Method for screening marker protein or marker gene related to protein production of strain, method for evaluating protein production of strain and related equipment

Abstract

The present invention belongs to the field of biological information processing, and discloses a method for screening marker proteins or marker genes related to the protein yield of a strain, as well as a method and related equipment for evaluating the protein yield of a strain, wherein the method for screening marker proteins related to the protein yield of a strain comprises: obtaining multiple proteins encoded by the genome of each strain in multiple Fusarium strains, and clustering to obtain multiple protein groups. The number of each protein group in each Fusarium strain and the proteins contained therein correspond to the protein yield of the Fusarium strain, to form a data set containing multiple data pairs. Based on the data set, a correlation analysis between the proteome and the protein yield is performed, and a protein group whose correlation coefficient with the protein yield of the Fusarium strain meets a preset threshold is screened as a marker protein. The screening method of the present invention can quickly and accurately obtain marker proteins or marker genes that can effectively characterize the protein yield of the strain.

Description

Technical Field

[0001] The present invention generally relates to the technical field of strain screening, and more specifically, to a method for screening a marker protein or marker gene related to the protein yield of a strain, and a method and related equipment for evaluating the protein yield of a strain. Background Art

[0002] As a potential fungus for producing edible protein, Fusarium has great potential value. However, since the mechanism of Fusarium regulating protein production is still unclear, the only way to screen Fusarium strains with high protein production is through large-scale fermentation experiments, which has the disadvantages of large workload, long experimental cycle and high cost.

[0003] The effects of the proteins encoded in the genome of Fusarium strains on the protein production of the strains mainly include: promoting effect, inhibiting effect, and slight effect. For example, whether the protein encoded by the genome is expressed and the amount of expression are affected by many factors. For example, due to the limitations of culture medium conditions, the protein encoded by the genome may not be expressed or the expression amount is small; when a protein encoded by a certain genome is not expressed, it may have a limited effect on the protein production of the strain, and in some cases the effect is very small; in addition, the number of proteins encoded by the genome is also a relatively important factor affecting the protein production of the strain. In the strain, multiple proteins with different effects and degrees of influence are mixed together, making it difficult to analyze and predict the protein production of the strain based on each protein and its quantity.

[0004] In view of this, there is an urgent need to provide a marker protein related to the protein yield of Fusarium strains and a scheme for evaluating the protein yield of Fusarium strains, so as to quickly and accurately screen Fusarium strains with high protein yield. Summary of the invention

[0005] In order to at least solve one or more of the technical problems mentioned above, according to one aspect of the present invention, a method for screening marker proteins related to the protein production of a strain is provided, comprising: obtaining multiple proteins encoded by the genome of each strain in multiple Fusarium strains, clustering the multiple proteins encoded by the genomes of the multiple Fusarium strains to obtain multiple proteomes; corresponding the number of each proteome and the proteins contained therein in each Fusarium strain to the protein production of the Fusarium strain, to form a data set containing multiple data pairs; based on the data set composed of the data pairs, performing a correlation analysis between the proteome and the protein production, obtaining a correlation coefficient and / or significance, and screening the proteome whose correlation coefficient and / or significance with the protein production of the Fusarium strain meets a preset threshold as the marker protein.

[0006] According to one embodiment of the present invention, clustering multiple proteins encoded by the genomes of multiple Fusarium strains to obtain multiple proteomes includes: performing sequence alignment and protein phylogenetic tree analysis on the proteins encoded by the genomes of each Fusarium strain, dividing them into multiple categories, each category corresponds to a proteome, and obtaining multiple proteomes; and / or clustering the proteins encoded by the genomes of each Fusarium strain according to proteins with similar functions, clustering proteins with similar functions but different structures, or clustering proteins with similar functions but different protein sequences, each category corresponds to a proteome, and obtaining multiple proteomes.

[0007] According to one embodiment of the present invention, each proteome in each Fusarium strain and the number of proteins contained therein correspond to the protein yield of the Fusarium strain to form a data set containing multiple data pairs, wherein the data pairs are in the form of:

[0008] {strain n, protein group 1, protein amount a, protein yield x},

[0009] {strain n, protein group 2, protein quantity b, protein yield x}, ......

[0010] {strain m, proteome 1, protein quantity d, protein yield y},

[0011] {strain m, proteome 2, protein quantity e, protein yield y};

[0012] m and n represent strain numbers; a, b, c, d represent protein quantities; x and y represent protein yields.

[0013] According to one embodiment of the present invention, the correlation analysis between the proteome and the protein yield based on the data set composed of the data pair includes: obtaining multiple subsets of the data set; performing correlation analysis based on each subset respectively; screening the proteome whose correlation coefficient and / or significance with the protein yield of the Fusarium strain is greater than a preset threshold, and obtaining multiple analysis results; taking the intersection of the multiple analysis results to obtain the marker protein;

[0014] Preferably, the correlation coefficient and / or significance between the proteome and the protein yield is obtained by the cor() function;

[0015] Preferably, the correlation coefficient and / or significance between the proteome and the protein yield is calculated using cor(df￥x, df￥y, method='spearman'), wherein df￥x represents a vector set [a, d, ...] of the number of proteins in each strain of the proteome to be analyzed for correlation, df￥y represents a vector set [protein yield 1, protein yield 2, ...] of the protein yield of each strain, and 'spearman' indicates that the spearman correlation analysis is selected in the cor() function;

[0016] Preferably, the correlation coefficient between the proteome of each strain and the protein yield is obtained by taking the significance value.

[0017] According to one embodiment of the present invention, the method for screening marker proteins related to the protein yield of the strain also includes: obtaining positively correlated marker proteins and negatively correlated marker proteins of the Fusarium strain according to the values ​​of the correlation coefficient and / or significance, the positively correlated marker proteins promoting protein yield, and the negatively correlated marker proteins inhibiting protein yield; wherein, the multiple Fusarium strains include high-yield strains, medium-yield strains and low-yield strains according to protein yield, and the strains with a protein percentage (g) in the product measured by the Kjeldahl nitrogen method not higher than 50 are low-yield strains, those higher than 50 and lower than 55 are medium-yield strains, and those higher than 55 are high-yield strains.

[0018] According to another aspect of the present invention, a method for evaluating the protein production of a strain is provided, comprising: obtaining a correlation coefficient between the proteome and the protein production based on the aforementioned method for screening marker proteins related to the protein production of the strain; calculating a score value of the protein production of the strain to be evaluated based on the correlation coefficient, and evaluating the protein production of the strain to be evaluated based on the score value.

[0019] The scoring formula for the protein yield of the strain to be evaluated is:

[0020]

[0021] Wherein, Score is the score value of the strain; n = 1, 2, ..., N, represents the number of the proteome; N represents the total number of proteomes in the strain to be evaluated;

[0022] Preferably, the correlation coefficient is obtained by the cor() function;

[0023] Preferably, the correlation coefficient is calculated using cor(df￥x, df￥y, method='spearman'), wherein df￥x represents a vector set [a, d, ...] of the number of proteins in each strain of the protein group to be analyzed for correlation; df￥y represents a vector set [protein yield 1, protein yield 2, ...] of the protein yield of each strain, and 'spearman' indicates that spearman correlation analysis is selected in the cor() function;

[0024] Preferably, the correlation coefficient between the proteome of each strain and the protein yield is obtained by taking the significance value.

[0025] According to another aspect of the present invention, a method for evaluating the accuracy of a method for evaluating the protein production of a strain is provided, comprising: obtaining a scoring data pair for the protein production of the strain, the scoring data pair for the protein production of the strain being: [strain 1, score Score1, protein production 1], [strain 2, score Score2, protein production 2], ..., [strain n, score Score n, protein production n], wherein n is the strain number; performing a correlation analysis between the score in the scoring data pair for the protein production of the strain and the protein production to obtain a correlation coefficient R and a significance p value; and evaluating the accuracy of the method of claim 6 according to the correlation coefficient R and the significance p value.

[0026] Preferably, the correlation analysis is calculated by calling the cor() function of the R language;

[0027] More preferably, the correlation analysis performed by the cor() function includes: cor(df￥x, df￥y, method='spearman'), wherein df￥x represents the score value vector set of each strain [Score1, Score2, ..., Score n], df￥y represents the protein yield vector set of each strain [protein yield 1, protein yield 2, ..., protein yield n], and 'spearman' indicates that spearman correlation analysis is selected in the cor function.

[0028] According to another aspect of the present invention, a method for evaluating the protein yield of a Fusarium strain is provided, comprising: obtaining the proteins encoded by the genome of the strain to be evaluated and the number thereof; traversally comparing the proteins encoded by the genome of the strain to be evaluated with marker proteins, screening the proteins encoded by the genome whose sequence similarity with the marker protein reaches at least 90%, 95% or 99%, or the proteins encoded by the genome whose sequence identity percentage is Percent identity>70, sequence coverage is Query coverage>70 and E-value is E-value<0.05, to obtain similar proteins; normalizing the number of similar proteins to obtain the normalized number of similar proteins; taking the correlation coefficient of the marker protein corresponding to the similar protein as the weight, summing the normalized numbers to obtain the evaluation value of the strain; wherein the marker protein and the correlation coefficient are screened by the aforementioned method for screening marker proteins related to the protein yield of the strain.

[0029] According to another aspect of the present invention, a method for training a machine learning model for evaluating the protein yield of Fusarium strains is provided, comprising: constructing a machine learning model including a clustering module, a protein yield correlation analysis module, a first comparison module, a scoring module, and a second comparison module that are logically connected in sequence; inputting a plurality of proteins encoded by genomes of a plurality of Fusarium strains in a training set into the clustering module for clustering to obtain a plurality of protein groups; transmitting the number of proteins contained in each protein group in each Fusarium strain in the training set and the protein yield of the Fusarium strain to the protein yield correlation analysis module for correlation analysis and screening to obtain marker proteins and correlation coefficients; and The proteins and marker proteins encoded by the group and their correlation coefficients are transmitted to the first comparison module for comparison to obtain similar proteins and their quantities; the similar proteins and their quantities, and the correlation coefficients of the marker proteins corresponding to the similar proteins are transmitted to the scoring module for evaluation to obtain a strain evaluation value; the strain evaluation value and the protein yield of each strain in the test set are input into the second comparison module for comparison to obtain an accuracy evaluation result; the accuracy evaluation result is fed back to at least one of the clustering module, the protein yield correlation analysis module and the first comparison module, and the parameters of at least one of the clustering module, the protein yield correlation analysis module and the first comparison module are adjusted until the accuracy evaluation result reaches a preset evaluation threshold.

