Method for screening plant root growth promoting microbial inoculants

Abstract

The application relates to the technical field of microbial agent, and discloses a plant root growth-promoting microbial agent screening method, which comprises the following steps: firstly, collecting rhizosphere microbial group data and corresponding crop growth phenotype data under different target crops, growth stages and soil environments; then, establishing a microbe-phenotype prediction model based on the data, obtaining two groups of key growth-promoting strains through characteristic importance analysis and microbial interaction network analysis; and finally, screening a microbial agent formula with the optimal expected growth-promoting effect according to the key strains and the prediction model. The application establishes a systematic plant root growth-promoting microbial agent screening scheme, can screen out efficient and synergistic dominant strain combinations by comprehensively considering the growth-promoting effects and interaction relationships of the strains, significantly improves the pertinence and effectiveness of the plant root growth-promoting microbial agent, and has a good application prospect in the field of microbial fertilizer development.

Description

Technical Field

[0001] This invention relates to the field of microbial inoculants, and more specifically, to a method for screening microbial inoculants that promote plant root growth. Background Technology

[0002] Plant roots are vital organs for plant growth and development, and their growth status directly affects crop nutrient absorption and stress resistance. Therefore, effectively regulating root growth to improve crop stress resistance and increase agricultural production efficiency has always been a research hotspot in agriculture. In recent years, with the rapid development of microbiome and bioinformatics, the preparation of microbial inoculants using rhizosphere growth-promoting microorganisms has become a promising technological approach.

[0003] Chinese patent CN110029077B discloses a salt-tolerant growth-promoting bacterial strain Y4 and its application. This technique screened out a Bacillus strain Y4, which significantly improved the salt tolerance and growth of tomato seedlings under salt stress and had a significant promoting effect on root growth. However, this technique only screened one growth-promoting bacterium, limiting its application scope and effectiveness, and it neglected the influence of inter-strain interactions on the efficacy of the inoculant. Another Chinese patent CN111793680B discloses a method for screening growth-promoting strains based on high-throughput sequencing and its application. This method uses high-throughput sequencing technology to analyze the composition of soil microbial communities and performs regression analysis based on changes in the abundance of dominant strains and plant biomass to screen potential growth-promoting strains. However, this method does not consider the direct correlation between microorganisms and plant phenotypes, nor does it integrate the interactions between growth-promoting strains, and it remains insufficient in optimizing inoculant formulations.

[0004] In summary, existing technologies have failed to establish quantitative predictive models for microbial diversity and plant phenotype in the screening of plant root growth-promoting microbial agents. The screening basis is insufficient, and the ecological niche and interaction relationships of growth-promoting strains in the microbial community are not fully considered, which affects the stability of the agents. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of existing technologies, this invention provides a method for screening plant root growth-promoting microbial agents. This method first collects rhizosphere microbiome data and corresponding crop growth phenotypic data from different target crops, growth stages, and soil environments. Then, based on these data, a microbial-phenotypic prediction model is established. Through feature importance analysis and microbial interaction network analysis, two groups of key growth-promoting strains are obtained. Finally, based on the key strains and the prediction model, the agent formulation with the optimal expected growth-promoting effect is screened. The screening system established by this invention is expected to significantly improve the targeting and effectiveness of plant root growth-promoting microbial agents.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Methods for screening plant root-promoting microbial agents include:

[0008] Rhizosphere microbiome data were collected from different target crops, at different growth stages, and under different soil conditions; growth phenotypic data of the target crops corresponding to the rhizosphere microbiome data were also collected; the rhizosphere microbiome data included species composition and relative abundance data.

[0009] Based on rhizosphere microbiome data and growth phenotype data, a microbiome-phenotype prediction model was established. Based on this model, feature importance analysis was performed to obtain the first group of key growth-promoting strains. A microbial interaction network was constructed, and based on this network, the centrality index of microbial species was calculated. According to the centrality index of microbial species, the second group of key growth-promoting strains was determined. Based on the first group of key growth-promoting strains, the second group of key growth-promoting strains, and the established microbiome-phenotype prediction model, N3 strain combinations with the optimal expected growth-promoting effect were selected and marked as the optimal inoculum formulation.

[0010] Furthermore, the feature importance analysis based on the microbial-phenotype prediction model includes:

[0011] Based on the microbiome-phenotype prediction model, the Shapley value and permutation importance score of each species in the microbiome were calculated.

[0012] The Shapley value and permutation importance score were standardized and then weighted and averaged to obtain the overall importance score for each species.

[0013] Based on the overall importance score, the top N1 microbial species with the highest scores are selected as the first group of key growth-promoting strains, where N1 is a positive integer.

[0014] Furthermore, the construction of the microbial interaction network includes:

[0015] Species composition and relative abundance data from rhizosphere microbiome data were used to calculate correlation coefficients between species.

[0016] A primary microbial interaction network is constructed based on the correlation coefficients between species; in the primary microbial interaction network, nodes represent microbial species, and edges represent the correlations between species.

[0017] Furthermore, the construction of the microbial interaction network also includes:

[0018] In the primary microbial interaction network, edges with correlation coefficients between species greater than the correlation threshold are marked as strongly correlated edges, and edges with correlation coefficients less than or equal to the correlation threshold are marked as weakly correlated edges. The weakly correlated edges are removed to obtain the final microbial interaction network.

[0019] Furthermore, the identification of the second group of key growth-promoting strains includes:

[0020] Calculate the overall centrality score based on the centrality index of microbial species;

[0021] Based on the comprehensive centrality score of the species, the top N2 species were selected from high to low and identified as the second group of key growth-promoting strains.

[0022] Furthermore, the process of selecting the N3 strain combinations with the best expected growth-promoting effect includes: forming candidate bacterial agent formulations based on the first group of key growth-promoting strains and the second group of key growth-promoting strains; and screening the candidate bacterial agent formulations based on a microbial-phenotypic prediction model.

[0023] Further, the formulation for forming the candidate microbial agent includes:

[0024] The first group of key growth-promoting strains and the second group of key growth-promoting strains were combined to generate the third group of key growth-promoting strains.

[0025] The third group of key growth-promoting strains were combined and formulated in different proportions to form N4 candidate bacterial agent formulations.

