; Qin et al., 2022b; Qin & Gaggiotti, 2022). Using these technologies to clarify the geographic genetic structure of the population at the potential source of the sample will help reveal the source, migration route and endangerment mechanism of the sample.

[0005] Kernel Local Fisher Discriminant Analysis of Principal Components (KLFDAPC) is a nonlinear supervised machine learning method that has the same principle as principal component discriminant analysis (DAPC), but uses Kernel Local Fisher Discriminant Analysis (KLFDA) instead of LDA (Fisher Linear Discriminant Analysis) in the original method.

[0006] DAPC is a hybrid dimension reduction statistical technique that combines LDA and PCA (principal component analysis). It makes up for the deficiency of PCA that only represents the total variance of the sample. This multivariate statistical method divides the variance of the sample into two parts: between-group and within-group, which can distinguish different groups to the greatest extent. However, DAPC is more suitable for linear association analysis, and its statistical performance for nonlinear association is poor.

[0007] KLFDAPC uses KLFDA instead of LDA to help overcome these limitations. By introducing the kernel function, KLFDA can effectively expand the linear discriminant to a nonlinear discriminant. This analysis method that combines linear, nonlinear and principal components can be applied to more complex genetic scenarios, more effectively detect the geographic genetic structure of the population and predict the geographic origin of individuals. Summary of the invention

[0008] To solve the above problems, the present invention provides a geographical tracing method for endangered wild animals and their products based on machine learning and DNA imprinting and its application, develops a geographical tracing model for endangered animals and their products based on machine learning and DNA imprinting, and provides new theories and new and superior technologies for the exploration and monitoring of endangered wild animals and plants.

[0009] To achieve the above object, the present invention provides a method for geographical tracing of endangered wild animals and their products based on machine learning and DNA imprinting, comprising the following steps:

[0010] S1. Use KLFDAPC to trace the samples and establish a linear regression model;

[0011] The process of tracing the source and establishing a linear regression model includes:

[0012] S11, sampling and sequencing to obtain the whole genome sequence of the sample;

[0013] S12, aligning the whole genome sequence obtained by sequencing in step S1 to the reference genome and detecting mutations, merging the mutation detection files of each sample and setting parameter filtering to obtain high-quality mutation sites;

[0014] S13, using 0, 1, 2 to encode genotypes, and performing KLFDAPC analysis on the individual × genotype matrix, including:

[0015] S131, using PCA to reduce the dimension of the individual × genotype matrix to obtain the principal components PCs containing the geographical location information in the genotype;

[0016] S132, KLFDA analysis was performed on the first 20 to 50 standardized PCs to further extract nonlinear features in the principal components;

[0017] S133, using Procrustes analysis to calculate the correlation between the first two principal components KLFDAPC1 and KLFDAPC2 and the geographic coordinates of the sampling points, establishing a transformation between the principal components and the geographic coordinates through translation, rotation and scaling, and establishing a linear regression model between the transformed principal components and longitude and latitude;

[0018] S2. Apply linear regression model to trace the samples with unknown geographic coordinates.

[0019] Preferably, during the sequencing process of step S11, tissues that are difficult to grind and have low nucleic acid content are removed from the sample, and after the tissues are removed, they are quickly frozen in liquid nitrogen, stored at -80°C, and transported on dry ice, and repeated freezing and thawing should be avoided during sample delivery.

[0020] Preferably, in the filtering process of step S12, filtering is performed with minimum allele frequency MAF>0.05 and deletion rate ≤0.2.

[0021] Preferably, in step S13, 012 respectively represent homozygous, heterozygous and homozygous minor alleles of the locus.

[0022] Preferably, step S2 includes the following steps: inputting samples KLFDAPC1 and KLFDAPC2 of unknown geographic coordinates into a linear regression model, decoding the output of the linear regression model as the geographic coordinates corresponding to KLFDAPC1 and KLFDAPC2, and mapping the geographic coordinates onto a map for tracing.

[0023] The present invention provides a method for tracing the geographical origin of endangered wild animals and their products based on machine learning and DNA imprinting and its application, which has the following beneficial effects:

[0024] (1) Through dimensionality reduction, the large number of genetic markers obtained from the original sequencing are simplified into a set of genetic features that cannot be directly compared. At the same time, each feature contains the geographical origin information of the individuals in the sample;

[0025] (2) Establishing a linear regression model through genetic feature training greatly reduces the training cost;

[0026] (3) The genetic characteristics with the largest variance are mapped to the geographic coordinates of the individual, and the effect of tracing the individual's geographic origin is achieved through a linear regression model.