[0030] According to another aspect of the present invention, a device for evaluating protein yield of Fusarium strains is provided, comprising: a protein comparison module and a strain yield scoring module, wherein the protein comparison module pre-stores marker proteins and correlation coefficients obtained by screening through a method of screening marker proteins related to the protein yield of the strain, and is used to receive the proteins encoded by the genome of the strain to be evaluated and the quantity thereof, traversely compare the proteins encoded by the genome of the strain to be evaluated with the marker proteins, screen the proteins encoded by the genome with a sequence similarity of at least 90%, 95% or 99%, or the proteins encoded by the genome with a blast comparison result of Percentidentity>70, Query coverage>70 and E-value<0.05, and obtain similar proteins; the strain yield scoring module is used to normalize the quantity of similar proteins to obtain the normalized quantity of similar proteins, and sum the normalized quantities with the correlation coefficients corresponding to the similar proteins as weights to obtain the evaluation value of the strain.

[0031] According to another aspect of the present invention, a method for screening marker genes related to the protein production of a strain is provided, comprising: clustering multiple genes in the genomes of multiple Fusarium strains according to the categories of proteins encoding the same or similar functions to obtain multiple gene families; making each gene family in each Fusarium strain and the number of genes contained therein correspond to the protein production of the Fusarium strain to form a data set containing multiple data pairs; based on the data set, performing a correlation analysis between the gene family and the protein production, and screening the gene family whose correlation coefficient with the protein production of the Fusarium strain meets a preset threshold as the marker gene.

[0032] According to one embodiment of the present invention, clustering multiple genes in the genomes of multiple Fusarium strains includes: performing gene family analysis on the genes of each Fusarium strain and dividing them into multiple family categories; clustering the genes according to the family categories to obtain multiple gene families.

[0033] According to one embodiment of the present invention, based on a data set, performing a correlation analysis between gene families and protein production includes: obtaining multiple subsets of the data set; performing a correlation analysis based on each subset, screening gene families whose correlation coefficients with the protein production of the Fusarium strain are greater than a preset threshold, and obtaining multiple analysis results; taking the intersection of the multiple analysis results to obtain a marker gene.

[0034] According to one embodiment of the present invention, the Fusarium strains include high-yield strains, medium-yield strains and low-yield strains; among them, the strains with a protein percentage (g) in the product measured by the Kjeldahl method not higher than 50 are low-yield strains, those higher than 50 and not higher than 55 are medium-yield strains, and those higher than 55 are high-yield strains.

[0035] According to another aspect of the present invention, a method for evaluating the protein yield of a Fusarium strain is provided, comprising: obtaining a genome of the strain to be evaluated; traversing and comparing genes in the genome of the strain to be evaluated with marker genes, screening genes whose sequence similarity with the marker genes reaches at least 90%, 95% or 99%, or genes whose blast comparison results reach a sequence identity percentage Percentidentity>70 and a sequence coverage Query coverage>70 and an E-value E-value<0.05, to obtain similar genes; normalizing the number of similar genes to obtain a normalized number of similar genes; taking the correlation coefficient between the marker genes corresponding to the similar genes and the protein yield as a weight, summing the normalized numbers to obtain an evaluation value of the strain; wherein the marker genes are screened by a method of screening marker genes related to the protein yield of the strain.

[0036] According to another aspect of the present invention, there is provided a method for training a machine learning model for evaluating the protein yield of Fusarium strains, comprising: constructing a machine learning model comprising a gene family analysis module, a protein yield correlation analysis module, a third comparison module, an evaluation module and a fourth comparison module which are logically connected in sequence; inputting the genes of the genome of the strains in the training set into the gene family analysis module for clustering to obtain the gene family; transmitting the protein yield of each strain in the training set and the number of gene families and genes contained therein to the protein yield correlation analysis module for correlation analysis to obtain marker genes and correlation coefficients; transmitting the genes and marker genes and correlation coefficients of the genome of the strains in the test set to the third comparison module for comparison to obtain similar genes and their numbers; transmitting the similar genes and their numbers and the correlation coefficients of the marker genes corresponding to the similar genes to the evaluation module for evaluation to obtain strain scores; inputting the strain scores and protein yields of each strain in the test set into the fourth comparison module for comparison to obtain accuracy evaluation results; feeding back the accuracy evaluation results to at least one of the gene family analysis module, the protein yield correlation analysis module and the third comparison module, and adjusting the parameters of at least one of the gene family analysis module, the protein yield correlation analysis module and the third comparison module until the accuracy evaluation results reach a preset evaluation threshold.

[0037] According to another aspect of the present invention, a device for evaluating the protein yield of a Fusarium strain is provided, comprising: a gene comparison module and a strain yield evaluation module, wherein the first comparison module pre-stores marker genes and correlation coefficients obtained by screening by a method for screening marker genes related to the protein yield of the strain, and is used to receive genes of the strain to be evaluated and their quantity, traverse and compare the genes of the strain to be evaluated with the marker genes, screen genes with sequence similarity of at least 90%, 95% or 99%, or genes with blast comparison results of Percentidentity>70, Query coverage>70 and E-value<0.05, to obtain similar genes; the strain yield evaluation module is used to normalize the quantity of similar genes to obtain the normalized quantity of similar genes, use the correlation coefficient between the marker genes corresponding to the similar genes and the protein yield as a weight, sum the normalized quantities, and obtain the evaluation value of the strain.

[0038] According to another aspect of the present invention, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the aforementioned method of screening marker proteins related to the protein production of the strain, the method of evaluating the protein production of the strain, the method of evaluating the accuracy of the method of evaluating the protein production of the strain, the method of evaluating the protein production of the Fusarium strain, the method of training a machine learning model for evaluating the protein production of the Fusarium strain, the method of screening marker genes related to the protein production of the strain, the method of evaluating the protein production of the Fusarium strain, and the method of training a machine learning model for evaluating the protein production of the Fusarium strain are implemented.

[0039] According to another aspect of the present invention, there is provided a device, comprising: a processor; and a memory storing program instructions, wherein when the program instructions are executed by the processor, the device implements the aforementioned method.

[0040] In an embodiment of the present invention, by clustering the proteins encoded by the genome of the Fusarium strain, a protein group with similar functions and structures is obtained. Correlation analysis is performed on the basis of the protein group, which reduces the amount of data processing. By adopting the correlation analysis between the protein group and the protein yield of the strain, the authenticity and reliability of the results of the correlation analysis are improved, and the marker protein that affects the protein yield of the Fusarium strain can be screened. By dividing the data set into a plurality of different subsets and finding the intersection of the results after analysis, the influence of the outlier effect is avoided, so that the marker protein characterizes the protein yield of the strain more accurately. The correlation coefficient and / or significance of the proteome and the protein yield are obtained by the cor() function, and the quantification of the correlation coefficient and / or significance can be achieved, which is convenient for comparison with a preset threshold. By clustering the genes in the genome of the Fusarium strain according to the type of the corresponding protein, a gene family is obtained, and correlation analysis is performed on the basis of the gene family, so that the marker gene obtained by screening can accurately reflect the protein yield, and by dividing the data set into a plurality of different subsets for analysis, the influence of the outlier effect is avoided, and the accuracy of the marker gene reflecting the protein yield is improved.

[0041] The present invention uses a protein group with similar or identical structure and function to mean that the regulatory effect of the protein group has similar effects on the protein yield of the strain. The protein groups that have the same promoting or inhibiting effect on the protein yield of the strain are respectively taken as an object for correlation analysis, and a plurality of protein sequences with the same or similar function are taken as a protein group for correlation analysis, which can improve the accuracy and representativeness of the correlation analysis and reduce the data processing amount of the correlation analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0043] Figure 1 A schematic diagram showing the steps of a method for screening a marker protein associated with protein yield of a strain according to an embodiment of the present invention is shown;

[0044] Figure 2 A schematic diagram showing the steps of a method for evaluating protein yield of a Fusarium strain according to an embodiment of the present invention is shown;

[0045] Figure 3 A schematic diagram of a process flow of screening marker protein data related to protein yield of a strain by performing machine learning on a training set and a test set according to an embodiment of the present invention is shown;

[0046] Figure 4 A block diagram of a machine learning model according to an embodiment of the present invention is shown;

[0047] Figure 5 A block diagram of a device for evaluating protein yield of a strain according to an embodiment of the present invention is shown;

[0048] Figure 6 A schematic diagram showing the steps of a method for screening marker genes associated with protein production of a strain;

[0049] Figure 7 A schematic diagram showing the steps of a method for evaluating protein yield of a Fusarium strain according to another embodiment of the present invention;

[0050] Figure 8 A schematic diagram of a data flow of performing machine learning on a training set and a test set to screen marker genes related to protein production of a strain according to another embodiment of the present invention is shown;

[0051] Fig. 9 A block diagram of a machine learning model according to another embodiment of the present invention is shown;

[0052] Fig.10 A block diagram of a device for evaluating protein yield of strains according to another embodiment of the present invention is shown. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0054] It should be understood that the terms "include" and "comprising" used in the description and claims of the present invention indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0055] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the claims, the singular forms of "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in the specification of the present invention and the claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0056] As used in this specification and claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0057] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0058] Figure 1 A schematic diagram showing the steps of a method for screening a marker protein associated with protein production of a strain according to an embodiment of the present invention is shown.

[0059] like Figure 1 As shown, first in the first step S101, multiple proteins encoded by the genome of each of the multiple Fusarium strains are obtained, and the multiple proteins encoded by the genomes of the multiple Fusarium strains are clustered to obtain multiple proteomes. Among them, the genome of each strain includes multiple gene fragments, so the genome can encode multiple proteins, and the multiple proteins encoded by the genome can be, for example, multiple protein sequences obtained by gene transcription and translation, also known as amino acid sequences. A proteome is a collection of a class of protein sequences with the same or similar sequences, that is, a class of protein sequences with the same or similar sequences are classified into the same proteome. For example, one strain corresponds to a whole genome, and one genome will encode multiple types of proteins, and each type of protein may have multiple protein sequences.

[0060] The proteins encoded by the genome of Fusarium strains include proteins with different functions. Each functional protein can have multiple identical or similar protein sequences. Different protein sequences can be assembled and predicted based on the whole genome data of second-generation sequencing or third-generation sequencing, and the quantity of each protein can also be obtained. The genome of Fusarium strains theoretically includes most of the genetic information of the strains. The proteins encoded by the genome have corresponding functions. The combination of these functions constitutes the main factor affecting the protein production of the strains.

[0061] The present invention classifies proteins with similar sequences or functions into the same proteome and names them "proteome 1", and so on to obtain proteome 2, .... N. The proteome of the present invention can also be collected through literature, and different protein structures with the same function that have been reported and studied can be classified into one proteome. For example, proteins with alcohol dehydrogenase activity include alcohol dehydrogenase A, alcohol dehydrogenase B, alcohol dehydrogenase 1, alcohol dehydrogenase 2, .... alcohol dehydrogenase N, and these different alcohol dehydrogenases can be classified into one proteome 1.