[0026] Furthermore, the screening of candidate microbial agent formulations based on the microbial-phenotypic prediction model includes:

[0027] Input the candidate microbial agent formulation into the microbial-phenotype prediction model to obtain the growth phenotype prediction data of the candidate microbial agent formulation;

[0028] Based on the growth phenotype prediction data of the candidate bacterial agent formulations, the N3 strain combinations with the best expected growth-promoting effect were screened out.

[0029] Furthermore, the screening of the N3 strain combinations with the optimal expected growth-promoting effect based on the growth phenotype prediction data of the candidate bacterial agent formulations includes:

[0030] The growth phenotype prediction data of the candidate bacterial agent formulations were standardized to obtain standardized growth phenotype prediction data.

[0031] The standardized growth phenotype prediction data were weighted and summed to obtain the comprehensive score of each candidate inoculant formulation;

[0032] Based on the comprehensive scores of the candidate bacterial agent formulations, the N3 strain combinations with the highest scores were selected.

[0033] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0034] This invention constructs a large-scale rhizosphere microbiome dataset covering major crops, growth stages, and environmental factors. This dataset systematically records the composition and diversity characteristics of the rhizosphere microbiome, providing rich raw data for analyzing the association between rhizosphere microorganisms and crop growth. Using this dataset, microbial-plant interactions can be compared and analyzed at the community level, identifying key growth-promoting microbial groups and providing theoretical guidance for subsequent screening of dominant strains. The acquisition of this dataset overcomes the shortcomings of previous studies, such as small sample sizes and poor representativeness, and is of great significance for expanding our understanding of rhizosphere microbial ecology.

[0035] This invention innovatively integrates microbial diversity information and phenotypic data to construct a microbial-phenotypic prediction model. Traditional microbial agent screening mainly relies on experience and randomness, evaluating agent effects through extensive field trials, resulting in long screening cycles and high costs. This invention utilizes machine learning algorithms, using microbial diversity indicators as independent variables and plant growth indicators as dependent variables, to train a prediction model with high accuracy and good generalization performance. This model can directly predict the growth-promoting effects of microbial combinations, significantly shortening the screening cycle and reducing field trial costs. This strategy pioneers a new approach to targeted design of microbial agents, achieving precise screening based on big data and artificial intelligence, representing a significant technological advancement in this field.

[0036] This invention comprehensively considers the ecological niche and interactions of bacterial strains, improving the effectiveness and stability of screening inoculants. Previous work primarily screened strains based on their growth-promoting effects, neglecting the complexity of soil microbial communities. This invention, in addition to examining the growth-promoting effects of individual strains, also analyzes the key roles and interaction patterns of strains within the community. Characteristic importance analysis identifies key strains that significantly contribute to plant growth; network analysis reveals the interactions between these key strains. Integrating these two types of analysis results prioritizes combinations of strains with strong growth-promoting effects, key ecological niches, and synergistic interactions, resulting in more stable and efficient inoculant formulations in complex environments. This strategy effectively solves problems such as low colonization rates and diminishing effects, enhancing the practical application value of inoculants.

[0037] This invention establishes a systematic screening scheme for rhizosphere growth-promoting microbial agents. This scheme covers key aspects such as substrate preparation, data acquisition, bioinformatics analysis, modeling and prediction, and formulation optimization, forming a standardized technical process that provides a basis for the quality assessment of rhizosphere growth-promoting microbial agents. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating the principle of the plant root growth-promoting microbial inoculant screening method in this invention.

[0040] Figure 2 This is a flowchart of the method for obtaining the first group of key growth-promoting strains in the plant root growth-promoting microbial inoculant screening method of the present invention.

[0041] Figure 3 This is a flowchart of the method for constructing a microbial interaction network in the plant root growth-promoting microbial inoculant screening method of the present invention;

[0042] Figure 4 This is a flowchart of the method for determining the second group of key growth-promoting strains in the plant root growth-promoting microbial inoculant screening method of the present invention.

[0043] Figure 5 This is a flowchart illustrating the method for screening optimal microbial combinations based on growth phenotype prediction data of candidate microbial agent formulations in the plant root growth-promoting microbial agent screening method of the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Example 1

[0046] Please see Figure 1 As shown, this embodiment provides a method for screening plant root-promoting microbial agents, including:

[0047] Step S1000: Collect rhizosphere microbiome data for different target crops, different growth stages, and different soil environments; collect growth phenotype data of the target crops corresponding to the rhizosphere microbiome data;

[0048] Further, step S1000 includes:

[0049] Step S1100: Collect rhizosphere microbiome data for different target crops, different growth stages, and different soil environments; the rhizosphere microbiome data includes species composition, relative abundance data, and Alpha diversity index.

[0050] Specifically, target crops are selected, such as major food crops like wheat, corn, and rice, as well as vegetable crops like tomatoes and peppers, and fruit trees like apples and peaches. Because the rhizosphere microbiota of different crops varies significantly, sampling and analysis of multiple crops are necessary. By selecting major food crops like wheat, corn, and rice, representative vegetables like tomatoes and peppers, and representative fruit trees like apples and peaches, most crop types can be covered. The sampling period is determined according to the crop's growth process, collecting rhizosphere soil samples at different growth stages, such as seedling, jointing, heading, and grain-filling stages. The composition and quantity of root exudates change dynamically at different growth stages, thus affecting the rhizosphere microbial community. Therefore, setting the sampling period according to the crop's growth process allows for tracking the dynamic succession of the microbial community. When collecting samples, select several representative farmlands, taking into account factors such as soil type (sand, clay, loam, etc.), fertility status (high, medium, low fertility), and irrigation conditions (irrigated land, dry land). Randomly select 5-10 crops from each farmland, collect rhizosphere soil, and mix multiple samples to form a single sample. By selecting farmlands with different soil types, fertility status, and irrigation conditions, the shaping effect of the soil environment on the microbial community can be compared and analyzed.

[0051] Soil samples were pretreated by sieving to remove plant debris and stones. DNA from soil microorganisms was extracted using commercially available kits. High-throughput sequencing was then used to amplify and sequence the 16S rRNA (bacteria) and ITS (fungi) fragments of the microorganisms using PCR, obtaining raw sequencing data for each sample. Scientific sampling methods and sample pretreatment are fundamental to ensuring the accuracy and reliability of subsequent analysis data. Multi-point pooled sampling improves sample representativeness, and the use of commercial kits maximizes the efficiency and purity of DNA extraction. High-throughput sequencing technologies, such as the Illumina Hiseq platform, enable massively parallel sequencing of microbial 16S rRNA and ITS fragments, generating millions of reads to reveal the diversity and compositional characteristics of microorganisms at the community level. Bioinformatics analysis was performed on the raw sequencing data to remove low-quality and chimeric sequences. OTU clustering and species annotation were performed on the effective sequences to obtain information such as species composition, relative abundance, and alpha diversity indices (Chao1, Shannon) of bacterial and fungal communities.

