Cross-queue multi-bound depression diagnosis marker screening method and system and storage medium

Abstract

The invention discloses a cross-queue multi-bound depression diagnosis marker screening method and system and a storage medium, and relates to the technical field of biochemical detection.The method comprises the steps that a multi-bound microorganism database is constructed, and the multi-bound microorganism database comprises bacteria, fungi, viruses and archaea; carrying out metagenome sequencing on feces of cross-queue patients and healthy controls based on the multi-bound microbial database to obtain a cross-queue species annotation abundance table; carrying out microorganism difference abundance analysis on the cross-queue species annotation abundance table by utilizing MMUPHin to obtain a difference species table; performing feature selection on the differential species table by using a Boruta algorithm in Python to obtain differential features; and constructing a prediction model based on the difference characteristics, and performing diagnosis marker screening by using the prediction model. According to the invention, by combining the depression diagnosis marker combination of cross-queue multi-boundary microorganisms, more comprehensive information can be provided for the diagnosis of depression, so that the accuracy and reliability of diagnosis are improved.

Description

Technical Field

[0001] The present invention relates to the field of biochemical detection technology, and in particular to a method, system and storage medium for screening cross-cohort multi-community depression diagnostic markers. Background Art

[0002] Depression is a common mental illness, also known as clinical depression or depressive disorder. It is a persistent mental health problem characterized by persistent sadness, loss of interest or happiness, negative thinking, sleep problems, changes in appetite, fatigue, and possible suicidal thoughts or behaviors. Therefore, the screening of diagnostic markers for depression has become increasingly important; the current screening of diagnostic markers for depression has the following shortcomings:

[0003] (1) Single data dimension: The intestinal microbial community is a highly complex community, which is mostly composed of bacteria and viruses, fungi and archaea. However, at present, most studies and prediction models based on intestinal microbial screening for depression risk markers are still limited to bacterial results, while ignoring other microorganisms such as archaea, viruses and fungi, as well as the interactions between these markers. This single-dimensional data limits the comprehensive understanding of depression biomarkers and causes potential information loss and misunderstanding.

[0004] (2) Lack of cross-cohort marker screening: Existing marker screening results are usually only applicable to specific geographical locations or populations, and do not take into account the possible differences in biomarkers between different regions and ethnic groups. This limitation limits the applicability of the screened markers across populations and regions, and reduces their promotion and application value in clinical practice.

[0005] (3) Low prediction accuracy: Due to over-reliance on a single bacterium as a marker, the prediction accuracy of current markers is limited. The predictive ability of a single bacterium is limited and it is difficult to fully capture the complex biological characteristics of depression. This limitation may limit the effectiveness and reliability of the marker in clinical practice.

[0006] Therefore, there is an urgent need for a depression diagnostic marker screening program that can take into account multiple data dimensions from different regions and ethnic groups. Summary of the invention

[0007] The purpose of the present invention is to provide a method, system and storage medium for screening cross-cohort multi-kingdom depression diagnostic markers, which, combined with a combination of cross-cohort multi-kingdom microbial depression diagnostic markers, can provide more comprehensive information for the diagnosis of depression, thereby improving the accuracy and reliability of the diagnosis.

[0008] To achieve the above object, the present invention provides the following solutions:

[0009] A method for screening cross-cohort multi-community depression diagnostic markers, comprising the following steps:

[0010] S1. Construct a multi-kingdom microbial database, wherein the multi-kingdom microbial database includes a bacterial database, a fungal database, a viral database, and an archaea database;

[0011] S2. Performing metagenomic sequencing on the feces of patients and healthy controls across cohorts based on the multi-kingdom microbial database to obtain an annotated species abundance table across cohorts;

[0012] S3, using MMUPHin to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts to obtain a differential species table;

[0013] S4, using the Boruta algorithm in Python to perform feature selection on the differential species table to obtain the minimum differential feature group;

[0014] S5. Constructing a depression diagnostic marker prediction model based on the minimum difference feature group, and using the depression diagnostic marker prediction model to screen depression diagnostic markers.

[0015] Furthermore, the S2, performing metagenomic sequencing on the feces of patients across cohorts and healthy controls based on the multi-kingdom microbial database to obtain a species annotation abundance table across cohorts, specifically comprises the following steps:

[0016] S2.1. Use KneadData to remove low-quality sequences and contaminants from host-related sequences and laboratory-related sequences through bowtie2 to obtain clean data.