[0062] Then, in the second step S102, each proteome in each Fusarium strain and the number of proteins contained therein are matched with the protein yield of the Fusarium strain to form a data set containing multiple data pairs of proteomes and protein yields. The number of proteins refers to the number of protein sequences of the same type encoded by the genome of the strain. For example, the genome of strain 1 can encode N class A protein sequences and M class B proteins, which means that the number of class A proteins encoded by the strain is N, and the number of class B proteins encoded is M, and so on. For class C proteins and class D proteins, the number of proteins of the corresponding categories encoded by the corresponding strains is counted accordingly. Strain 1 can contain multiple proteomes, and the number of proteins contained in each proteome is calculated according to the above method.

[0063] The data pairs of proteome and protein yield can be composed of:

[0064] {strain 1, proteome 1, protein quantity a, protein yield 1},

[0065] {strain 1, proteome 2, protein quantity b, protein yield 1},

[0066] {strain 2, proteome 1, protein quantity d, protein yield 2},

[0067] {strain 2, proteome 2, protein quantity e, protein yield 2} etc.

[0068] The multiple data pairs thus obtained constitute a data set, in which the strains and protein yields correspond to each other.

[0069] Finally, in the third step S103, based on the data set composed of the data pairs, a correlation analysis is performed between the proteome and the protein yield, and the proteome whose correlation coefficient and / or significance with the protein yield of the Fusarium strain meets the preset threshold is screened as a marker protein, also called a labeled protein.

[0070] The inventors have found that the effects of proteins encoded in the genome of Fusarium strains on the protein production of the strains mainly include: promoting effect, inhibiting effect, and slight effect. For example, whether the protein encoded by the genome is expressed and the amount of expression are affected by many factors. For example, due to the limitations of culture medium conditions, the protein encoded by the genome may not be expressed or the expression amount is small; when a protein encoded by a certain genome is not expressed, it may have a limited effect on the protein production of the strain, and in some cases the effect is very small; in addition, the number of proteins encoded by the genome is also a relatively important factor affecting the protein production of the strain. In the strain, a variety of proteins with different effects and degrees of influence are mixed together, making it difficult to analyze and predict the protein production of the strain based on each protein and its quantity.

[0071] The proteins encoded by the genome of Fusarium strains contain multiple proteins with similar functions and structures. If the correlation analysis and prediction of each protein with the protein yield of the strain are performed separately, there are many shortcomings: on the one hand, the quantitative factor of the protein is not considered, which affects the accuracy of the analysis results. On the other hand, due to the large number of protein types, the amount of data calculation is large, which affects the speed of generating analysis and prediction results.

[0072] Accordingly, the inventors clustered multiple proteins encoded by the genomes of multiple Fusarium strains, and classified proteins with similar structures and functions into a class of proteomes, thereby forming multiple categories of proteomes (hereinafter referred to as multiple proteomes). For example, by diffraction detection, proteins with similar domains are clustered, etc.; protein sequences with similar functions can also be clustered, or different protein structures or sequences with the same function can be clustered and attributed to the same proteome. The proteins encoded by the genomes of various Fusarium strains can also be clustered according to proteins with similar functions, proteins with the same function but different structures, or proteins with the same function but different protein sequences, with each category corresponding to a proteome, to obtain multiple proteomes.

[0073] A protein group with similar or identical structure and function means that the regulatory effect of the proteins in this group has similar effects on the protein production of the strain. The protein group that has the same promoting or inhibiting effect on the protein production of the strain is analyzed as a whole. The correlation analysis is performed by treating multiple protein sequences with the same or similar functions as a protein group, which can not only improve the accuracy and representativeness of the correlation analysis, but also reduce the amount of data processing for the correlation analysis.

[0074] According to an embodiment of the present invention, the result of clustering is to obtain multiple proteomes. As before, each proteome contains a certain number of proteins. For example, the Fusarium genus corresponds to N different Fusarium strains, wherein, among the proteins encoded by a genome of Fusarium strain 1, there are A proteins, the number of which is x, and there are B proteins, the number of which is y. When it is determined through clustering that protein A and protein B belong to the same proteome C, the number of proteins contained in proteome C of the strain is x+y. N Fusarium strains can be deduced by analogy to obtain the proteome encoded by the genome of each strain.

[0075] In an embodiment of the present invention, when constructing a data pair, the protein yield obtained by the actual strain shake flask culture can be measured in advance by the Kjeldahl method, that is, the actual protein yield of each strain is used as the protein yield of the strain in the data pair. By corresponding each proteome in each strain to the protein yield of the strain, a data pair of the proteome and protein yield of the strain is formed. The correlation between the protein yield of the strain and the protein encoded by the strain is introduced, and the data pair is constructed by the proteome and protein yield of multiple strains, and the training set is used for learning and the test set is used for testing. Among the numerous genome-encoded proteins of the Fusarium strain, the factors such as the type and quantity of proteins that have a promoting effect, an inhibitory effect, and a small effect on the protein yield of the strain can be determined, and according to the results obtained, the protein yield of other fungal protein strains in the strain resource library can be predicted. The screening of high protein yield strains can be achieved only by analyzing the proteins encoded by the genome, without the need to measure the protein yield of each strain by shaking flasks and fermentation culture, and the Kjeldahl method, thereby greatly saving manpower, screening time and screening cost. At the same time, the data pairs of proteome and protein yield constructed by the present invention, as well as the analysis method based on the data pairs, can overcome the problem that it is difficult to analyze and predict the protein yield of the strain based on each protein due to the mixing of multiple proteins with different effects and degrees of influence, and achieve the technical effect of accurately screening strains with high protein yields by analyzing the proteins encoded by the genome.

[0076] The data pair includes the protein yield of each strain, corresponding to the multiple proteomes contained in the strain and the number of proteins contained in each proteome. The data set constructed by these data pairs can reflect the association between the proteome categories and protein quantities encoded by the genomes of different strains and the protein yield of the corresponding strains. Table 1 exemplifies a composition form of the data pair.

[0077] Table 1.

[0078]

[0079] As shown in Table 1, the data pairs of the proteome and protein yield of each strain can be composed of:

[0080] {strain 1, proteome 1, protein quantity a, protein yield 1},

[0081] {strain 1, proteome 2, protein quantity b, protein yield 1},

[0082] {strain 2, proteome 1, protein quantity d, protein yield 2},

[0083] {strain 2, proteome 2, protein quantity e, protein yield 2}, etc. A data set is composed of multiple data pairs.

[0084] Based on the data set constructed by the proteome and protein yield of each strain, the correlation analysis between the proteome and protein yield is performed, and the correlation is the comprehensive result of the protein yield of the strain and the type correlation and quantity correlation of the protein in the strain. In an embodiment of the present invention, a linear correlation method can be used for analysis to simplify the process of correlation analysis and reduce the dependence on computing power.

[0085] The correlation analysis between the proteome of the strain and the protein yield can be: construct data pairs in the form of Table 1 to form a data set. For example, to analyze proteome 1, take the data set consisting of the data pairs of the number of proteins contained in proteome 1 in all strains of "multiple Fusarium strains" and perform correlation analysis with the protein yield of the corresponding strain. For example, the calculation can be performed by calling the cor function (cor()) of the R language.

[0086] An example of cor() function call for correlation analysis is: cor(df￥x, df￥y, method='spearman'), where df￥x represents the vector set [a, d, ...] of the number of proteins in all strains of the proteome to be analyzed for correlation (e.g., proteome 1), where a represents the number of proteins corresponding to proteome 1 in strain 1, and d represents the number of proteins corresponding to proteome 1 in strain 2. df￥y represents the vector set [protein yield 1, protein yield 2, ..., protein yield n] of protein yield of all strains, where protein yield 1 is the protein yield of strain 1, and protein yield 2 is the protein yield of strain 2. 'spearman' means that spearman correlation analysis is selected in the cor() function. The cor() function outputs the correlation coefficient R and the significance p value, and the cor() function value of each proteome is calculated in turn to obtain the corresponding correlation coefficient R and significance p value. According to an embodiment of the present invention, the correlation analysis may also select other well-known correlation analysis means or algorithms, including but not limited to: Pearson correlation analysis.

[0087] After correlation analysis, the correlation coefficient and significance of a specific protein group are obtained, and the protein group can be screened according to the values ​​of the correlation coefficient R and the significance p. For example, the protein group with the absolute value of R |R|>0.45 and p<0.05, for example, when R is a positive number, R>0.45, is a positively correlated protein group; when R is a negative number, R<-0.45, is a negatively correlated protein group; therefore, it is believed that the positively correlated protein group will promote protein production, and the negatively correlated protein group will inhibit protein production; the absolute value of R can also be greater than 0.5, 0.6, 0.7...0.99; P can be less than 0.95, 0.9, 0.8, ....0.05, 0.01, 0.005.

[0088] In an embodiment of the present invention, a data set constructed from multiple data pairs can be divided into a training set and a test set. The training set is used to perform correlation analysis and screen for genome-encoded proteins that are significantly correlated with the protein production of the strain, and the test set is used to verify the reliability of the screening results.

[0089] The above correlation analysis can be used for the training set, and the correlation coefficient and significance can be obtained in the same way, thereby obtaining the marker proteins in the training set data that affect protein production. Similarly, according to the values ​​of the correlation coefficient R and the significance p, positively correlated protein groups and negatively correlated protein groups are obtained. Among them, the positively correlated protein group will promote protein production, and the negatively correlated protein group will inhibit protein production. Through the correlation analysis of the data in the training set, the correlation coefficient R and significance p value corresponding to each protein group can be obtained.

[0090] The scores of strains in the test set were calculated as follows:

[0091] The strain scoring formula is: Where n=1, 2.....N, represents the number of the proteome. When n=1, it represents proteome 1, that is, it is the first marker protein (corresponding to proteome 1). N represents the total number of marker proteins or proteomes. The correlation coefficient of proteome n is obtained by the aforementioned cov() function; Score is the score of the strain. Each proteome is multiplied by the number of proteins corresponding to it, and the cumulative sum is obtained to obtain the score of the strain.

[0092] The data pairs used to construct the test set score are as follows: [strain 1, score Score1, protein yield 1], [strain 2, score Score2, protein yield 2], ..., [strain n, score Score n, protein yield n].