[0052] Relative abundance reflects the proportion of each microbial taxa in a community. The number of sequences from different microbial taxa in a sample can be obtained through 16S rRNA and ITS sequencing. Dividing the number of sequences from a particular taxa by the total number of sequences in the sample yields the relative abundance of that taxa, usually expressed as a percentage. For example, a relative abundance of 42.5% for Proteobacteria indicates that Proteobacteria sequences account for 42.5% of the total bacterial sequences in the bacterial community of that sample, making it the most abundant taxa. Relative abundance data can reflect the compositional structure of a community, revealing dominant and rare bacterial groups. By comparing the relative abundance of different samples, differences and dynamic changes in community composition can be analyzed. Alpha diversity indices are used to assess the level of species diversity within a single sample; common indices include the Chao1 index and the Shannon index. The Chao1 index is an index that estimates community species richness based on the abundance of rare species. Its calculation formula considers the number of OTUs that appear only once and twice in the sample, thus estimating the number of species that may exist in the community but have not been observed. A higher Chao1 index indicates a higher species richness in the community. The Shannon index comprehensively considers both species richness and evenness. Its calculation formula considers not only the number of species but also the relative abundance of each species. A higher Shannon index indicates a higher level of species diversity and a more even distribution of species. The Alpha diversity index can quantitatively compare the diversity differences of microbial communities under different samples or conditions. Higher index values ​​usually indicate a richer and more balanced community structure. By tracking changes in the Alpha diversity index, we can monitor the dynamics of microbial communities and study the impact of biotic and abiotic factors on soil microbial diversity. Relative abundance and the Alpha diversity index reflect the composition and diversity characteristics of microbial communities from different perspectives and are two key indicators commonly used in microbiome research. Combining these two indicators allows for a comprehensive and systematic analysis and comparison of rhizosphere microbial communities.

[0053] Step S1100 fully considers key factors such as crop type, growth stage, and environmental conditions, and comprehensively acquires rhizosphere microbial community structure data through scientific sampling and advanced sequencing analysis. Using this dataset, the compositional differences of microbial communities under different conditions can be systematically compared, and key microbial taxa related to crop growth, resistance, and other phenotypes can be preliminarily identified. This lays a solid foundation for subsequent construction of microbial resource libraries and association models. Large-scale community data can also be used to guide the isolation and enrichment of new growth-promoting strains, accelerating the development of inoculants. Furthermore, the rhizosphere microbiome dataset established in step S1100 integrates multi-dimensional biological and ecological information, serving as crucial foundational data in the field of rhizosphere microbial research and providing rich data support for future in-depth studies of rhizosphere microbial ecological mechanisms.

[0054] For example, consider the collection and analysis of wheat rhizosphere microbiome data:

[0055] 1. Selection of target crop and growth stage

[0056] Target crop: Wheat was selected as the main research subject.

[0057] Growth stages: Rhizosphere soil samples were collected at four key growth stages of wheat, including:

[0058] Seedling stage: The early growth stage of plants, when the root system begins to develop.

[0059] Jointing stage: Wheat stems begin to elongate, and nutrient absorption increases.

[0060] Heading stage: The wheat ear begins to emerge, and the plant enters the reproductive stage.

[0061] Grain filling stage: During the grain filling stage, root activity decreases, but the rhizosphere microbial community remains active.

[0062] 2. Selection of Soil Environment

[0063] Soil types: Three different soil types were selected for comparative analysis, including sandy soil, clay soil, and loam soil;

[0064] Fertility status: Soil fertility level was considered during sampling: high fertility farmland, medium fertility farmland, and low fertility farmland;

[0065] Irrigation conditions: Wheat rhizosphere soil samples were collected under different irrigation conditions, including irrigated land and dry land.

[0066] 3. Sample collection method

[0067] For each different soil type, fertility and irrigation condition, 5-10 wheat plants were randomly selected and rhizosphere soil samples were collected.

[0068] Multiple collected samples are combined into a representative sample.

[0069] 4. Sample pretreatment

[0070] The collected wheat rhizosphere soil samples were pretreated by first sieving to remove plant debris and stones.

[0071] Commercially available kits were used to extract microbial DNA from the soil to ensure extraction efficiency and purity.

[0072] 5. High-throughput sequencing

[0073] High-throughput sequencing of 16S rRNA (bacterial) and ITS (fungal) fragments was performed using the Illumina Hiseq platform;

[0074] Millions of reads were generated, covering wheat rhizosphere bacteria and fungal communities.

[0075] 6. Bioinformatics Analysis

[0076] The raw sequencing data is processed to remove low-quality sequences and chimeric sequences;

[0077] Effective sequences are clustered into OTUs (operational taxonomic units) and species are annotated to obtain the species composition and relative abundance of the community.

[0078] Calculate the Alpha diversity index for each sample, such as the Chao1 index (to assess species richness) and the Shannon index (to assess species richness and evenness).

[0079] 7. Results

[0080] Species composition:

[0081] Proteobacteria: Penicillium, Listeria, Bacillus, etc.

[0082] Phylum Actinobacteriae: Streptomyces, Nocardia, etc.;

[0083] Bacteroidetes: Flavobacterium, chitinophiles, etc.;

[0084] Acidobacteria: Gp6, Gp16, etc.;

[0085] Relative abundance (mean ± standard deviation, %):

[0086] Proteobacteria: 42.5±5.2;

[0087] Actinobacteria: 28.3±3.6;

[0088] Bacteroidetes: 15.8±2.4;

[0089] Acidobacteria: 7.2±1.5;

[0090] Other doors: 6.2±1.8;

[0091] Alpha diversity index (mean ± standard deviation):

[0092] Chao1 index: 1526±218;

[0093] Shannon index: 5.85±0.42.