[0017] S2.2, using Kraken to compare the clean data with the bacteria database, the fungus database, the virus database, and the archaea database, respectively, and annotate the bacteria, fungi, viruses, and archaea with kraken2 to obtain a species annotation abundance table for the first cross-cohort;

[0018] S2.3. Re-estimate the species annotation abundance table of the first cross-cohort through Braken, and calculate its classification abundance to obtain the final species annotation abundance table of the cross-cohort.

[0019] Further, the S3, using MMUPHin to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts to obtain a differential species table, specifically includes the following steps:

[0020] Parameters were set in MMUPHin to screen differential bacteria whose difference test P value and FDR-corrected P value were less than fixed values, and a differential species table was obtained;

[0021] PERMANOVA was used to eliminate the batch effect during the difference analysis. If the PERMANOVA analysis result was P < 0.05, gender and age were selected as covariates.

[0022] Further, the S4, using the Boruta algorithm in Python to perform feature selection on the difference species table to obtain the minimum difference feature group, specifically includes the following steps:

[0023] Use Boruta in Python to select the difference features and obtain the difference features;

[0024] Perform collinearity feature elimination on the difference features, eliminate the collinearity problem caused by highly correlated features, and return irrelevant valid features;

[0025] The weakest features in each cycle are eliminated recursively to screen out the smallest set of discriminative features.

[0026] Furthermore, the S5, constructing a depression diagnostic marker prediction model based on the minimum difference feature group, also includes: using the BayesianOptimization software package in Python to adjust hyperparameters.

[0027] The present invention also provides a cross-cohort multi-realm depression diagnostic marker screening system, which is used to perform the cross-cohort multi-realm depression diagnostic marker screening method, comprising:

[0028] A database module, used to construct a multi-kingdom microbial database, wherein the multi-kingdom microbial database includes bacteria, fungi, viruses and archaea;

[0029] A sequencing module, used for performing metagenomic sequencing on feces of patients across cohorts and healthy controls based on the multi-kingdom microbial database to obtain a species annotation abundance table across cohorts;

[0030] A differential abundance analysis module, used to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts using MMUPHin to obtain a differential species table;

[0031] A feature selection module is used to perform feature selection on the differential species table using the Boruta algorithm in Python to obtain the minimum differential feature group;

[0032] A model building module is used to build a prediction model based on the differential features and use the prediction model to screen diagnostic markers.

[0033] The present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is used by a processor to implement the cross-cohort multi-community depression diagnostic marker screening method.

[0034] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the cross-cohort multi-kingdom depression diagnostic marker screening method and system provided by the present invention, by integrating the information of multi-kingdom microorganisms across cohorts to construct a depression diagnostic marker prediction model, thereby improving the prediction accuracy of depression risk; the present invention combines a combination of depression diagnostic markers of cross-cohort multi-kingdom microorganisms to provide more comprehensive information for the diagnosis of depression, further improving the accuracy and reliability of the diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0036] Figure 1 This is a flow chart of the cross-cohort multi-community depression diagnostic marker combination and screening method described in an embodiment of the present invention;

[0037] Figure 2 This is a structural diagram of a cross-cohort multi-community depression diagnostic marker combination and screening system according to an embodiment of the present invention;

[0038] Figure 3 It is a volcano plot of the difference analysis in Example 2 of the present invention;

[0039] Figure 4 This is a box plot of the importance of differentially characteristic microorganisms in Example 2 of the present invention. DETAILED DESCRIPTION

[0040] 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 only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] The purpose of the present invention is to provide a method, system and storage medium for screening cross-cohort multi-kingdom depression diagnostic markers, which, combined with a combination of cross-cohort multi-kingdom microbial depression diagnostic markers, can provide more comprehensive information for the diagnosis of depression, thereby improving the accuracy and reliability of the diagnosis.

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] like Figure 1 As shown, the cross-cohort multi-community depression diagnostic marker screening method provided by the embodiment of the present invention comprises the following steps:

[0044] S1. Construct a multi-kingdom microbial database, wherein the multi-kingdom microbial database includes a bacterial database, a fungal database, a viral database, and an archaea database;

[0045] S2. Performing metagenomic sequencing on the feces of patients and healthy controls across cohorts based on the multi-kingdom microbial database to obtain an annotated species abundance table across cohorts;

[0046] S3, using MMUPHin to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts to obtain a differential species table;

[0047] S4, using the Boruta algorithm in Python to perform feature selection on the differential species table to obtain the minimum differential feature group;

[0048] S5. Constructing a depression diagnostic marker prediction model based on the minimum difference feature group, and using the depression diagnostic marker prediction model to screen depression diagnostic markers.