[0093] The correlation analysis between the score value and protein yield of the strain can be: according to the constructed data pairs [strain 1, score Score1, protein yield 1], [strain 2, score Score2, protein yield 2]...[strain n, score Score n, protein yield n], a data set is formed. The correlation analysis between the score value and protein yield of the strain can be calculated by calling the cor function of the R language: cor(df￥x, df￥y, method='spearman'), where df￥x represents the score value vector set of all strains [Score1, Score2,...Score n], df￥y represents the protein yield vector set of all strains [protein yield 1, protein yield 2,..., protein yield n], and 'spearman' indicates that the spearman correlation analysis is selected in the cor function. The correlation analysis between the score value and protein yield of the strain can also choose other well-known correlation analyses, including but not limited to: Pearson correlation analysis. The cor() function outputs the correlation coefficient R and the significance p value;

[0094] Thus, the correlation coefficient and significance of the relationship between the strain score and the yield are obtained. For example, according to the values ​​of the correlation coefficient R and the significance p, the accuracy of the scoring method of the present invention is evaluated. For example: the absolute value of R is taken as R>0.45, p<0.05, where R is a positive number, and the R value can also be greater than 0.5, 0.6, 0.7...0.99; P can be less than 0.95, 0.9, 0.8, ....0.05, 0.01, 0.005.

[0095] In the embodiment of the present invention, by clustering the proteins encoded by the genome of the strain according to similar structures and functions, a proteome containing similar proteins can be constructed, and each protein and its quantity in the proteome have a similar effect on the protein yield of the strain. Correlation analysis is performed on the proteome and protein yield, and the obtained marker proteins and their quantities can significantly characterize the protein yield of the strain.

[0096] According to an embodiment of the present invention, a plurality of Fusarium strains include high-yield strains, medium-yield strains and low-yield strains. The strains involved in the embodiments of the present invention may be obtained under the same culture environment, for example, the quantity ratio of any two of the high-yield strains, the medium-yield strains and the low-yield strains is 0.8-1.2. Among them, the percentage of protein in the product measured by the Kjeldahl nitrogen determination method (g) is not higher than 50 for low-yield strains, higher than 50 but not higher than 55 for medium-yield strains, and higher than 55 for high-yield strains. The number of Fusarium strains in high-yield, low-yield and medium-yield strains is evenly distributed, so that the protein yield is evenly distributed, thereby highlighting the influence of proteins encoded by different genomes on the protein yield of the strains, making the results of the correlation analysis more significant.

[0097] According to an embodiment of the present invention, clustering multiple proteins encoded by the genomes of multiple Fusarium strains includes: performing sequence comparison between the proteins encoded by the genomes of the various Fusarium strains and / or performing phylogenetic tree analysis on the proteins, dividing them into multiple categories, and clustering them according to the categories to obtain multiple protein groups.

[0098] Specifically, the proteins encoded by the genome of a certain bacterial species in the training set are compared one by one with the proteins encoded by the genome of itself and other strains, such as blast comparison. A phylogenetic tree is constructed for the proteins, and the proteins encoded by the genomes of the strains involved in the analysis are divided into multiple categories. For example: the 15,250 proteins encoded by the genome of strain A are blast compared with the proteins encoded by other strains in the analysis to construct a phylogenetic tree. The phylogenetic tree summarizes and classifies the encoded proteins of all strains involved in the analysis into 22,736 categories, that is, 22,736 protein groups, of which the 15,250 proteins encoded by the genome of strain A include 9,612 category protein groups.

[0099] According to an embodiment of the present invention, based on a data set constructed from multiple data pairs of proteomes and protein yields, performing correlation analysis between proteomes and protein yields includes: obtaining multiple subsets of the data set constructed from multiple data pairs, for example, the multiple data pairs of the data set can be randomly and evenly divided into N (N is a non-zero integer) parts, performing correlation analysis based on each subset, screening proteomes whose correlation coefficient with the protein yield of the Fusarium strain is greater than a preset threshold, and obtaining multiple analysis results. Take the intersection of multiple analysis results to obtain marker proteins.

[0100] In an embodiment of the present invention, a subset of a data set lacks a part of data pairs relative to the data set. By setting different data pairs missing in each subset, and then performing correlation analysis on each subset as a unit, an analysis result is obtained, and the intersection of each analysis result is obtained. This method of performing correlation analysis on different subsets and then obtaining the intersection can reduce the influence of outlier extreme values ​​of the sample. To obtain multiple subsets of a data set constructed by multiple data pairs, the data of the data set can be divided into N parts, for example, it can be optionally divided into ten parts, and a part of the data pairs are removed in turn for each data set, thereby constructing ten data subsets, and each data subset is subjected to correlation analysis, including correlation analysis of the proteome of the strain and the protein yield, thereby performing ten correlation analyses. This construction method of the subset makes the data pairs evenly distributed in each subset, and through multiple correlation analyses, it is beneficial to improve the accuracy of the analysis results.

[0101] According to an embodiment of the present invention, after the marker proteins are screened, redundancy removal is performed on the marker proteins. In each type of marker protein, if there are protein sequences with completely identical amino acid sequences, only one of the protein sequences is retained.

[0102] According to an embodiment of the present invention, the correlation analysis is a spearman correlation analysis. The preset threshold is a significance level p<0.95, which can be selected as less than 0.9, 0.8, .... 0.05, 0.01, 0.005, 0.001.

[0103] When performing the spearman correlation analysis, the proteome and the number of proteins it contains and the protein yield of the corresponding strain are rank-transformed according to the size, and the rank-transformed result is substituted into the spearman correlation coefficient calculation formula, wherein the spearman correlation coefficient calculation formula is well known. The correlation coefficient R representing the proteome and the protein yield is calculated, and the normal distribution statistic is obtained according to the correlation coefficient and the number of strains, and the corresponding significance p of the normal distribution statistic is obtained by looking up the table. The proteome is screened according to the values ​​of the correlation coefficient R and the significance p. For example: take the protein group with the absolute value of R |R| greater than 0.45 and p < 0.05. When R is a positive number, R>0.45, it is a positively correlated protein group; when R is a negative number, R<-0.45, it is a negatively correlated protein group; therefore, it is believed that the positively correlated protein group will promote protein production, and the negatively correlated protein group will inhibit protein production; the absolute value of R can also be greater than 0.5, 0.6, 0.7...0.99; P can be less than 0.95, 0.9, 0.8,....0.05, 0.01, 0.005.

[0104] In order to determine the stability of the correlation between the proteome and the protein yield of the strain, it can be verified by using the divided subsets. For example, the strain is randomly divided into 10 groups with the same number as much as possible, and each time one of the groups is removed, the spearman correlation analysis of the protein number of the proteome and the protein yield of the strain is performed. The significance p of the 10 correlation analyses is 0.000255613, 0.000104383, 0.00000227, 0.000169075, 0.000124116, 0.000129028,

[0105] 0.000396305, 0.000399709, 0.00009, 0.000181564, all are less than 0.001, so the protein sequence of this genome can be used for subsequent processing to generate marker proteins, and the correlation coefficient R = 0.45 is used as the weight.

[0106] In the embodiment of the present invention, by clustering the encoded proteins of the genome of the Fusarium strain, different proteins with similar effects on the protein yield of the strain can be analyzed collectively, thereby improving the accuracy of the obtained marker protein in characterizing the protein yield of the strain. By setting different subsets for correlation analysis and then finding the intersection to obtain the marker protein, the influence of outliers can be reduced, thereby improving the accuracy of screening marker proteins.

[0107] According to another aspect of the present invention, a method for evaluating the protein yield of a strain is provided, comprising: obtaining multiple proteins encoded by the genome of each of multiple Fusarium strains, clustering the multiple proteins encoded by the genome of multiple Fusarium strains to obtain multiple proteomes; corresponding the number of each proteome in each Fusarium strain and the protein yield of the Fusarium strain to form a data set containing multiple data pairs; based on the data set composed of the data pairs, performing a correlation analysis between the proteome and the protein yield to obtain a correlation coefficient; according to the correlation coefficient, calculating the score value of the protein yield of the strain to be evaluated, and evaluating the protein yield of the strain to be evaluated according to the score value. The calculation formula for the score value of the protein yield of the strain to be evaluated is:

[0108]

[0109] Where Score is the score value of the strain;

[0110] n = 1, 2, ..., N, represents the number of the protein group;

[0111] N represents the total number of proteomes in the strain to be evaluated;

[0112] Preferably, the correlation coefficient of protein group n is obtained by the cor() function;

[0113] Preferably, the correlation coefficient of protein group n is calculated using cor(df￥x, df￥y, method='spearman'),

[0114] Among them, df￥x represents the vector set [a, d, ...] of the number of proteins in each strain of the proteome to be analyzed for correlation;

[0115] df￥y represents the vector set of protein yield of each strain [protein yield 1, protein yield 2, ...],

[0116] Spearman means using the Spearman correlation analysis in the cor() function;

[0117] Preferably, the correlation coefficient between the proteome of each strain and the protein yield is obtained by taking the significance value.

[0118] That is, according to an embodiment of the present invention, the protein group and its correlation coefficient obtained by correlation analysis, as well as the number of proteins in the corresponding proteome in the strain to be evaluated, can be used to calculate the score of the strain, thereby evaluating the protein yield of the strain.

[0119] According to another aspect of the present invention, a method for evaluating the accuracy of the protein production of the above-mentioned evaluation strain is also provided, comprising: obtaining a scoring data pair for the protein production of the strain, the scoring data pair for the protein production of the strain being: [strain 1, score Score1, protein production 1], [strain 2, score Score2, protein production 2]…[strain n, score Score n, protein production n]; performing a correlation analysis between the score in the scoring data pair for the protein production of the strain and the protein production to obtain a correlation coefficient R and a significance p value; and evaluating the accuracy of the protein production of the above-mentioned evaluation strain according to the correlation coefficient R and the significance p value.

[0120] Preferably, the correlation analysis can be calculated by calling the cor() function of the R language;

[0121] Preferably, the correlation analysis performed by the cor() function includes: cor(df￥x, df￥y, method='spearman'), wherein df￥x represents the score value vector set of all strains [Score1, Score2, ... Score n], df￥y represents the protein yield vector set of all strains [protein yield 1, protein yield 2, ..., protein yield n], and 'spearman' indicates that spearman correlation analysis is selected in the cor function.

[0122] That is, the protein yield score of each strain and its actual protein yield can be combined to form new data pairs, and correlation analysis can be performed using multiple data pairs to evaluate whether these score values ​​are accurate.

[0123] Figure 2 A schematic diagram showing the steps of a method for evaluating protein production of a Fusarium strain according to an embodiment of the present invention is shown.

[0124] like Figure 2 As shown, a method 200 for evaluating protein production of Fusarium strains is provided, comprising: step S201, obtaining proteins encoded by the genome of the strain to be evaluated and their quantity. The genome comprises a plurality of gene fragments, and the genome comprising the plurality of gene fragments encodes a plurality of proteins. The plurality of proteins encoded by the genome are, for example, a plurality of protein sequences obtained by gene transcription and translation, also known as amino acid sequences.