[0094] 8. Analysis

[0095] In this batch of wheat rhizosphere bacterial communities, Proteobacteria and Actinobacteria were the dominant groups, accounting for 42.5% and 28.3% of the total bacterial abundance, respectively. At the genus level, common plant growth-promoting bacteria such as Penicillium and Streptomyces were detected. Bacteroidetes and Acidobacteria had relatively low abundance, accounting for 15.8% and 7.2%, respectively. The Chao1 index reflects the species richness of the community; this batch of samples had high species richness, averaging 1526 OTUs. The Shannon index, which considers both species richness and evenness, had a value of around 5.85, indicating a high level of community diversity. By calculating the mean and standard deviation of multiple batches of samples, the stability and variability of the community composition can be assessed.

[0096] 9. Conclusion

[0097] Using the above experimental methods, we successfully obtained rhizosphere microbiome data for wheat at different growth stages and under different soil conditions. By comparing and analyzing the effects of different environments and conditions on the rhizosphere microbiome, we can provide a scientific basis for optimizing crop growth and soil management.

[0098] Step S1200: Collect growth phenotypic data of the target crop corresponding to the rhizosphere microbiome data. The growth phenotypic data includes crop growth index data and resistance index data.

[0099] Specifically, crop growth indicators were measured, including plant height, leaf area, aboveground and belowground biomass, root-shoot ratio, chlorophyll content, and photosynthetic rate. Plant height was measured using a ruler, leaf area using a leaf area meter, biomass using the oven-drying and weighing method, chlorophyll content using a SPAD-502 chlorophyll meter, and photosynthetic rate using a Li-6400 portable photosynthesis meter. Growth indicators such as plant height, leaf area, and biomass directly reflect the growth and development of crops and are commonly used indicators for evaluating the growth-promoting effects of microorganisms. Leaf area indicates the crop's light-trapping capacity, biomass accumulation reflects the conversion efficiency of photosynthetic products, and the root-shoot ratio reflects the distribution pattern of carbohydrates between aboveground and belowground organs. These indicators can systematically assess the promoting effect of microorganisms on crop growth. Chlorophyll content and photosynthetic rate are closely related to crop photosynthesis and are also sensitive physiological indicators. By measuring multiple indicators of crop growth phenotypes, the plant-promoting effects of microorganisms can be quantitatively analyzed from multiple levels, including morphology and physiology.

[0100] The determination of crop resistance indicators includes disease resistance, insect resistance, drought resistance, salt tolerance, and cold resistance. Disease resistance is assessed by inoculating pathogens (such as Fusarium head blight and powdery mildew) and investigating disease incidence and disease index. Insect resistance is assessed by artificial inoculation with insects (such as beet armyworm and aphids) and statistically analyzing insect population density and damage severity. Drought resistance is assessed by controlling water usage and measuring physiological indicators such as relative leaf water content and malondialdehyde (MDA) content. Salt tolerance and cold resistance are assessed by irrigation with gradient NaCl solutions and low-temperature stress treatments, respectively, investigating phenotypic changes such as plant survival rate and leaf damage. The determination of resistance indicators is an important means of evaluating the enhancement of crop stress resistance by microorganisms. Inoculation with pathogens and pests allows for direct examination of the impact of microorganisms on crop disease and insect resistance; disease incidence and damage severity are intuitive indicators for evaluating plant protection effectiveness. Water control and salt stress treatments can be used to detect the ability of microorganisms to improve crop drought and salt tolerance. Relative water content reflects the water retention capacity of leaves, while malondialdehyde (MDA) content is closely related to the degree of membrane lipid peroxidation damage and is an ideal physiological indicator for identifying drought-resistant microorganisms. In addition, low-temperature stress treatment can screen for microbial strains with cold-resistant and growth-promoting functions.

[0101] For example, continuing with the wheat rhizosphere microbiome, we measured wheat growth index data and resistance index data:

[0102] 1. Measuring wheat growth index data

[0103] During the critical growth stages of wheat, such as the seedling stage, jointing stage, heading stage, and grain-filling stage, the corresponding growth indicators are measured simultaneously.

[0104] (1) Plant height: Ten wheat plants were randomly selected from each treatment group at the seedling stage, jointing stage, heading stage and grain filling stage, and the height was measured from the ground to the top of the plant with a ruler. The average value was calculated.

[0105] Example data:

[0106] Seedling stage: 12.5±1.2cm;

[0107] Joint elongation stage: 45.8±3.6cm;

[0108] Heading stage: 78.2±4.1cm;

[0109] Grouting period: 85.6±2.9cm.

[0110] (2) Leaf area: During the heading stage, representative leaves (second leaf from the top) of 5 wheat plants in each treatment group were selected and measured using a leaf area meter (such as Li-3000C), and the average value was calculated.

[0111] Example data: Leaf area index (LAI) = 4.52 ± 0.38.

[0112] (3) Biomass: At harvest, five wheat plants were randomly selected from each treatment group and divided into aboveground parts (stems, leaves, ears) and underground parts (roots). After blanching at 105℃ for 15 minutes, the samples were dried in an oven at 80℃ to constant weight, the biomass was measured, and the root-to-shoot ratio was calculated.

[0113] Example data:

[0114] Aboveground biomass: 12.6 ± 0.8 g / plant;

[0115] Underground biomass: 2.1 ± 0.3 g / plant;

[0116] Root-to-crown ratio R / S: 0.165±0.012.

[0117] (4) Chlorophyll content: During the jointing and heading stages, on sunny days from 9:00 to 11:00, the relative chlorophyll content of the second leaf from the top was measured using a SPAD-502 chlorophyll meter. Five parts of each leaf were measured, and 10 replicates were measured for each treatment group.

[0118] Example data:

[0119] During the jointing stage: SPAD value 45.2±2.6;

[0120] Heading stage: SPAD value 52.8±3.2.

[0121] 2. Determination of wheat resistance index data

[0122] (1) Disease resistance: Taking wheat scab as an example. During the wheat heading stage, the wheat ears were sprayed with a suspension of Fusarium spores (concentration of 10⁵ spores / mL), with 5 mL sprayed per ear. After inoculation, the ears were sealed with plastic bags to keep them moist for 48 hours. Two weeks later, the disease incidence was investigated, and the disease probability and disease index were calculated.

[0123] Example data:

[0124] Disease incidence rate in panicles: 15.2% ± 2.8%;

[0125] Disease index: 8.5±1.2.

[0126] (2) Insect resistance: Taking aphids as an example. During the jointing stage of wheat, aphid nymphs were inoculated, with 10 nymphs per plant. Two weeks later, the degree of aphid damage was investigated, and the number of aphids was counted.