[0049] In this embodiment, the S1, constructing a multi-kingdom microbial database, the multi-kingdom microbial database includes a bacterial database, a fungal database, a viral database and an archaea database, specifically including:

[0050] Build a bacterial database: Use the kraken2-build tool to build it, including 50 phylum-level bacteria, 190 class-level bacteria, 340 order-level bacteria, 736 family-level bacteria, 2031 genus-level bacteria, 13427 species-level bacteria, and the database size is 401G. Use the bracken-build tool to generate the required k-mer distribution files (including 50mers) for subsequent calculation of species abundance;

[0051] Constructing the Archaea database: using the kraken2-build tool, including 13 phylum-level bacteria, 24 class-level bacteria, 45 order-level bacteria, 70 family-level bacteria, 213 genus-level bacteria, 531 species-level bacteria, the database size is 65G, and the bracken-build tool is used to generate the required k-mer distribution file for subsequent calculation of species abundance;

[0052] Construct fungal database: Use kraken2-build tool to build, including 5 phylum-level fungi, 18 class-level fungi, 23 order-level fungi, 45 family-level fungi, 61 genus-level fungi, 103 species-level fungi, the database size is 53G, use bracken-build tool to generate the required k-mer distribution file for subsequent calculation of species abundance;

[0053] Construct a virus database: Use the kraken2-build tool to build it, including 20 phylum-level bacteria, 43 class-level bacteria, 87 order-level bacteria, 294 family-level bacteria, 3193 genus-level bacteria, 17845 species-level bacteria, and the database size is 4G. Use the bracken-build tool to generate the required k-mer distribution file for subsequent calculation of species abundance.

[0054] In this embodiment, S2, performing metagenomic sequencing on feces of patients across cohorts and healthy controls based on the multi-kingdom microbial database to obtain a species annotation abundance table across cohorts, specifically includes the following steps:

[0055] S2.1. Use KneadData to remove low-quality sequences and contaminants from host-related sequences and laboratory-related sequences through bowtie2 to obtain clean data.

[0056] S2.2, using Kraken to compare the clean data with the constructed bacterial, fungal, viral and archaea databases respectively, and annotate the bacteria, fungi, viruses and archaea with kraken2 to obtain the species annotation abundance table of the first cross-cohort;

[0057] S2.3. Re-estimate the species annotation abundance table of the first cross-cohort through Braken, and calculate its classification abundance to obtain the final species annotation abundance table of the cross-cohort.

[0058] In this embodiment, S3, using MMUPHin to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts to obtain a differential species table, specifically includes the following steps:

[0059] Parameters were set in MMUPHin to screen differential bacteria whose difference test P value and FDR-corrected P value were less than fixed values, and a differential species table was obtained; the P value of significant difference results was set to 0.05, and the FDR-corrected P value was set to 0.1;

[0060] PERMANOVA was used to eliminate the batch effect during the difference analysis. If the PERMANOVA analysis result was P < 0.05, gender and age were selected as covariates.

[0061] In this embodiment, the S4, using the Boruta algorithm in Python to perform feature selection on the difference species table to obtain the minimum difference feature, specifically includes the following steps:

[0062] S4.1. Select differential features using Boruta in Python.

[0063] Among them, the Boruta algorithm process:

[0064] (1) Randomly shuffle the feature values ​​of the original feature matrix X to create "shadow features" that have the same distribution characteristics as the original features (real features) but have no practical significance. The shuffled shadow features are concatenated with the real features to form a new feature matrix.

[0065] (2) Use the new feature matrix as input to the random forest model;

[0066] (3) Calculate the importance score Zscore of the real feature and shadow feature;

[0067] (4) Find the maximum Zscore among the shadow features, denoted as Zmax;

[0068] (5) Mark the real feature with a Zscore better than Zmax as "important", mark the real feature with a Zscore less than Zmax as "unimportant", and permanently remove the real feature marked as "unimportant" from the feature set;

[0069] (6) Delete all shadow features;

[0070] (7) Repeat 1 to 6 until all features are marked as "important" or "unimportant", and the iteration stops.

[0071] Taking the bacterial kingdom as an example, the results of differential characteristic microorganisms are shown in Table 1, where the first column is the differential characteristic microorganisms and the second column is the importance of the differential characteristic microorganisms.