[0125] Step S202, traversally compare the proteins encoded by the genome of the strain to be evaluated with the marker proteins, screen the proteins encoded by the genome whose sequence similarity with the marker proteins reaches 90%, 95% or 99%, or the proteins encoded by the genome whose sequence identity percentage Percentidentity>70 and sequence coverage Query coverage>70 and E-value E-value<0.05, and obtain similar proteins.

[0126] Step S203, normalizing the number of similar proteins to obtain a normalized number of similar proteins.

[0127] Step S204, using the correlation coefficient between the marker protein corresponding to the similar protein and the protein yield as a weight, summing the normalized quantities to obtain the evaluation value of the strain, characterizing the magnitude of its influence on the protein yield of the strain. The marker protein is obtained by screening the marker protein related to the protein yield of the strain as described above, and each marker protein has a determined correlation coefficient with the protein yield of the Fusarium strain, which can be obtained by calling the cor() function as described above, or by checking known literature or databases, or by artificially assigning the correlation coefficient based on experience after a large number of experiments.

[0128] The proteins encoded by the genome of the strain to be evaluated and their quantities can be obtained by assembling and predicting based on second-generation sequencing or third-generation sequencing data and measuring based on the Kjeldahl nitrogen method.

[0129] When traversing and comparing the proteins encoded by the genome of the strain to be evaluated with the marker proteins, blast comparison is used, and similar proteins are counted to obtain their number. Since the overall number of marker proteins is small, the marker proteins can be used to traverse and compare the proteins encoded by the genome of the strain to be evaluated to obtain similar proteins and their number. The proteins encoded by the genome with a sequence similarity of 90%, 95% or 99% with the marker protein are screened, and the remaining proteins with a similarity lower than this are qualified as non-related proteins, that is, they have no effect on the protein production of the strain or the effect is so small that it can be ignored. According to one embodiment of the present invention, the similarity screening conditions are blast comparison results Percentidentity>70, Query coverage>70, and E-value<0.05.

[0130] Normalize the number of similar proteins to unify the dimensions for easy calculation. This includes using Z Score normalization based on the mean and variance of the number of marker proteins and proteins in each protein group, and performing mean centering on the original data: subtracting the mean from the data, that is, subtracting the mean of the column from each number. Perform standard deviation normalization on the mean-centered data: divide the data after subtracting the mean from each data by the standard deviation of the data.

[0131] In the embodiment of the present invention, the influence of each marker protein on the protein yield of the strain is simplified into a multivariate linear correlation, and the weighted sum of the quantities is calculated as an evaluation value for evaluating the relative level of protein yield of the strain. Based on this, on the one hand, the differences in protein yield levels of different strains can be accurately compared, and on the other hand, the complex calculation of nonlinear correlation can be avoided, thereby improving the speed of obtaining the evaluation value.

[0132] Take strain A as an example to explain the evaluation process:

[0133] The results of obtaining the proteins encoded by the genome of strain A and comparing them with 50 groups of marker proteins are as follows: Figure 2 As shown, the numbers of the marker proteins and the quantities of similar proteins present in strain A are shown, respectively.

[0134] Table 2.

[0135]

[0136]

[0137] The number of similar proteins of the strains was summed up with the correlation coefficient of the corresponding marker protein as the weight, and the score was calculated.

[0138] The correlation coefficient is a weight, which can be calculated by calling the cor function (i.e., function cor()) of the R language. An example of calling the cor() function for correlation analysis is:

[0139] cor(df￥x,df￥y,method='spearman'),

[0140] Wherein, df￥x represents the vector set [a, d, ..., m] of the number of proteins of labeled protein n in all strains, where a represents the number of proteins corresponding to labeled protein 1 in strain 1, d represents the number of proteins corresponding to labeled protein 2 in strain 2, and m represents the number of proteins corresponding to strain n;

[0141] df￥y represents the vector set of protein yields of all strains [protein yield 1, protein yield 2, ..., protein yield n], protein yield 1 is the protein yield of strain 1, protein yield 2 is the protein yield of strain 2, and protein yield n is the protein yield of strain n;

[0142] 'spearman' means using spearman correlation analysis in the cor function.

[0143] Other well-known correlation analysis methods may also be used for correlation analysis, including but not limited to: Pearson correlation analysis. The cor() function outputs the correlation coefficient R and the significance p value; the cor() function value of each marker protein n is calculated in turn to obtain the corresponding correlation coefficient R and significance p value.

[0144] Thus, the correlation coefficient and significance of specific marker proteins are obtained. For example, the marker proteins are screened according to the values ​​of the correlation coefficient R and the significance p. For example, the absolute value of R is taken as |R|>0.45, p<0.05. For example, when R is a positive number, R>0.45, it is a positively correlated marker protein; when R is a negative number, R<-0.45, it is a negatively correlated marker protein; therefore, it is believed that positively correlated marker proteins will promote protein production, and negatively correlated marker protein groups will inhibit protein production; the absolute value of R can also be greater than 0.5, 0.6, 0.7...0.99; P can be less than 0.95, 0.9, 0.8, ....0.05, 0.01, 0.005;

[0145] Positively correlated marker proteins were selected for further strain scoring calculation:

[0146] The calculation formula is Where n = 1, 2.....N; Score is the score of the strain; n = 1, representing marker protein 1, the number of similar proteins corresponds to the number of similar proteins of marker protein 1; each marker protein is multiplied by the number of similar proteins corresponding to it and then added up to obtain the score of the strain. For example, the score is calculated to be 13.06, which is a relatively low-scoring evaluation value. The protein yield of strain A is measured by the Kjeldahl method to verify the accuracy of the score. For example: if the protein yield is 47.80, it is a low-yield strain, corresponding to the score. For example, the score of the strain can be judged based on the actual protein yield, which is used to feedback the correction algorithm, and the correlation coefficient and score value calculation are continuously recalculated based on the adjusted algorithm until the score of the strain can accurately evaluate the actual protein yield.

[0147] In order to further determine whether the results of the correlation analysis between the marker protein and the protein production of the strain are reasonable, the inventors conducted functional prediction and judgment on the marker proteins obtained by screening, specifically including: comparing the marker protein sequences through a search algorithm based on the hidden Markov model of the pfam database, predicting the possible domains of the marker protein sequences, screening the common domains in the marker proteins, and investigating the relationship between such domains and the physiological activities of Fusarium through literature, thereby predicting the function of the marker protein sequence.

[0148] Positively correlated marker proteins can be obtained by calling a correlation function (such as the cor() function in the R language).

[0149] Among the marker proteins (also called marker proteins) that are significantly negatively correlated with the protein yield of the strains in Table 2:

[0150] The phosphoenolpyruvate phosphatase corresponding to OG0000335 is a common gene cluster in the Fusarium solan ispecies complex, to which many crop pathogens of Fusarium are classified. Therefore, it may be related to Fusarium infection of plants. As a compensation, the protein production ability of the carrying strain is weakened.

[0151] The alcohol dehydrogenase corresponding to OG0002432 is highly expressed under anaerobic conditions of tomato Fusarium, so it may be related to the survival of Fusarium under anaerobic conditions, which weakens the protein production ability of Fusarium strains in aerobic fermentation of protein production.

[0152] The amine oxidase corresponding to OG0009357 is involved in oxidative deamination in Fusarium oxysporum, which may affect the absorption of ammonium salts in the culture medium during protein production fermentation.

[0153] The glutathione S-transferase corresponding to OG0010079 has the effect of improving benzimidazole resistance in Fusarium graminearum, so it may weaken the protein production capacity as a compensation.

[0154] Among the marker proteins (also called marker proteins) that are significantly positively correlated with the protein yield of the strains in Table 2,

[0155] The phospholipase corresponding to OG0014423 has the function of promoting hyphal growth, conidia germination, perithecia and colony formation in Fusarium graminearum;

[0156] The glutaminase corresponding to OG0014589 has the function of promoting hyphal growth and sexual reproduction in Fusarium graminearum;

[0157] The lipase corresponding to OG0014598 is involved in the decomposition of complex substrates and subsequent biomass synthesis in filamentous fungi including Fusarium, and may play a role in decomposing and utilizing the oil added as a defoaming agent during the fermentation process and promoting protein synthesis.

[0158] The marker protein screened by the method of the embodiment of the present invention can effectively and accurately characterize the protein yield of the strain.

[0159] The above scheme of the present invention can be implemented with the help of program instructions. Therefore, the present invention also provides a device, including: a processor, which is used to execute program instructions; and a memory, which stores program instructions, when the program instructions are loaded and executed by the processor, the processor executes the method of screening marker proteins related to the protein yield of the strain or the method of evaluating the protein yield of the Fusarium strain in the above embodiment.

[0160] Figure 3 A schematic diagram of a data flow of performing machine learning on a training set and a test set to screen marker proteins related to protein production of a strain according to an embodiment of the present invention is shown.

[0161] Process 300 includes:

[0162] The proteins encoded by the genome of the training set are clustered and analyzed in step S301 to obtain proteomic data, including proteins and their quantities in each proteome of each strain, as well as protein yield of each strain, and then protein yield correlation analysis and redundancy removal are performed to obtain marker proteins and correlation coefficients. Correlation analysis includes but is not limited to: Spearman, Pearson, etc.

[0163] When performing correlation analysis S302 on the proteins encoded by the genome, in order to ensure that the analysis process is as little affected by outliers as possible, the training set data is divided into N parts (e.g., 10 parts, 20 parts, 30 parts, ..., 100 parts), the number of gene family proteins in one of the training set data is removed in turn, and the data set after removal is subjected to correlation analysis of protein production, and gene family proteins in which N correlation analyses are significantly correlated with protein production are selected for redundancy removal as marker proteins, and the corresponding correlation coefficients, such as the mean and variance of the number of proteins, are retained; wherein the gene family proteins can be: a plurality of genes in the genomes of a plurality of Fusarium strains are clustered to obtain a plurality of gene families;

[0164] When using the test set for verification, the proteins encoded by the genome of each strain in the test set are compared with the marker proteins S303, and the proteins encoded by the genome similar to each marker protein are counted, and the counts are normalized S304 according to the mean and variance of the marker protein, and the correlation coefficient is used as a weight to sum up and obtain the evaluation value. The evaluation value of each strain in the test set and the protein yield corresponding to each strain are compared S305 to evaluate the accuracy of the marker protein and its corresponding correlation coefficient. For example, the protein yields of three strains are ranked and compared with the ranking of the evaluation values ​​of the three strains. If they are consistent, the accuracy is high, and if they are inconsistent, the accuracy is low.