[0127] Example data:

[0128] Aphid density: 56.3 ± 8.2 aphids / plant;

[0129] Severity of hazard: Level 3, mild hazard.

[0130] (3) Cold resistance: Cold resistance was tested when the average daily temperature was ≤-10℃ for 5 consecutive days during the wheat overwintering period. Fifty wheat plants were randomly selected, and the frost damage symptoms were observed. The number of surviving plants was counted after 30 days, and the survival rate was calculated.

[0131] Example data:

[0132] Symptoms of frost damage: Leaves turn brown, dry out and curl;

[0133] Survival rate: 65.8% ± 4.6%.

[0134] Step S2000: Based on rhizosphere microbiome data and growth phenotype data, establish a microbiome-phenotype prediction model; based on the microbiome-phenotype prediction model, perform feature importance analysis to obtain the first group of key growth-promoting strains; construct a microbiome interaction network, and calculate the centrality index of microbial species based on the microbiome interaction network; determine the second group of key growth-promoting strains based on the centrality index of microbial species; based on the first group of key growth-promoting strains, the second group of key growth-promoting strains, and the constructed microbiome-phenotype prediction model, screen out N3 strain combinations with the best expected growth-promoting effect, and mark them as the optimal strain combinations.

[0135] Further, step S2000 includes:

[0136] Step S2100: Based on rhizosphere microbiome data and growth phenotype data, establish a microbiome-phenotype prediction model;

[0137] Specifically, based on the rhizosphere microbiome data and growth phenotype data collected in step S1000, a training set and a test set for a neural network-based microbiome-phenotype prediction model are constructed. Microbiome data such as species composition and relative abundance are used as input features of the model, while growth phenotype data such as plant height, leaf area, and disease resistance are used as output targets, establishing a nonlinear mapping relationship between the microbiome and phenotype. 80% of the data is used as the training set, and 20% as the test set. The training set is used for model training and parameter optimization, while the test set is used to evaluate the model's generalization performance. To improve the model's robustness, a k-fold cross-validation method is used. The training set is randomly divided into k mutually exclusive subsets. Each time, k-1 subsets are selected for training, and the remaining subset is used for validation. This process is repeated k times, and the average value is taken. Finally, the model with the best generalization performance is selected.

[0138] A neural network model suitable for microbiome-phenotype association analysis was designed and constructed. A densely connected multilayer perceptron (MLP) was selected as the basic network structure. Multiple fully connected layers were stacked to progressively extract high-level features from the microbiome data, and nonlinear activation functions such as Sigmoid and ReLU were used to enhance the network's expressive power. The number of nodes in the input layer equaled the feature dimension of the microbiome, and the number of nodes in the output layer equaled the number of plant phenotypic indicators. The number of hidden layers and the number of nodes in each layer were appropriately set to balance the model's fitting ability and generalization performance. To prevent overfitting, Dropout regularization was added between hidden layers to randomly disable some neurons and improve the model's generalization ability. Mean squared error (MSE) was selected as the loss function, and the Adam optimization algorithm was used to iteratively update the network weights. The hyperparameters of the model were optimized using methods such as grid search. Using the trained microbiome-phenotype prediction model, plant phenotypes were predicted based on microbiome data, and the influence of different microbial groups on the phenotype was evaluated. Microbiome data from the test set was input into the trained model to predict the corresponding plant phenotypic indicators, and the results were compared with the true values ​​to calculate the coefficient of determination (R²). 2 Evaluation metrics such as root mean square error (RMSE) are used to quantitatively assess the predictive performance of the model.

[0139] The microbial-phenotype prediction model constructed in this step has broad application prospects in revealing the interaction mechanisms between rhizosphere microorganisms and host crops, and guiding agricultural production practices. For example, this model can be used to predict the impact of soil microbiome changes on crop yield and quality, and guide the optimization of farmland management and control measures; based on the microbial-phenotype prediction model, highly efficient plant growth-promoting agents can be screened and designed to reduce the use of chemical fertilizers and pesticides, and achieve green and sustainable agricultural development.

[0140] Step S2200: Based on the microbial-phenotypic prediction model, feature importance analysis is performed to obtain the first group of key growth-promoting strains;

[0141] Furthermore, such as Figure 2 As shown, step S2200 includes:

[0142] Step S2210: Based on the microbial-phenotypic prediction model, calculate the Shapley value and permutation importance score for each species in the microbiome;

[0143] Step S2220: Standardize the Shapley value and the permutation importance score, and then perform a weighted average to obtain the comprehensive importance score for each species;

[0144] Step S2230: Sort the microbial species by comprehensive importance score from high to low, and select the top N1 microbial species with the highest scores as the first group of key growth-promoting strains, where N1 is a positive integer.

[0145] Specifically, machine learning interpretation methods, such as Shapley scores or permutation importance, are applied to calculate the feature importance scores for each species and taxa in the microbiome. Shapley scores, a game-theoretic method, measure the average marginal contribution of each feature to the model's prediction results. It considers all possible feature combinations, fairly and reasonably allocating joint contributions, making it suitable for assessing the relative importance of species to phenotypes. Permutation importance, on the other hand, measures the importance of a feature by randomly permuting it and comparing the decrease in model performance before and after the permutation. These two methods reveal the intrinsic link between microbial taxa and plant phenotypes from different perspectives. Using Python's SHAP library, Shapley scores are calculated for each species in a trained microbiome-phenotype prediction model. Based on the idea of ​​permutation importance, permutation tests are performed on the test set. First, the trained microbiome-phenotype prediction model is used to predict the test set to obtain baseline performance. Then, the abundance of a specific microbial species in the test set is randomly permuted, and predictions are made using the permuted data. This permutation is repeated multiple times, recording the change in model performance after each permutation. Finally, the average performance decrease of the species' feature is calculated as its importance score. Repeat the above process to obtain the permutation importance score for all species.

[0146] Shapley values ​​and substitution importance scores were standardized and assigned weights, and a weighted average was used to obtain the overall importance score for each microbial group. The top N1 groups with the highest overall importance scores were selected as key growth-promoting groups. The determination of the N1 value needs to balance the importance and operability of the taxa; too many taxa are not conducive to subsequent functional validation and application development, while too few may miss important candidate strains. Considering different criteria, the top 30-100 important taxa can be selected. During further screening, the sign of the Shapley value for each group is considered; a positive value indicates that the group has a positive promoting effect on the phenotype, while a negative value suggests a possible antagonistic effect. Groups with larger positive Shapley values ​​are given priority, as they are likely to become new plant growth-promoting strains. Furthermore, by comparing and analyzing the importance ranking of this group in multiple phenotypic prediction models, if certain groups have high importance in multiple indicators such as plant height, leaf area, and disease resistance, it suggests that they may promote host growth through multiple mechanisms and have a broad-spectrum effect, and should be given special attention.