[0072] Table 1 Results of differential characteristic microorganisms

[0073]

[0074] S4.2. Perform collinearity elimination on the difference features, eliminate the collinearity problem caused by highly correlated features, and return irrelevant valid features. In this embodiment, the correlation coefficient is calculated using Spearman, and features with high correlation coefficients (higher than 0.8) are considered to be collinear.

[0075] S4.3. Recursively eliminate the weakest feature in each cycle to screen out the smallest difference feature group, that is, the optimal feature group.

[0076] In this embodiment, the S5, constructing a depression diagnostic marker prediction model based on the minimum difference feature group, also includes:

[0077] S5.1. Use the BayesianOptimization package in Python to adjust the hyperparameters and obtain the optimal hyperparameters.

[0078] S5.2. Construct a depression diagnostic marker prediction model based on the optimal feature group and optimal hyperparameters.

[0079] Result file: The output file param.txt stores the various indicators of each repeated training model evaluation, as shown in Table 2:

[0080] Table 2 Various indicators for each repeated training model evaluation

[0081]

[0082] The results are explained as follows:

[0083] iter: indicates the number of iterations in the optimization process.

[0084] target: represents the optimized objective function value.

[0085] learning_rate: The learning rate is used to control the magnitude of each update step.

[0086] max_depth: The maximum depth of the decision tree, which limits the maximum depth of the tree to prevent overfitting.

[0087] max_features: Maximum number of features, the maximum number of features considered when splitting nodes.

[0088] n_estimators: The number of weak learners, that is, the number of iterations.

[0089] subsample: Subsampling ratio, which is the ratio of samples randomly drawn when building each tree.

[0090] Example 2

[0091] This example is a specific example of the cross-cohort multi-community depression diagnostic marker screening method described in Example 1. In this example, three groups of cross-cohort (CNP0001162, PRJNA762199, PRJNA1083304) metagenomic data of 425 people were obtained, including 227 patients with depression and 198 healthy controls. The specific steps are as follows:

[0092] Step 1: Metagenomic sequencing of feces of 425 samples: KneadData was used to remove low-quality reads and contaminant reads, including host-related and laboratory-related sequences, through bowtie2; kraken2 was used to annotate bacteria + archaea + viruses + fungi, and Braken was used to re-estimate the abundance, finally obtaining a species annotation abundance table of bacteria + archaea + viruses + fungi across cohorts.

[0093] Step 2, use MMUPHin to perform differential abundance analysis on the species abundance table at the species level of the four kingdoms: age and gender are used as grouping variables for the random effects of the model. Finally, the difference table of the four kingdoms is obtained, and P<0.05 and FDR-corrected P<0.1 are used as screening criteria to obtain the final differential bacteria. Among them, there are 143 differential bacteria in bacteria, 45 differential bacteria in archaea, 108 differential bacteria in viruses, and 2 differential bacteria in fungi. Taking the bacterial kingdom as an example, the differential analysis volcano plot is as follows Figure 3 shown.

[0094] Step 3, combine the four kingdoms of microorganisms obtained in the previous step in groups of two, three, or four, and finally obtain 15 tables of one, two, three, or four kingdoms:

[0095] Kingdom 1: (1) Bacteria, (2) Viruses, (3) Fungi, (4) Archaea;

[0096] Kingdom II: (5) Bacteria + Viruses; (6) Bacteria + Fungi, (7) Bacteria + Archaea, (8) Viruses + Fungi, (9) Viruses + Archaea, (10) Fungi + Archaea;

[0097] Three kingdoms: (11) Bacteria + Fungi + Viruses, (12) Bacteria + Fungi + Archaea, (13) Bacteria + Archaea + Viruses (14) Viruses + Fungi + Archaea;

[0098] Four kingdoms: (15) Bacteria + Fungi + Viruses + Archaea.

[0099] Step 4: Use the Boruta algorithm in Python to select features for the above 15 difference tables. After the selected difference features are eliminated through collinearity elimination and recursive feature elimination, the optimal features are used as candidate difference features. Taking the bacterial kingdom as an example, the importance box plot of the difference feature microorganisms is shown in Figure 4 shown.

[0100] Step 5: Use the above differential features to build a prediction model and use the BayesianOptimization package in Python to adjust the hyperparameters. Taking the bacterial kingdom as an example, the optimal hyperparameter combination is shown in Table 3.