[0027] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A in the middle is the distinction of population structure by the first 16 principal components of PCA, B is the first 4 principal component graphs, C is the gradient change of the first two genetic characteristics along longitude, and D is the gradient change of the first two genetic characteristics along latitude;

[0029] Figure 2 is the correlation between the reduced-dimensional genetic characteristics and geographic coordinates; A is the relationship between KLFDAPC1 and longitude on the projected map and its gradient distribution on the geographic coordinates, and B is the relationship between KLFDAPC2 and latitude on the projected map and its gradient distribution on the geographic coordinates;

[0030] Figure 3 is the projection of the individual on the real map, where A is the KLFDAPC tracing result and B is the PCA tracing result;

[0031] Figure 4 Projection of KLFDAPC traceability results on real maps for individuals with unknown geographic locations. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical scheme and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention are further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions.

[0033] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products or devices.

[0034] Embodiment 1

[0035] A method for tracing the geographical origin of endangered wild animals and their products based on machine learning and DNA imprinting is used to infer the geographical origin of musk deer. A total of 274 samples were used for sequencing in this embodiment, of which 59 were from the Tp breeding farm in the southernmost part of Shaanxi, 192 were from the Ql breeding farm in southern Shaanxi and north of the Qinling Mountains, and 23 were from the WS breeding farm in northern Sichuan Province, with the origins being the Tapa Mountains, the middle Qinling Mountains, and the western Sichuan Plateau, respectively.

[0036] These farms are typical representatives of the musk artificial breeding industry. The three farms are independent and do not cross-breed with each other. Musk deer from the same farm have similar breeding conditions. The 274 samples were randomly sampled from adult male individuals aged 2.5-6.5 years old in each farm. Sample collection was carried out under the permission of the forestry department and in accordance with China's National Wildlife Protection Law.

[0037] The method specifically comprises the following steps:

[0038] S1. KLFDAPC was used to trace the origin of 274 musk deer samples and a linear regression model was established.

[0039] S11. First, sequence the musk deer samples in the training set and remove tissues that are difficult to grind and have low nucleic acid content. After the tissues are removed, quickly freeze them in liquid nitrogen, store them at -80℃, and transport them on dry ice. Repeated freezing and thawing should be avoided during the period.

[0040] S12. Align the simplified genome sequence obtained by sequencing to the reference genome and detect SNP single nucleotide polymorphism variation. Merge the variation detection files of each sample and filter them with the minimum allele frequency <0.05 and the missing rate <0.2 to obtain high-quality SNP sites.

[0041] S13, 012 coded genotypes, representing homozygous, heterozygous and homozygous minor alleles of the locus, respectively. KLFDAPC analysis was performed on the individual × genotype matrix to trace the origin.

[0042] S131. PCA is used to reduce the dimension and obtain the principal components PCs containing the geographical location information in the genotype. The PCA results are as follows: Figure 1 As shown in a and b.

[0043] It shows the degree of differentiation of the three musk deer populations on PC1-PC16. Both results show that the first two principal components of PCA can clearly divide the samples into three groups, while the remaining principal components have little effect in distinguishing different populations. This shows that the first two principal components contain most of the individual's geographical location information, retaining useful information to the greatest extent while reducing the dimension, and providing support for the subsequent use of the first two principal components for geographical tracing.

[0044] S132, perform KLFDA analysis on the first 20 to 50 standardized PCs to further extract nonlinear features in the principal components. At this time, the principal components (KLFDAPCs) not only contain decisive geographic location information, but also retain linear and nonlinear associations between samples.

[0045] S133. Use Procrustes analysis to calculate the correlation between the first two principal components (KLFDAPC1 and KLFDAPC2) and the geographic coordinates of the sampling points, and compare and calculate the geometric similarity between KLFDAPCs and the actual geographic locations of the samples by translation, rotation and scaling, and draw the first two principal component axes after rotation. The results are shown in Figure 1 As shown in c and d, c is longitude and d is latitude.

[0046] The variance explanation rates (PVE) of the first two principal components after Proctor transformation were 42.6% and 3.84%, respectively, which were higher than the corresponding PCA results, and they also showed longitude and latitude gradients, indicating that the reduced-dimensional genetic features had a better explanation for longitude.

[0047] S134, perform linear regression on KLFDAPC1 and KLFDAPC2 with the sample longitude and latitude, and establish the corresponding relationship between the dimensionality reduction features and the longitude and latitude, so as to realize the mapping of PC on the real map.

[0048] Figure 2 The linear correlation between the KLFDAPCs after the Pu-type transformation and the geographic longitude and latitude coordinates is shown, Y = aX + b, Y is the longitude and latitude coordinates, and X is the KLFDAPC genetic characteristic value. The correlation coefficients (R 2 ) reached 75% and 73% respectively, indicating that the principal components after Pu-type transformation have a high geometric correlation with longitude and latitude and also have a certain linear correlation. At the same time, R 2 It also ensures the accuracy of geographical location of traceability to a certain extent. Considering that the actual traceability is generally carried out at the national or regional scale, R 2 To a certain extent, it can also effectively ensure the application of traceability in actual scenarios.