[0147] Step S2300: Construct a microbial interaction network, and calculate the centrality index of microbial species based on the microbial interaction network; the centrality index includes degree centrality, compact centrality, and betweenness centrality; determine the second group of key growth-promoting strains based on the centrality index of microbial species.

[0148] Further, step S2300 includes:

[0149] Step S2310: Construct a microbial interaction network;

[0150] Furthermore, such as Figure 3 As shown, step S2310 includes:

[0151] Step S2311: Calculate the correlation coefficients between species based on the species composition and relative abundance data in the rhizosphere microbiome data;

[0152] The method for calculating the correlation coefficient between species includes:

[0153] Suppose there are two microbial species, A and B, and their relative abundance under different samples or conditions is expressed as x. A ={x A1 ,x A2 ,…,x An} and x B ={x B1 ,x B2 ,…,x Bn}, where n is the number of samples, x Ai x represents the relative abundance of species A in the i-th sample. Bi Let represent the relative abundance of species B in the i-th sample, where 1 ≤ i ≤ n.

[0154]

[0155] in:

[0156] Γ AB : Represents the correlation coefficient between species A and species B;

[0157] μ A : Represents the average relative abundance of species A across all samples;

[0158] μ B : Represents the average relative abundance of species B across all samples;

[0159] α: Adjustment parameter used to control the overall magnitude of the correlation coefficient. It is usually determined through model optimization or experimental data fitting. This parameter is used to ensure that the correlation coefficient is consistent under different sample conditions.

[0160] β AB : Represents the potential symbiotic or competitive relationship coefficient between species A and species B, estimated from historical literature or experimental results (such as interactions between functional groups);

[0161] κ: Sensitivity coefficient controlling for relative abundance differences between species, reflecting the extent to which abundance differences between two species affect their correlation, obtained through experiments or model fitting.

[0162] α and β AB The inclusion of κ allows the formula to not only rely on numerical species abundance but also incorporate functional interactions between species, providing a more accurate measure of correlation. The introduction of κ makes the formula more flexible in adapting to abundance differences among different species, avoiding the sensitivity of purely linear models to extreme data. This formula can capture complex inter-species interactions in microorganisms, especially in the rhizosphere environment, where species may exhibit symbiotic, competitive, and other relationships. By calculating correlation coefficients, it is possible not only to reveal basic correlations between species but also to assess functional interactions, which is of great significance for further constructing microbial interaction networks.

[0163] Step S2312: Construct a primary microbial interaction network based on the correlation coefficients between species; in the primary microbial interaction network, nodes represent microbial species, and edges represent the correlations between species.

[0164] Step S2313: Mark the edges in the primary microbial interaction network where the correlation coefficient between species is greater than the correlation threshold as strongly correlated edges, and mark the edges where the correlation coefficient between species is less than or equal to the correlation threshold as weakly correlated edges; remove the weakly correlated edges to obtain the final microbial interaction network.

[0165] Specifically, the microbial interaction network is a simplified model of the rhizosphere microbial community, revealing the interaction patterns within the community. After obtaining species composition and relative abundance data of the microbial community through 16S rRNA and ITS sequencing, inter-species correlations can be calculated. When constructing a network based on inter-species correlations, positive correlation coefficients correspond to positive ecological relationships, such as mutualism and synergistic metabolism; negative correlation coefficients indicate negative interactions between species, such as resource competition and chemical antagonism. To simplify the network structure, a correlation threshold needs to be set to filter out weakly correlated edges. The selection of the correlation threshold must balance the complexity and information content of the network; too high a threshold will miss important interaction information, while too low a threshold will introduce excessive noise interference. Typically, a significance level (e.g., P < 0.01 or P < 0.001) can be used as a reference, and adjustments can be made based on ecological knowledge and research experience.

[0166] Step S2320: Calculate the centrality index of microbial species based on the microbial interaction network;

[0167] Specifically, based on the constructed microbial interaction network, the centrality index of species nodes can be calculated through topological analysis. Degree centrality reflects a species' connectivity; species that establish connections with more species often play a core role in the community. Tight centrality assesses the importance of a species from the perspective of the entire network, considering the distance of a species to all other species. Species with high tight centrality are located at the center of the network and are crucial to the stability and function of the community. Betweenness centrality focuses on evaluating a species' ability to absorb information in the network flow; species with high betweenness centrality are like the network's "hubs," playing a key role in connecting other species and transmitting interactions. Degree centrality, tight centrality, and betweenness centrality are fundamental concepts in the field of network analysis, and their calculation methods are common knowledge to those skilled in the art; therefore, they will not be elaborated upon in this embodiment.

[0168] Step S2330: Determine the second group of key growth-promoting strains based on the centrality index of microbial species;

[0169] Furthermore, such as Figure 4 As shown, step S2330 includes:

[0170] Step S2331: Calculate the overall centrality score based on the centrality index of the microbial species;

[0171] The calculation of the overall centrality score includes:

[0172]

[0173] in:

[0174] The overall centrality score of species j is used to assess the importance of species j in the microbial interaction network. The higher the score, the more likely the species is to be a key growth-promoting strain.

[0175] The score of species j on the k-th centrality index; 1≤k≤3;

[0176] μ k The mean of the centrality index k represents the average value of all species on this index. It is used for standardization to eliminate dimensional differences between different centrality indices.

[0177] σ k The standard deviation of the centrality index k measures the distribution range of a species on this index and is used for standardization so that the scores of different species on different centrality indices can be compared horizontally.

[0178] ω kThe weights of the centrality metric k reflect the relative importance of different centrality metrics to the overall centrality score. These weights allow for adjustments to the influence of different metrics based on actual needs, reflecting the priority of certain metrics in different scenarios. They are determined based on domain expert experience or experimental data. These weights can be optimized through historical data analysis.