[0165] When the marker protein and its corresponding correlation coefficient are inaccurate, feedback can be given to the clustering step and the correlation analysis step in S306 to adjust the clustering grouping strategy. The significance parameter can also be adjusted in the correlation analysis step until the accuracy of the evaluation value reaches a preset evaluation threshold, for example, the accuracy of the evaluation value reaches 80%.

[0166] The above steps can be achieved by loading a program with relevant functions on the computer.

[0167] Figure 4 A block diagram of a machine learning model according to an embodiment of the present invention is shown.

[0168] like Figure 4 As shown, a machine learning model 400 is constructed, including a clustering module 401, a protein yield correlation analysis module 402, a first comparison module 403, a scoring module 404, and a second comparison module 405 which are logically connected in sequence.

[0169] The clustering module 401 is provided with an interface, and has the function of receiving the data of the protein encoded by the genome, the data of the protein yield of the corresponding strain, and performing cluster analysis on the protein encoded by the genome, and its output result is the data of the protein group and the protein yield of the corresponding strain. For example, the protein group 1 contains multiple protein sequences.

[0170] The protein yield correlation analysis module 402 has the functions of protein yield correlation analysis and redundancy removal, receives the output result of the clustering module, and outputs it as a marker protein and a correlation coefficient after processing. The marker protein is a protein group after screening.

[0171] The first comparison module 403 has the function of comparing proteins, receives the output result of the protein yield correlation analysis and the protein encoded by the strain genome in the test set, compares the two, and outputs similar proteins and their quantities.

[0172] The scoring module 404 receives the output result of the first comparison module, performs normalization operation and weighted sum calculation based on the correlation coefficients corresponding to similar proteins, and outputs the evaluation value of the strain. The correlation coefficients corresponding to similar proteins are the correlation coefficients of the corresponding marker proteins.

[0173] The second comparison module 405 receives the protein production of the test set strains and the evaluation values ​​of the strains, and compares the two. For example, multiple strains are sorted according to the evaluation values, and at the same time, multiple strains are sorted according to the protein production. The difference between the two sortings can be compared and the difference can be quantified as an accuracy evaluation result in the form of a percentage.

[0174] The second comparison module 405 feeds back the accuracy evaluation result to at least one of the clustering module 401, the protein yield correlation analysis module 402 and the first comparison module 403 to adjust the parameters. For example, the second comparison module 405 feeds back the result to the clustering module 401 and the protein yield correlation analysis module 402, and adjusts the parameters of the two modules until the accuracy evaluation result reaches a preset evaluation threshold, which can be set to 50%, 60%, 70%, 80%, 90%, 95%, 96%, 97%, 98%, 99%, etc.

[0175] Figure 5 A block diagram of a device for evaluating protein yield of a strain according to an embodiment of the present invention is shown.

[0176] When the result of the accuracy evaluation reaches a preset evaluation threshold, for example, the accuracy rate reaches 80%, the machine learning model training is completed, and an evaluation model 500 loaded in the evaluation device is obtained, including a protein comparison module 501 and a strain yield scoring module 502. The protein comparison module 501 is Figure 4 The first comparison module 403 after training is completed in FIG4, and the strain yield scoring module 502 is the scoring module 404 after training is completed in FIG4. The marker protein and the correlation coefficient are pre-stored in the protein comparison module 501, which characterizes the influence of the marker protein on the strain protein yield. For example, the encoded protein of the genome of the strain to be evaluated is input into the protein comparison module, and the correlation coefficient of the similar protein and the quantity and the corresponding marker protein is output to the strain yield scoring module, and then processed by the scoring module to output the strain score of the strain to be evaluated.

[0177] Figure 6 A schematic diagram showing the steps of a method for screening marker genes associated with protein production of a strain.

[0178] like Figure 6As shown, the method 600 for screening marker genes related to protein production of strains includes: in step S601, clustering multiple genes in the genomes of multiple Fusarium strains to obtain multiple gene families. In step S602, the number of each gene family and the genes contained therein in each Fusarium strain is matched with the protein production of the Fusarium strain to form a data set containing multiple data pairs. In step S603, based on the data set, a correlation analysis between the gene family and the protein production is performed, and a gene family whose correlation coefficient with the protein production of the Fusarium strain meets a preset threshold is screened as a marker gene.

[0179] In the genome of Fusarium strains, there are structural genes encoding proteins, regulatory genes with regulatory functions, and other genes, such as ribosomal RNA genes. Protein-encoding genes are limited by the surrounding environment, and their expression and expression amount are inconsistent with the actual copy number of the encoding gene. Therefore, the inventors proposed a method for screening markers related to the protein production of the strain, so that the screened markers can accurately characterize the protein production of the strain, while reducing the dependence on computing power.

[0180] In step S601, a certain number of strains are obtained, and the genes in the genome of the strains and the protein yield of the strains are obtained. The genome genes of the strains can be obtained by gene sequencing or querying an existing database. The protein yield of the strain can be obtained by fermentation test. The genomes of multiple Fusarium strains contain multiple genes, and multiple gene families can be obtained through gene family analysis.

[0181] In step S602, for a certain strain, its genome contains multiple genes, and the multiple genes of the strain are mapped to the gene families obtained in step S601, then the strain contains multiple gene families, each gene family contains one or more genes, thereby obtaining gene families containing different genes and data pairs corresponding to the protein production of the strain.

[0182] For other strains, the same method can be used to obtain multiple data pairs, and the multiple data pairs constitute a data set.

[0183] Table 3 shows the components of the data pairs comprising the gene families.

[0184] Table 3.

[0185]

[0186] As shown in Table 3, the data pair can be composed of:

[0187] {strain 1, gene family 1; quantity f; yield X},

[0188] {strain 1, gene family 2; quantity g; yield X},

[0189] {strain 2, gene family 1; quantity l; yield Y},

[0190] {strain 2, gene family 2; quantity m; yield Y}, etc. A data set is composed of multiple data pairs.

[0191] In step S603, a correlation analysis with protein production is performed on a gene family basis, and a gene family with a greater correlation is obtained as a marker gene.

[0192] In the embodiment of the present invention, the genes contained in the genome of the Fusarium strain are first clustered according to the gene family, and the genes with similar sequences are divided into the same gene family. The principle is that the coding function, regulatory function or other functions of the same gene family are basically the same, and the influence on the protein expression of the strain is basically the same. The correlation analysis of protein yield based on this unit can accurately determine the correlation relationship, so that the marker gene finally obtained can effectively and accurately characterize the protein yield of the strain.

[0193] According to one embodiment of the present invention, clustering multiple genes in the genomes of multiple Fusarium strains includes: performing gene family analysis on the genes of each Fusarium strain and dividing them into multiple family categories; clustering the genes according to the family categories to obtain multiple gene families.

[0194] Gene family analysis includes constructing a gene library from the genes of the genomes of multiple Fusarium strains, performing blast comparison on the genes in the gene library, and classifying genes with similar sequences into the same gene family.

[0195] According to an embodiment of the present invention, based on a data set, performing a correlation analysis between a gene family and protein production includes: obtaining multiple subsets of the data set; performing a correlation analysis based on each subset, screening a gene family whose correlation coefficient with the protein production of the Fusarium strain is greater than a preset threshold, and obtaining multiple analysis results; taking the intersection of the multiple analysis results to obtain a marker gene. The marker gene is a collection of multiple gene families.

[0196] In an embodiment of the present invention, a subset of a data set lacks a portion of data pairs relative to the data set. By setting different data pairs missing from each subset, correlation analysis is performed on each subset as a unit to obtain analysis results, and the intersection of the respective analysis results is calculated. This method of performing correlation analysis on different subsets and then obtaining the intersection can reduce the influence of outlier extreme values ​​of the sample. For example, the data of the data set is divided into ten parts, and one of the parts is removed in turn to construct ten subsets, and ten correlation analyses are performed. This method of constructing subsets makes the data pairs evenly distributed in each subset, which is conducive to improving the accuracy of the analysis results.

[0197] According to an embodiment of the present invention, after the marker genes are screened and obtained, the same genes in the marker genes are merged and de-redundancy processing is performed.

[0198] According to an embodiment of the present invention, multiple Fusarium strains include high-yield strains, medium-yield strains and low-yield strains; the quantity ratio of any two of the high-yield strains, the medium-yield strains and the low-yield strains is 0.8-1.2; wherein, the protein percentage (g) in the product measured by the Kjeldahl nitrogen method is not higher than 50 for the low-yield strain, higher than 50 and lower than 55 for the medium-yield strain, and higher than 55 for the high-yield strain.

[0199] Figure 7 A schematic diagram showing the steps of a method for evaluating protein production of a Fusarium strain according to another embodiment of the present invention is shown.

[0200] like Figure 7 As shown, a method 700 for evaluating the protein production of Fusarium strains is provided, comprising: step S701, obtaining the genome of the strain to be evaluated. Step S702, traversing and comparing the genes in the genome of the strain to be evaluated with the marker genes, screening genes whose sequence similarity with the marker genes reaches at least 90%, 95% or 99%, or genes whose blast comparison results reach a sequence identity percentage Percentidentity>70 and a sequence coverage Query coverage>70 and an E-value E-value<0.05, to obtain similar genes. Step S703, normalizing the number of similar genes to obtain a normalized number of similar genes. Step S704, using the correlation coefficient between the marker genes corresponding to the similar genes and the protein production as a weight, summing the normalized numbers to obtain the evaluation value of the strain. Among them, the marker genes are obtained by Figure 6 The method was screened.

[0201] When the genome of the strain to be evaluated is traversed and compared with the marker gene, blast comparison is used, and similar genes are counted to obtain their number. The number of similar genes is normalized to unify the number dimensions of similar genes and marker genes. Z Score normalization can be used to mean-center the original data and normalize the standard deviation of the mean-centered data.

[0202] When calculating the evaluation value of the strain, it is obtained in a linear polynomial manner, and the number of similar genes of the strain is added according to the correlation coefficient of the corresponding marker gene as the weight to obtain the evaluation value score. The protein production of the strain can be judged according to the score.

[0203] The above method can also be implemented through machine learning.

[0204] Figure 8 A schematic diagram of a data flow for screening marker genes related to protein production of strains by performing machine learning through training sets and test sets according to another embodiment of the present invention is shown.

[0205] In the training set, the genomes of multiple strains are clustered to obtain gene families through the gene family analysis step S801. The genes of each strain are counted according to the gene family and corresponded to the protein production of the strain to obtain the gene family data of the strain, that is, the data set.

[0206] In the test set, each gene and its copy number in the genome of each strain can be directly obtained.

[0207] In step S802 of performing strain protein yield correlation analysis using training set data, a gene family strongly correlated with the protein yield of the strain can be obtained, and its correlation coefficient can be obtained as a purpose for evaluating the protein yield of the strain. In this process, there are two steps in which parameters can be adjusted, namely the gene family analysis step and the protein yield correlation analysis step. For example, when dividing the gene family, the ratio of similarity is adjusted. For another example, when performing correlation analysis, the parameters of the spearman algorithm are adjusted. By adjusting these parameters, different marker genes and correlation coefficients can be obtained.