[0049] Figure 2 It also shows the gradient distribution of the first two principal components in geographic coordinates. Figure 2 A in the figure shows the relationship between KLFDAPC1 and longitude on the projection map. The results show that KLFDAPC1 is positively correlated with the longitude of the three source farms. Figure 2 Figure B shows the relationship between KLFDAPC2 and latitude on the projection map. Here, KLFDAPC2 well illustrates the latitude correlation between Ql and the other two farms, but for WS and Tp farms, there is no good representation, probably because their latitudes are not much different.

[0050] The above results show that the geographic location information contained in the principal components can be used to trace the geographical origin of wild animals. However, both KLFDAPCs and PCs contain this information, so their effects on tracing are compared. At the same time, KLFDAPCs and PCs are transformed and linearly correlated, and the degree of fit of the projections on the map and the Pullman correlation coefficient are compared to illustrate the tracing effect.

[0051] The final individual traceability results are as follows: Figure 3 The results showed that the sample points matched well with the geographic coordinates of individuals from Ql and WS farms, whether directly mapped by PCs or mapped by KLFDAPCs after KLFDA analysis, but for the traceability of Tp farm, the results of KLFDAPCs were more consistent.

[0052] In particular, for the KLFDAPCs results, the geometric similarity between the first two dimensionality reduction features and the geographic coordinates reached 0.8486, which is higher than the 0.7355 of the PCs method. This is also well demonstrated in the figure for the traceability of individuals from the Tp farm.

[0053] S2. Apply the above trained linear regression model for verification.

[0054] The model constructed by the above KLFDAPC results covers the geographical genetic structure of the population sampling site. For unknown samples obtained, after simplified genome sequencing, the sequencing data is input into the KLFDAPC model to predict the geographical genetic characteristic values. The output value of the genetic characteristic decoding linear regression model is the geographical coordinates corresponding to the genetic characteristic, which can be mapped on the map for tracing ( Figure 4 ).

[0055] In this example, five individuals from different sources were selected to obtain their simplified genomes, and the five simplified genomes were input into KLFDAPC to obtain the predicted genetic geographic genetic characteristic values. After predicting these characteristic values ​​using the linear model constructed previously, the geographic coordinates of the individuals were obtained.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for geographical tracing of endangered wild animals and their products based on machine learning and DNA imprinting, characterized in that: The steps include: S1. Use KLFDAPC to deduce the genetic geographic structure of the samples and establish a linear regression model, including: S11, sampling and sequencing to obtain a simplified genome sequence of the sample; S12, aligning the simplified genome sequence obtained by sequencing in step S1 to the reference genome and detecting mutations, merging the mutation detection files of each sample and setting parameter filtering to obtain high-quality mutation sites; S13, using 012 coded genotypes, conduct KLFDAPC analysis and traceability on the individual × genotype matrix, including: S131, using PCA to reduce the dimension of the individual × genotype matrix to obtain the principal components PCs containing the geographical location information in the genotype; S132, KLFDA analysis was performed on the first 20 to 50 standardized PCs to further extract nonlinear features in the principal components; S133, using Procrustes analysis to calculate the correlation between the first two principal components KLFDAPC1 and KLFDAPC2 and the geographic coordinates of the sampling points, establishing a transformation between the principal components and the geographic coordinates through translation, rotation and scaling, and establishing a linear regression model between the transformed principal components and longitude and latitude; S2. Apply linear regression model to trace the samples with unknown geographic coordinates.

2. The method for geographical tracing of endangered wild animals and their products based on machine learning and DNA imprinting according to claim 1, characterized in that: During the sequencing process of step S11, tissues that are difficult to grind and have low nucleic acid content are removed from the sample. After the tissues are removed, they are quickly frozen in liquid nitrogen, stored at -80°C, and transported on dry ice. Repeated freezing and thawing should be avoided during sample delivery.

3. The method for geographical tracing of endangered wild animals and their products based on machine learning and DNA imprinting according to claim 1, characterized in that: In the filtering process of step S12, filtering is performed with minimum allele frequency MAF>0.05 and missing rate ≤0.

2.

4. The method for geographical tracing of endangered wild animals and their products based on machine learning and DNA imprinting according to claim 1, characterized in that: In step S13, 012 respectively represent homozygous, heterozygous and homozygous minor alleles of the locus.

5. The method for geographical tracing of endangered wild animals and their products based on machine learning and DNA imprinting according to claim 1, characterized in that: Step S2: The process of tracing the source of samples with unknown geographical coordinates includes: The samples KLFDAPC1 and KLFDAPC2 with unknown geographic coordinates are input into the linear regression model. The output of the decoded linear regression model is the geographic coordinates corresponding to KLFDAPC1 and KLFDAPC2. The geographic coordinates can be mapped onto the map for traceability.

6. An application of the method for geographical tracing of endangered wild animals and their products based on machine learning and DNA imprinting as claimed in any one of claims 1 to 5, characterized in that: Apply KLFDAPC to the geographical traceability of endangered wild animals and their products.

Images