[0179] α k : The nonlinear moderating index of the k-th centrality indicator; adjusting the nonlinear effect of an indicator to help capture nonlinear characteristics under specific conditions; for example, the centrality of certain species may have a greater impact on the overall score within a certain value range; through empirical setting, α is a common choice. k ∈[1,3], based on the analysis of the centrality distribution, decide whether to amplify the score of high-value species;

[0180] λ k The logarithmic moderating factor is used to control the centrality index. By introducing a moderating factor through a logarithmic function, the influence of species with lower scores on the overall score can be softened, while the differences of high-scoring species can be amplified. An appropriate value can be set according to the distribution of the centrality index, with a common range of values ​​being λ. k ∈[0.1,1.0].

[0181] When species j is at a certain centrality index When the score on the statistic increases, the overall score... It will also increase, but due to the introduction of nonlinear adjustment α k and logarithmic adjustment The increase in score will be non-linear and moderate. Weight ω k Increasing the degree centrality score will significantly improve the contribution of this centrality indicator to the overall score. For example, if degree centrality is given a higher weight, a high degree centrality score for species j will significantly increase its contribution to the overall score. To have a greater impact. α k The increase in will enhance the nonlinear effect of species j on this indicator, making the score differences of high-scoring species more obvious.

[0182] This formula introduces a weight ω k The relative importance of different centrality metrics can be dynamically adjusted according to different scenarios or needs. For example, in some cases, betweenness centrality may be more important and therefore can be given a higher weight. The mean μ is used. k and standard deviation σ k Standardization eliminates the differences in the dimensions of different centrality indicators, allowing each indicator to be compared and weighted on a uniform scale. A nonlinear adjustment exponent α is introduced. k It allows for the capture of complex network features and adapts to the extreme performance of some species on certain metrics within the network, which helps in identifying key species. Logarithmic regulation function. This approach minimizes the impact of low-scoring species in the calculation, while amplifying high-scoring species, thus highlighting key species. This formula allows for a more accurate calculation of the overall centrality score for each species j, thereby identifying growth-promoting strains that play a crucial role in the microbial interaction network.

[0183] In step S2332, based on the comprehensive centrality score of the species, the top N2 species are sorted from high to low and identified as the second group of key growth-promoting strains.

[0184] Specifically, different centrality indicators reflect the importance of a species in a community from different perspectives. Therefore, when selecting key species, it is necessary to comprehensively consider various indicators. Standardizing each indicator to ensure consistent dimensions and calculating using formulas yields a comprehensive centrality score that takes multiple factors into account. Ranking the species according to their scores and listing the top N2 species with the highest scores identifies the second group of key growth-promoting strains. These key strains have high node degrees in the microbial interaction network, close connections with other species, and play an important role in information transmission. They are crucial for maintaining the structural and functional stability of the rhizosphere microbial community and are expected to become candidate resources for novel plant growth-promoting agents.

[0185] Microbial interaction network analysis organically combines microbiome science with network science, elucidating complex interactions among microorganisms at the community level and screening out key species. This analytical method breaks through the traditional "one microorganism, one strain" research model, identifying core microorganisms that significantly influence community structure and function from a holistic perspective. By focusing on these key species, dominant strains can be isolated more effectively, accelerating the development of microbial agents. Furthermore, in-depth research on key species helps elucidate the interaction mechanisms and functional positioning of rhizosphere microbial communities, providing new insights for the quantitative and predictive development of microbiome science. The construction and topological analysis of microbial interaction networks also provide new perspectives for studying the dynamic changes and environmental adaptations of rhizosphere microbial communities. By comparing differences in interaction networks under different conditions, the response mechanisms of interspecies interactions to environmental stress can be discovered, predicting the impact of environmental factor changes on community structure and function. In conclusion, microbial interaction network analysis is an important method in rhizosphere microbiome research, providing a new research approach for understanding the relationship between microorganisms, plants, and the environment by integrating microbial diversity information and interaction patterns.

[0186] In step S2400, based on the first group of key growth-promoting strains, the second group of key growth-promoting strains, and the constructed microbial-phenotypic prediction model, the N3 strain combinations with the best expected growth-promoting effect are screened out and marked as the optimal strain combinations.

[0187] Further, step S2400 includes:

[0188] Step S2410: The first group of key growth-promoting strains and the second group of key growth-promoting strains are merged to generate the third group of key growth-promoting strains.

[0189] Step S2420: Combine and formulate the third group of key growth-promoting strains in different proportions to form N4 candidate bacterial agent formulations;

[0190] Step S2430: Input the candidate microbial agent formulation into the constructed microbial-phenotype prediction model to obtain the growth phenotype prediction data of the candidate microbial agent formulation;

[0191] Step S2440: Using the growth phenotype prediction data of the candidate bacterial agent formulations, select the N3 strain combinations with the best expected growth-promoting effect.

[0192] Furthermore, such as Figure 5 As shown, step S2440 includes:

[0193] Step S2441: Standardize the growth phenotype prediction data of the candidate bacterial agent formulation to obtain standardized growth phenotype prediction data.

[0194] Step S2442: Weighted summation of the standardized growth phenotype prediction data to obtain the comprehensive score of each candidate bacterial agent formulation;

[0195] Step S2443: Based on the comprehensive score ranking of the candidate bacterial agent formulations, select the N3 strain combinations with the highest scores.

[0196] Specifically, candidate inoculant formulations were developed based on the first and second groups of key growth-promoting strains. These candidate formulations were then screened using a microbial-phenotypic prediction model. Specifically, the first and second groups of key growth-promoting strains were merged to generate a third group of key growth-promoting strains. The first group of key growth-promoting strains was selected from rhizosphere microbiome data using machine learning algorithms, representing microbial groups highly correlated with crop growth and resistance. The second group of key growth-promoting strains was obtained using network analysis methods, representing microbial groups occupying core nodes and playing key ecological niches in the microbial interaction network. These two groups of strains reflect the importance of rhizosphere microorganisms from different perspectives. Integrating them yields a more comprehensive and representative key microbial resource library, namely the third group of key growth-promoting strains. The third group of key growth-promoting strains were then combined in different proportions to form N4 candidate inoculant formulations. Due to the complex interactions between microorganisms, different strain combinations may produce synergistic or antagonistic effects. Therefore, it is necessary to design multiple candidate inoculant formulations with different ratios to cover various possible combined effects. The design of candidate formulations needs to comprehensively consider factors such as the ecological niche, metabolic function, and interactions of the strains. It should reflect the role of key strains while also taking into account the interactions between strains. For example, different proportions of dominant bacteria can be set (e.g., 30%, 50%, 70%), or combinations can be made based on interaction modules identified through network analysis. The specific value of N4 depends on the research objective and experimental scale; generally, 20–50 candidate formulations can be set to balance screening efficiency and combination diversity.