[0208] In addition, when performing the test, the similarity parameter of the comparison between the genome of the test set and the marker gene in step S803 may be adjusted.

[0209] For a certain strain in the test set, after the comparison in step S803, similar genes and their numbers are obtained, and then normalized and summed in step S804 to obtain the score of the strain, that is, the evaluation value of the strain. Then enter step S805. At this time, according to the comparison of the protein yield of the strain and the strain score, the result of whether the marker gene and its correlation coefficient are accurate is obtained. When the result is inaccurate, in step S806, return to the gene family analysis in step S801 and the protein yield correlation analysis and redundancy removal step in step S802 and the comparison step of step S803 to adjust the parameters of each step and improve the accuracy of the marker gene and its correlation coefficient. When the result is accurate, the machine learning ends and a trained model is obtained.

[0210] Fig. 9 A block diagram of a machine learning model according to another embodiment of the present invention is shown.

[0211] like Fig. 9 As shown, the machine learning model 900 for evaluating the protein production of Fusarium strains includes: a gene family analysis module 901, a protein production correlation analysis module 902, a third comparison module 903, an evaluation module 904 and a fourth comparison module 905 which are logically connected in sequence.

[0212] During training, the genes of the genome of the strain in the training set are input into the gene family analysis module 901 for clustering to obtain the gene family, and the protein yield of each strain in the training set and the number of genes contained in the gene family are transmitted to the protein yield correlation analysis module 902 for correlation analysis to obtain the marker gene and the correlation coefficient. The genes, marker genes and correlation coefficients of the genome of the strain in the test set are transmitted to the third comparison module 903 for comparison to obtain similar genes and their numbers. Similar genes and their numbers, and the correlation coefficients of the marker genes corresponding to the similar genes are transmitted to the evaluation module 904 for evaluation to obtain the strain score or strain evaluation value. The strain score and protein yield of each strain in the test set are input into the fourth comparison module 905 for comparison to obtain the accuracy evaluation result. The accuracy evaluation result is fed back to at least one of the gene family analysis module 901, the protein yield correlation analysis module 902 and the third comparison module 903, and the parameters of at least one of the gene family analysis module 901, the protein yield correlation analysis module 902 and the third comparison module 903 are adjusted until the accuracy evaluation result reaches the preset evaluation threshold. The evaluation threshold may be set to 50%, 60%, 70%, 80%, 90%, 95%, 96%, 97%, 98%, 99%, etc.

[0213] Fig.10 A block diagram of a device for evaluating protein yield of strains according to another embodiment of the present invention is shown.

[0214] like Fig.10 As shown, a device 1000 for evaluating protein production of a Fusarium strain comprises:

[0215] Gene comparison module 1001 and strain yield evaluation module 1002,

[0216] The gene comparison module 1001 has marker genes and correlation coefficients pre-stored therein, which are Fig. 9 The third comparison module 903 after the training is completed is used to receive the genes and their numbers of the strain to be evaluated, traverse and compare the genes of the strain to be evaluated with the marker genes, and screen the genes with sequence similarity of at least 90%, 95% or 99%, or the genes with blast comparison results of Percentidentity>70, Query coverage>70 and E-value<0.05, to obtain similar genes;

[0217] The strain yield evaluation module 1002 is Fig. 9 The evaluation module 904 at the end of the training is used to normalize the number of similar genes to obtain the normalized number of similar genes, and to sum the normalized numbers using the correlation coefficient between the marker gene corresponding to the similar gene and the protein yield as the weight to obtain the evaluation value of the strain.

[0218] The method for screening marker proteins related to protein yield of strains, the method for evaluating protein yield of Fusarium strains, the method for training a machine learning model for evaluating protein yield of Fusarium strains, the device for evaluating protein yield of Fusarium strains, and the method for screening marker genes related to protein yield of strains in the embodiments of the present invention are all completed by a computer system. The system may include a device and its peripheral devices and an external network, wherein the device executes the above method or the function of the above device to realize the above technical solution.

[0219] The system may include a CPU, which may be a general-purpose CPU, a dedicated CPU, or other information processing and program execution unit. It may also include a large-capacity memory and a read-only memory ROM, wherein the large-capacity memory may be configured to store various types of data, including various programs such as pre-training sets, sample training sets, weight parameters, and detection results, and the ROM may be configured to store power-on self-test for the system, initialization of various functional modules in the system, basic input / output drivers for the system, and data required to boot the operating system.

[0220] The system also includes other hardware platforms or components, such as TPU, GPU, FPGA and MLU. It is understandable that those skilled in the art can add or remove corresponding hardware according to actual needs.

[0221] The system also includes a communication interface, so that it can be connected to a local area network / wireless local area network (LAN / WLAN) through the communication interface, and then connected to a local server or to the Internet ("Internet") through the LAN / WLAN. Alternatively or additionally, the system can also be directly connected to the Internet or a cellular network through the communication interface based on wireless communication technology, such as wireless communication technology based on the third generation ("3G"), fourth generation ("4G") or fifth generation ("5G") generation. In some application scenarios, the system can also access servers and possible databases of external networks as needed to obtain various known neural network models, data and modules, and can remotely store various measured data.

[0222] The peripherals of the system may include a display device, an input device, and a data transmission interface. In one embodiment, the display device may, for example, include one or more speakers and / or one or more visual displays, which are configured to perform voice prompts and / or image video displays on the computing process or screening results of the device of the present invention. The input device may include, for example, a keyboard, a mouse, a microphone, a gesture capture camera, or other input buttons or controls, which are configured to receive input or user instructions for training data. The data transmission interface may include, for example, a serial interface, a parallel interface or a universal serial bus interface ("USB"), a small computer system interface ("SCSI"), a serial ATA, a FireWire ("FireWire"), a PCI Express, and a high-definition multimedia interface ("HDMI"), etc., which are configured for data transmission and interaction with other devices or systems.

[0223] The CPU, mass storage, read-only memory ROM, TPU, GPU, FPGA, MLU and communication interface can be interconnected through a bus, and data interaction with peripheral devices can be achieved through the bus. In one embodiment, through the bus, the CPU can control other hardware components and peripheral devices in the system.

[0224] The processor CPU can receive training data through an input device or a data transmission interface, and call the computer program instructions or codes (such as codes involving neural networks) stored in the memory to train or screen the marker proteins related to the protein yield of the strain to obtain the weight parameters or screening results of the trained marker proteins. After the CPU determines the result by executing the program instructions, the result can be displayed on a display device or output by voice prompts. In addition, the system can also upload the results to a network, such as a remote database, through a communication interface.

[0225] It should also be understood that any module, unit, component, server, computer, terminal or device that executes instructions of the present invention examples may include or otherwise access computer-readable media, such as storage media, computer storage media or data storage devices (removable) and / or non-removable) such as disks, optical disks or tapes. Computer storage media may include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules or other data.

[0226] The present invention also provides a computer-readable storage medium having computer-readable instructions stored thereon. When the computer-readable instructions are executed by one or more processors, the method of screening a marker protein related to the protein production of a strain or the method of evaluating the protein production of a Fusarium strain according to any of the foregoing embodiments is implemented.

[0227] The computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as RRAM (Resistive Random Access Memory), DRAM (Dynamic Random Access Memory), SRAM (Static Random-Access Memory), EDRAM (Enhanced Dynamic Random Access Memory), HBM (High-Bandwidth Memory), HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by the application, module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0228] Although multiple embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art may conceive of many changes, modifications, and alternatives without departing from the thought and spirit of the present invention. It should be understood that in the process of practicing the present invention, various alternatives to the embodiments of the present invention described herein may be adopted. The appended claims are intended to define the scope of protection of the present invention, and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A method for screening a marker protein associated with protein production of a Fusarium strain, characterized in that: include: obtaining multiple proteins encoded by the genome of each of the multiple Fusarium strains, and clustering the multiple proteins encoded by the genomes of the multiple Fusarium strains to obtain multiple proteomes; The number of each proteome and the proteins contained therein in each Fusarium strain is made to correspond to the protein yield of the Fusarium strain to form a data set including a plurality of data pairs; Based on the data set composed of data pairs, a correlation analysis is performed between the proteome and the protein yield to obtain a correlation coefficient and / or significance, and the proteome whose correlation coefficient and / or significance with the protein yield of the Fusarium strain meets a preset threshold is screened as a marker protein.

2. The method according to claim 1, characterized in that The multiple proteins encoded by the genomes of multiple Fusarium strains were clustered to obtain multiple proteomes including: Perform sequence alignment and protein phylogenetic tree analysis on the proteins encoded by the genomes of the Fusarium strains, and divide them into multiple categories, each category corresponding to a protein group, to obtain multiple protein groups; and / or The proteins encoded by the genome of each Fusarium strain are clustered according to proteins with similar functions, proteins with similar functions but different structures, or proteins with similar functions but different protein sequences. Each category corresponds to a protein group, and multiple protein groups are obtained.

3. The method according to claim 1 or 2, characterized in that: The number of each proteome in each Fusarium strain and the proteins contained therein are made to correspond to the protein yield of the Fusarium strain to form a data set containing multiple data pairs, wherein the data pairs are in the form of: {strain n, protein group 1, protein amount a, protein yield x}, {strain n, protein group 2, protein quantity b, protein yield x}, ...... {strain m, proteome 1, protein quantity d, protein yield y}, {strain m, proteome 2, protein quantity e, protein yield y}; m and n represent strain numbers; a, b, c, and d represent the amount of protein; x and y represent protein production.

4. The method according to claim 3, characterized in that: Based on the data set composed of data pairs, performing correlation analysis between the proteome and the protein yield comprises: obtaining a plurality of subsets of the data set; Correlation analysis was performed based on each subset separately; Screening a protein group whose correlation coefficient and / or significance with the protein yield of the Fusarium strain is greater than a preset threshold to obtain multiple analysis results; The intersection of multiple analysis results is taken to obtain the marker protein.

5. The method according to claim 1, characterized in that The correlation coefficient and / or significance between the proteome and protein production is obtained by the cor() function.

6. The method according to claim 5, characterized in that The correlation coefficient and / or significance between the proteome and protein yield is calculated using cor(df$x, df$y, method = 'spearman'), wherein df$x represents a vector set [a, d, ...] of the number of proteins in each strain of the proteome to be analyzed for correlation, df$y represents a vector set [protein yield 1, protein yield 2, ...] of the protein yield of each strain, and 'spearman' indicates that spearman correlation analysis is selected in the cor() function.