[0101] Table 3 Optimal hyperparameter combinations

[0102]

[0103]

[0104] After summarizing 15 different boundary combinations, AUROC was calculated using the Python package sklearn to evaluate the performance of the classification model.

[0105] Table 4 Performance evaluation of different microbial combination prediction models

[0106]

[0107] From Table 4, we can see that the AUC values ​​of the mental illness diagnosis model established by merging one by one, two by two, three by three, and four by four groups are:

[0108] Kingdom: (1) Bacteria B (0.765), (2) Viruses V (0.833), (3) Fungi E (0.658), (4) Archaea A (0.708); the average value of the four different kingdoms is 0.741.

[0109] Kingdom 2: (5) Bacteria + Viruses BV (0.908); (6) Bacteria + Fungi EB (0.792), (7) Bacteria + Archaea AB (0.826), (8) Viruses + Fungi EV (0.846), (9) Viruses + Archaea AV (0.869), (10) Fungi + Archaea AE (0.710); the average AUC after merging the six kingdoms was 0.825.

[0110] The three kingdoms were (11) bacteria + fungi + viruses EBV (0.919), (12) bacteria + fungi + archaea AEB (0.731), (13) bacteria + archaea + viruses ABV (0.927), and (14) viruses + fungi + archaea AEV (0.863); the average AUC after merging the four kingdoms was 0.86.

[0111] Four kingdoms (15): Bacteria + Fungi + Viruses + Archaea AEBV (0.934).

[0112] Obviously, the four-boundary AUC (0.934) > the three-boundary AUC (0.86) > the two-boundary AUC (0.825) > the one-boundary AUC (0.741). Therefore, the combination of depression diagnostic markers across cohorts and multi-kingdom microorganisms can provide more comprehensive information for the diagnosis of depression and improve the accuracy and reliability of diagnosis.

[0113] In order to explore the breadth of applicability of the method of the present invention, the published metagenomic data of depression in different regions (PRJNA860656, PRJNA634666) were downloaded from public databases to verify the cross-cohort multi-kingdom depression diagnostic marker combination kingdom screening method. The analysis results showed that after combining 15 different kingdoms, the four-kingdom microbial model showed better predictive ability compared with the other kingdoms. Therefore, the combination of depression diagnostic markers combined with cross-cohort multi-kingdom microorganisms can provide more comprehensive information for the diagnosis of depression, and may improve the accuracy and reliability of diagnosis.

[0114] Example 3

[0115] like Figure 2 As shown, the embodiment of the present invention provides a cross-cohort multi-realm depression diagnostic marker screening system, which is used to execute the cross-cohort multi-realm depression diagnostic marker screening method described in Example 1, including:

[0116] A database module, used to construct a multi-kingdom microbial database, wherein the multi-kingdom microbial database includes a bacterial database, a fungal database, a viral database, and an archaea database;

[0117] A sequencing module, used for performing metagenomic sequencing on feces of patients across cohorts and healthy controls based on the multi-kingdom microbial database to obtain a species annotation abundance table across cohorts;

[0118] A difference analysis module, used for performing microbial difference abundance analysis on the species annotation abundance table across the cohorts using MMUPHin, so as to obtain a difference species table;

[0119] A feature selection module is used to perform feature selection on the differential species table using the Boruta algorithm in Python to obtain the minimum differential feature group;

[0120] A model building module is used to build a depression diagnostic marker prediction model based on the minimum difference feature group, and use the depression diagnostic marker prediction model to screen depression diagnostic markers.

[0121] In a further embodiment, the database module, the sequencing module, the difference analysis module, the feature selection module and the model building module are connected in sequence.

[0122] Example 4

[0123] An embodiment of the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is used by a processor to implement the steps of the cross-cohort multi-community depression diagnostic marker screening method described in Example 1.

[0124] In summary, the cross-cohort multi-kingdom depression diagnostic marker combination and screening method and system provided by the present invention have constructed a multi-kingdom microbial database including bacteria, fungi, viruses and archaea. The depression prediction model established on this basis has a more accurate prediction effect. It can be seen that the combination of depression diagnostic markers combined with cross-cohort multi-kingdom microorganisms can provide more comprehensive information for the diagnosis of depression, and may improve the accuracy and reliability of the diagnosis.