[0197] Candidate inoculant formulations are input into a pre-constructed microbial-phenotypic prediction model to obtain growth phenotypic prediction data for the candidate formulations. The microbial-phenotypic prediction model is based on machine learning algorithms and is trained by integrating rhizosphere microbiome data and crop phenotypic data. This model can predict crop growth and resistance phenotypes based on microbial composition, reflecting the intrinsic link between microbial communities and plant phenotypes. By inputting the composition information of candidate inoculant formulations into the prediction model, the expected effects of each formulation on crop growth, disease resistance, and stress resistance can be obtained, providing a quantitative basis for selecting the optimal formulation. The use of the prediction model can greatly improve the efficiency of inoculant formulation optimization and reduce the cost of blind screening.

[0198] The growth phenotype prediction data of candidate microbial agent formulations were standardized. Due to significant differences in the dimensions and numerical ranges of different growth phenotype indicators, standardization of the prediction data is necessary for easier comprehensive comparison. Commonly used standardization methods include Min-Max standardization and Z-score standardization. Min-Max standardization eliminates the influence of dimensions by mapping the data to the [0,1] interval; Z-score standardization transforms the data into a standard normal distribution by calculating the standard deviation of the data from the mean. After standardization, the prediction data for different indicators have the same scale, facilitating subsequent weighted calculations and ranking / screening.

[0199] Based on the importance of different growth phenotypic indicators, indicator weights are set, and the standardized prediction data are weighted and summed to obtain a comprehensive score for each candidate microbial agent formulation. Since different growth phenotypic indicators contribute differently to crop growth and agricultural production, weighting coefficients need to be set according to the importance of the indicators. For example, yield indicators are usually assigned higher weights, while certain resistance indicators have relatively lower weights. The setting of indicator weights needs to comprehensively consider factors such as agricultural production needs, crop characteristics, and cultivation environment, and can be determined through expert evaluation, farmer surveys, and literature references. During weighted summation, the standardized prediction data of each indicator are multiplied by the corresponding weighting coefficient, and then summed to obtain the comprehensive score. The comprehensive score reflects the overall performance of the candidate microbial agent formulation on multiple growth phenotypic indicators; the higher the value, the better the expected comprehensive growth-promoting effect.

[0200] Candidate microbial agent formulations were ranked from highest to lowest based on their comprehensive scores. Higher scores indicate better expected performance across multiple growth phenotypic indicators and a better overall growth-promoting effect. Based on research objectives and application needs, the top N3 strain combinations with the highest comprehensive scores were selected as the focus for subsequent product development and field trials. The specific value of N3 can be set according to actual conditions, typically ranging from 1 to 5. The optimal microbial agent formulations selected will be submitted for field validation as candidate products to evaluate their growth-promoting effects and application value in actual agricultural production. The selection of the top N3 strain combinations based on comprehensive scores provides a key direction for subsequent product development and field application, improving R&D efficiency and success rate. This data-driven screening method fully utilizes microbiome data and phenotypic prediction data, reducing the influence of subjective human judgment and enhancing the scientific rigor and objectivity of the screening process.

[0201] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0202] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for screening plant root growth promoting microbial inoculants, characterized by, The method comprises: Collect rhizosphere microbiome data of different target crops, different growth stages, and different soil environments; collect growth phenotype data of the target crops corresponding to the rhizosphere microbiome data; the rhizosphere microbiome data comprises species composition and relative abundance data; Based on the rhizosphere microbiome data and the growth phenotype data, a microbe-phenotype prediction model is established; based on the microbe-phenotype prediction model, feature importance analysis is performed to obtain a first group of key growth-promoting strains; a microbial interaction network is constructed, and based on the microbial interaction network, a centrality index of the microbial species is calculated; and based on the centrality index of the microbial species, a second group of key growth-promoting strains is determined; The first group of key growth-promoting strains and the second group of key growth-promoting strains are combined to generate a third group of key growth-promoting strains; the third group of key growth-promoting strains is combined at different ratios to form N4 candidate microbial agent formulations; The candidate microbial agent formulations are input into the established microbe-phenotype prediction model to obtain growth phenotype prediction data of the candidate microbial agent formulations; the growth phenotype prediction data of the candidate microbial agent formulations is standardized to obtain standardized growth phenotype prediction data; and the standardized growth phenotype prediction data is weighted and summed to obtain a comprehensive score of each candidate microbial agent formulation; the candidate microbial agent formulations are ranked according to the comprehensive score to select the top N3 strain combinations with the highest score, which are marked as optimal microbial agent formulations.

2. The method for screening plant root-promoting microbial agents according to claim 1, characterized in that, The feature importance analysis based on the microbe-phenotype prediction model comprises: Based on the microbe-phenotype prediction model, the Shapley value and the permutation importance score of each species in the microbiome are calculated; The Shapley value and the permutation importance score are standardized and then weighted and averaged to obtain a comprehensive importance score of each species; The top N1 microbial species with the highest score are selected according to the comprehensive importance score from high to low, which are used as the first group of key growth-promoting strains, and N1 is a positive integer.

3. The method for screening plant root-promoting microbial agents according to claim 1, characterized in that, The construction of the microbial interaction network comprises: The species composition and relative abundance data in the rhizosphere microbiome data are used to calculate the correlation coefficient between species; Based on the correlation coefficient between species, a primary microbial interaction network is constructed; the nodes in the primary microbial interaction network represent microbial species, and the edges represent the correlation between species.

4. The method for screening plant root-promoting microbial agents according to claim 3, characterized in that, The construction of the microbial interaction network further comprises: The edges in the primary microbial interaction network with a correlation coefficient greater than a correlation threshold are marked as strong correlation edges, and the edges with a correlation coefficient less than or equal to the correlation threshold are marked as weak correlation edges; the weak correlation edges are removed to obtain the final microbial interaction network.

5. The method for screening plant root-promoting microbial agents according to claim 1, characterized in that, The determination of the second group of key growth-promoting strains comprises: The comprehensive centrality score is calculated according to the centrality index of the microbial species; The top N2 species are selected according to the comprehensive centrality score from high to low to determine the second group of key growth-promoting strains.

Images