7. The method according to claim 1, characterized in that By taking the significance value, the correlation coefficient between the proteome and protein yield of each strain was obtained.

8. The method according to claim 1 or 2, characterized in that: Also includes: According to the values ​​of the correlation coefficient and / or significance, positively correlated marker proteins and negatively correlated marker proteins of the Fusarium strain are obtained, wherein the positively correlated marker proteins promote protein production, and the negatively correlated marker proteins inhibit protein production; Among them, multiple Fusarium strains include high-yield strains, medium-yield strains and low-yield strains according to protein yield. The strains with a protein percentage (g) in the product measured by Kjeldahl nitrogen method not higher than 50 are low-yield strains, those higher than 50 and lower than 55 are medium-yield strains, and those higher than 55 are high-yield strains.

9. A method for evaluating protein production of a strain, characterized in that include: The method for screening a marker protein associated with protein production of a Fusarium strain according to any one of claims 1 to 8, obtaining a correlation coefficient between the proteome and the protein production; According to the correlation coefficient, the score value of the protein yield of the strain to be evaluated is calculated, and the strain to be evaluated is evaluated according to the score value. of protein yield; The scoring formula for the protein yield of the strain to be evaluated is: (correlation coefficient of proteome n × number of proteins in the corresponding proteome), Among them, Score is the score value of the strain; n = 1, 2, …, N, represents the number of the protein group; N represents the total number of proteomes in the strain to be evaluated.

10. The method according to claim 9, characterized in that The correlation coefficient is obtained by the cor() function.

11. The method according to claim 10, characterized in that The correlation coefficient is calculated using cor(df$x, df$y, method = 'spearman'), where df$x represents a vector set [a, d, ...] of the number of proteins in each strain of the protein group to be analyzed for correlation; df$y represents a vector set [protein yield 1, protein yield 2, ...] of the protein yield of each strain, and 'spearman' indicates that spearman correlation analysis is selected in the cor() function.

12. The method according to any one of claims 9 to 11, characterized in that: By taking the significance value, the correlation coefficient between the proteome and protein yield of each strain was obtained.

13. A method for evaluating the accuracy of the method according to any one of claims 9 to 12, characterized in that: include: Obtaining a score data pair for protein production of the strain, wherein the score data pair for protein production of the strain is: [strain 1, score Score 1, protein production 1], [strain 2, score Score 2, protein production 2], ..., [strain n, score Score n, protein production n], wherein n is the strain number; Performing correlation analysis on the score of the protein production score of the strain and the protein yield to obtain the correlation coefficient R and the significance p value; The accuracy of the method described in any one of claims 9 to 12 is evaluated based on the correlation coefficient R and the significance p value.

14. The method according to claim 13, characterized in that The correlation analysis is calculated by calling the cor() function of the R language.

15. The method according to claim 14, characterized in that The cor() function performs correlation analysis including: cor(df$x, df$y, method = 'spearman'), wherein df$x represents the score value vector set of each strain [Score1, Score2, ... Score n], df$y represents the vector set of protein yield of each strain [protein yield 1, protein yield 2, ..., protein yield n], and 'spearman' indicates that spearman correlation analysis is selected in the cor function.

16. A method for evaluating protein production of a Fusarium strain, characterized in that: include: Obtain the proteins and their quantities encoded by the genome of the strain to be evaluated; The proteins encoded by the genome of the strain to be evaluated were compared with the marker proteins, and the proteins encoded by the genome whose sequence similarity with the marker proteins reached at least 90%, 95% or 99%, or the proteins encoded by the genome whose sequence identity percentage Percentidentity>70 and sequence coverage Query coverage>70 and E-value E-value<0.05 were screened to obtain similar proteins; The numbers of similar proteins are normalized to obtain the normalized numbers of similar proteins; Taking the correlation coefficient of the marker protein corresponding to the similar protein as a weight, summing the normalized quantities to obtain the evaluation value of the strain; Wherein, the marker protein and its correlation coefficient are obtained by screening through the method described in any one of claims 1 to 8.

17. A method for training a machine learning model for evaluating protein production of Fusarium strains, characterized in that: include: Constructing a machine learning model including a clustering module, a protein yield correlation analysis module, a first alignment module, a scoring module, and a second alignment module that are logically connected in sequence; Input multiple proteins encoded by genomes of multiple Fusarium strains in the training set into a clustering module for clustering to obtain multiple proteomes; The number of each protein group and the proteins contained therein in each Fusarium strain in the training set and the protein yield of the Fusarium strain are transmitted to a protein yield correlation analysis module for correlation analysis and screening to obtain marker proteins and correlation coefficients; The proteins encoded by the genome of the strain in the test set and the marker proteins and the correlation coefficients are transmitted to the first comparison module for comparison to obtain similar proteins and their quantities; Transmitting the similar proteins and their quantities, and the correlation coefficients of the marker proteins corresponding to the similar proteins to a scoring module for evaluation to obtain a strain evaluation value; Inputting the strain evaluation value and protein yield of each strain in the test set into a second comparison module for comparison to obtain an accuracy evaluation result; The accuracy evaluation result is fed back to at least one of the clustering module, the protein yield correlation analysis module and the first comparison module, and the parameters of at least one of the clustering module, the protein yield correlation analysis module and the first comparison module are adjusted until the accuracy evaluation result reaches a preset evaluation threshold.

18. A device for evaluating protein production of Fusarium strains, characterized in that: include: Protein comparison module and strain yield scoring module, The protein comparison module pre-stores marker proteins and correlation coefficients obtained by screening according to any one of claims 1 to 8, and is used to receive the proteins encoded by the genome of the strain to be evaluated and their quantity, traverse and compare the proteins encoded by the genome of the strain to be evaluated with the marker proteins, screen the proteins encoded by the genome with a sequence similarity of at least 90%, 95% or 99%, or the proteins encoded by the genome with a blast comparison result of Percent identity>70 and Query coverage>70 and E-value<0.05, and obtain similar proteins; The strain yield scoring module is used to normalize the number of similar proteins to obtain the normalized number of similar proteins, and to sum the normalized numbers using the correlation coefficients corresponding to the similar proteins as weights to obtain the evaluation value of the strain.

19. A method for screening marker genes related to protein production of Fusarium strains, characterized in that: include: Clustering multiple genes in the genomes of multiple Fusarium strains according to the categories of proteins encoding the same or similar functions to obtain multiple gene families; Corresponding the number of each gene family and the genes contained therein in each Fusarium strain to the protein production of the Fusarium strain to form a data set containing multiple data pairs; Based on the data set, a correlation analysis between the gene family and protein production is performed, and the gene family whose correlation coefficient with the protein production of the Fusarium strain meets a preset threshold is screened as a marker gene.

20. The method according to claim 19, characterized in that Clustering of multiple genes in the genomes of multiple Fusarium strains includes: Gene family analysis was performed between the genes of each Fusarium strain and divided into multiple family categories; The genes are clustered according to the family categories to obtain multiple gene families.

21. The method according to claim 19, characterized in that Based on the data set, the correlation analysis between the gene family and protein production includes: obtaining a plurality of subsets of the data set; Correlation analysis was performed based on each subset, and gene families whose correlation coefficients with the protein production of Fusarium strains were greater than a preset threshold were screened to obtain multiple analysis results; The intersection of multiple analysis results is taken to obtain the marker gene.

22. The method according to claim 19, characterized in that The Fusarium strains include high-yield strains, medium-yield strains and low-yield strains; Among them, the strains with a protein percentage (g) in the product measured by the Kjeldahl method of no more than 50 are low-yield strains, those with a protein percentage (g) between 50 and 55 are medium-yield strains, and those with a protein percentage (g) above 55 are high-yield strains.

23. A method for evaluating protein production of a Fusarium strain, characterized in that: include: Obtain the genome of the strain to be evaluated; The genes in the genome of the strain to be evaluated were compared with the marker genes, and genes with sequence similarity of at least 90%, 95% or 99% with the marker genes were screened, or genes with sequence identity percentage>70, query coverage>70 and E-value<0.05 in blast comparison results were obtained to obtain similar genes; The number of similar genes is normalized to obtain the normalized number of similar genes; Taking the correlation coefficient between the marker gene corresponding to the similar gene and the protein yield as a weight, summing the normalized quantities to obtain the evaluation value of the strain; Wherein, the marker gene is obtained by screening through the method described in any one of claims 19 to 22.

24. A method for training a machine learning model for evaluating protein production of Fusarium strains, characterized in that: include: Constructing a machine learning model including a gene family analysis module, a protein yield correlation analysis module, a third alignment module, an evaluation module, and a fourth alignment module that are logically connected in sequence; The genes of the genome of the strains in the training set are input into the gene family analysis module for clustering to obtain the gene family; The protein yield of each strain in the training set and the number of gene families and genes contained therein are transmitted to the protein yield correlation analysis module for correlation analysis to obtain marker genes and correlation coefficients; The genes in the genome of the strain in the test set and the marker genes and the correlation coefficients are transmitted to the third comparison module for comparison to obtain similar genes and their numbers; The similar genes and their numbers, and the correlation coefficients of the marker genes corresponding to the similar genes are transmitted to an evaluation module for evaluation to obtain a strain score; Inputting the strain score and protein yield of each strain in the test set into the fourth comparison module for comparison to obtain an accuracy evaluation result; The accuracy evaluation result is fed back to at least one of the gene family analysis module, the protein yield correlation analysis module and the third comparison module, and the parameters of at least one of the gene family analysis module, the protein yield correlation analysis module and the third comparison module are adjusted until the accuracy evaluation result reaches a preset evaluation threshold.

25. A device for evaluating protein production of a Fusarium strain, characterized in that: include: Gene comparison module and strain yield evaluation module, The gene comparison module pre-stores marker genes and correlation coefficients obtained by screening according to any one of claims 19 to 22, and is used to receive genes and their numbers of the strain to be evaluated, traverse and compare the genes of the strain to be evaluated with the marker genes, screen genes with sequence similarity of at least 90%, 95% or 99%, or genes with blast comparison results of Percent identity>70, Query coverage>70 and E-value<0.05, to obtain similar genes; The strain yield evaluation module is used to normalize the number of similar genes to obtain the normalized number of similar genes, and to sum the normalized numbers using the correlation coefficient between the marker gene corresponding to the similar gene and the protein yield as a weight to obtain the evaluation value of the strain.

26. A non-volatile computer-readable storage medium having computer-readable instructions stored thereon, which, when executed by one or more processors, implement the method according to any one of claims 1 to 17 or any one of claims 19 to 24.

27. An apparatus comprising: processor; as well as A memory storing program instructions, which, when executed by the processor, enables the device to implement the method according to any one of claims 1 to 17 or the method according to any one of claims 19 to 24.

Images