[0125] The remaining technical features in this embodiment can be flexibly selected by those skilled in the art according to actual conditions to meet different specific practical needs. However, it is obvious to those skilled in the art that it is not necessary to adopt these specific details to implement the present invention. In other examples, in order to avoid confusing the present invention, the well-known components, structures or parts are not specifically described, which are all within the technical protection scope defined by the technical solution claimed for protection in the claims of the present invention.

[0126] Modifications and changes made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the scope of protection of the claims attached to the present invention. In the above description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details are not necessary to practice the present invention. In other examples, in order to avoid confusing the present invention, well-known technologies, such as specific construction details, operating conditions and other technical conditions, are not specifically described.

[0127] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for screening cross-cohort multi-community depression diagnostic markers, characterized in that: The following steps are involved: S1. Construct a multi-kingdom microbial database, wherein the multi-kingdom microbial database includes a bacterial database, a fungal database, a viral database, and an archaea database; S2. Performing metagenomic sequencing on the feces of patients and healthy controls across cohorts based on the multi-kingdom microbial database to obtain an annotated species abundance table across cohorts; S3, using MMUPHin to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts to obtain a differential species table; S4, using the Boruta algorithm in Python to perform feature selection on the differential species table to obtain the minimum differential feature group; S5. Constructing a depression diagnostic marker prediction model based on the minimum difference feature group, and using the depression diagnostic marker prediction model to screen depression diagnostic markers.

2. The cross-cohort multi-community depression diagnostic marker screening method according to claim 1, characterized in that: S2, performing metagenomic sequencing on feces of patients across cohorts and healthy controls based on the multi-kingdom microbial database to obtain a species annotation abundance table across cohorts, specifically comprises the following steps: S2.

1. Use KneadData to remove low-quality sequences and contaminants from host-related sequences and laboratory-related sequences through bowtie2 to obtain clean data. S2.2, using Kraken to compare the clean data with the bacteria database, the fungus database, the virus database, and the archaea database, respectively, and annotate the bacteria, fungi, viruses, and archaea with kraken2 to obtain a species annotation abundance table for the first cross-cohort; S2.

3. Re-estimate the species annotation abundance table of the first cross-cohort through Braken, and calculate its classification abundance to obtain the final species annotation abundance table of the cross-cohort.

3. The cross-cohort multi-community depression diagnostic marker screening method according to claim 1, characterized in that: S3, using MMUPHin to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts to obtain a differential species table, specifically includes the following steps: Parameters were set in MMUPHin to screen differential bacteria whose difference test P value and FDR-corrected P value were less than fixed values, and a differential species table was obtained; PERMANOVA was used to eliminate the batch effect during the difference analysis. If the PERMANOVA analysis result was P < 0.05, gender and age were selected as covariates.

4. The cross-cohort multi-community depression diagnostic marker screening method according to claim 1, characterized in that: S4, using the Boruta algorithm in Python to perform feature selection on the difference species table to obtain the minimum difference feature group, specifically includes the following steps: Use Boruta in Python to select the difference features and obtain the difference features; Perform collinearity feature elimination on the difference features, eliminate the collinearity problem caused by highly correlated features, and return irrelevant valid features; The weakest features in each cycle are eliminated recursively to screen out the smallest set of discriminative features.

5. The cross-cohort multi-community depression diagnostic marker screening method according to claim 1, characterized in that: The S5, constructing a depression diagnostic marker prediction model based on the minimum difference feature set, also includes: using the BayesianOptimization software package in Python to adjust hyperparameters.

6. A cross-cohort multi-community depression diagnostic marker screening system, used to perform the cross-cohort multi-community depression diagnostic marker screening method according to any one of claims 1 to 5, characterized in that: include: A database module, used to construct a multi-kingdom microbial database, wherein the multi-kingdom microbial database includes bacteria, fungi, viruses and archaea; A sequencing module, used for performing metagenomic sequencing on feces of patients across cohorts and healthy controls based on the multi-kingdom microbial database to obtain a species annotation abundance table across cohorts; A differential abundance analysis module, used to perform microbial differential abundance analysis on the species annotation abundance table across the cohorts using MMUPHin to obtain a differential species table; A feature selection module uses the Boruta algorithm in Python to perform feature selection on the differential species table to obtain the minimum differential feature group; A model building module is used to build a prediction model based on the differential features and use the prediction model to screen diagnostic markers.

7. A computer storage medium, characterized in that: The computer storage medium stores a computer program, and the computer program is used by the processor to implement the cross-cohort multi-community depression diagnostic marker screening method as described in any one of claims 1-5.

Images