Alzheimer's disease biomarker identification method based on microbiome-host interaction
By combining the integrated analysis of microbiomics and host genomics, the identification of microbial markers of Alzheimer's disease has been solved, and the problem of poor early diagnosis and prediction model performance in the existing technology has been achieved, and the problem of Alzheimer's disease has been achieved. the improvement of effective marker identification and prediction model.
Patent Information
- Application Number
- CN202310098268.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-02-10
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2043-02-10
AI Technical Summary
The prior art is difficult to effectively identify early markers of Alzheimer's disease, especially when there is no symptoms, and the predictive model constructed by metagenomic data alone is not ideal, especially in cross-cohort cross-validation.
By combining the integrated analysis of microbiomics and host genomics, patients' feces and blood samples were collected, the microbial genome and host genome were extracted, and the subject model was used to identify the association of sub-society structures and host gene mutation sites, and Mendel randomization analysis was performed. Identify the causal relationship between microbial characteristics and Alzheimer's disease and build a disease prediction model.
It has achieved the ability to generalize the Alzheimer's phenotype prediction model in microbiome-host genome association research, revealed the interaction mode between microorganisms and hosts, excavated the microbial disease markers of Alzheimer's disease, and constructed A disease prediction model with generalization ability.
Smart Images

Figure CN116052767B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical testing technology, and in particular to a method for identifying Alzheimer's disease markers based on the interaction between microorganisms and hosts. Background Art
[0002] To date, early diagnosis of Alzheimer's disease (AD) remains a major problem. Even though various pathological changes have been shown several years before symptoms, such as excessive production and accumulation of β-amyloid protein in the brain and hyperphosphorylation of tau protein, the detection of these biomarkers requires positron emission tomography or lumbar puncture. When there are no or only mild symptoms, few people can undergo these radioactive or invasive examinations. In addition, some genetic factors may also aggravate the onset of AD, such as the APOEε4 allele, which is recognized as an important risk gene for AD. Therefore, host gene-related research identifies more new risk gene loci, further reveals the pathogenesis of AD, and provides new directions for treatment, which has great clinical application value.
[0003] Studies have shown that the characteristics of intestinal microbial imbalance in AD patients are an increase in pro-inflammatory bacteria and a decrease in anti-inflammatory bacteria, and their microbial-derived metabolites jointly affect intestinal permeability and blood-brain barrier integrity. For example, Firmicutes are related to the regulation of inflammatory response and metabolic function, which in turn affects behavioral and cognitive functions. However, most of the AD prediction models currently constructed solely based on metagenomic data are not ideal, especially for cross-cohort cross-validation. Therefore, how to obtain more consistent AD markers between cohorts is particularly important and a major challenge.
[0004] In the case-control cohort study of Alzheimer's disease (AD), there have been many studies on the association between microbiome and host genome and AD. At present, the exploration of the interaction pattern between microorganisms and hosts mainly involves the joint analysis of microbiome and host genome, but most of the existing studies are based on healthy cohorts, and then explore the association between microbial characteristics (mainly species composition and functional pathways) and host immune and metabolic related traits. Few studies have included case-control cohorts at the same time. Summary of the invention
[0005] The purpose of the present invention is to provide a method for identifying Alzheimer's disease markers based on the interaction between microorganisms and hosts, and to solve the following technical problems:
[0006] How to utilize prior information and combine integrative analysis of microbiome and host genomics to jointly reveal disease markers of AD.
[0007] The purpose of the present invention can be achieved through the following technical solutions:
[0008] The method for identifying Alzheimer's disease markers based on the interaction between microorganisms and hosts includes the following steps:
[0009] Fecal and blood samples were collected from patients, and the microbial genome and host genome were extracted and quality checked respectively. The microbial-host genome paired samples that passed the quality check were marked as the discovery set;
[0010] Use topic models to identify subcommunity structures in microbial genomes that are associated with Alzheimer's disease phenotypes, associate the subcommunity structures with host genomes, and perform gene annotation and enrichment analysis on significantly associated host gene variation sites;
[0011] Using significantly associated gene variant sites as instrumental variables, single-sample Mendelian randomization and two-sample Mendelian randomization were performed on the subcommunity structure to identify the causal relationship between the subcommunity structure characteristics and the Alzheimer's disease phenotype and the phenotype of existing human diseases;
[0012] SNPs covered by risk genes associated with Alzheimer's disease were collected, and the SNPs were associated with the abundance of microbial bacteria. The microbial species with significant differences between the disease and healthy groups were extracted as the differential species set. The candidate key microbial species set significantly associated with the disease risk genes was tested to see whether it was enriched in the differential species of Alzheimer's disease. The microbial features that passed the test were marked as Alzheimer's disease markers.
[0013] As a further solution of the present invention: the process of quality inspection of the microbial genome is:
[0014] Metagenomic sequencing samples of microorganisms in the fecal samples were extracted, and the original gene sequences of the metagenomic sequencing samples were quality checked using Trimmomatic and Bowtie2, and species annotation was performed using MetaPhlAn3 to obtain a relative abundance table from kingdom to species level, and the metagenomic sequencing samples that passed the quality inspection were marked as microbial genomes.
[0015] As a further solution of the present invention: the process of quality inspection of the host genome is:
[0016] Whole genome sequencing data from blood samples were extracted and compared with reference gene sequences to obtain SNP variation information. The samples and SNPs were quality-checked with the help of plink tools, and valid SNPs that met preset conditions were marked as host genomes.
[0017] As a further solution of the present invention: the preset conditions for SNP are:
[0018] The missing rate was less than 2%, the minimum allele frequency MAF was greater than or equal to 5%, and the P value of HWE was less than 1*10 -6 SNP variation information.
[0019] As a further solution of the present invention: the two-sample Mendelian random inference of subcommunity structure that may be associated with phenotypes is based on the GWAS database of Biobank Japan.
[0020] As a further solution of the present invention: the process of determining the causal relationship between the sub-community structure and the disease phenotype is:
[0021] Using the existing AD risk gene variant sites, single-sample Mendelian randomization was performed on the subcommunity structure to infer the causal relationship between the subcommunity structure characteristics and AD; GWAS summary data from Biobank Japan was collected, and two-sample Mendelian randomization was used to infer the causal relationship between the subcommunity structure and the existing disease phenotype.
[0022] As a further solution of the present invention: single-sample Mendelian randomization is directly implemented using two-stage logistic regression, and two-sample Mendelian randomization is implemented through the standard process of TwoSampleMR R package.
[0023] As a further solution of the present invention: before testing the candidate risk SNP-associated species enrichment analysis, the LinDAR package was used to perform inter-group difference analysis on single bacteria, and species with p < 0.05 and |log2(Foldchange)| > 1 in each subgroup were selected as differences; in the construction of the disease prediction model, a nested random forest classifier with 5 folds was used three times, and the model AUC and F1score were used as classification accuracy indicators.
[0024] Beneficial effects of the present invention:
[0025] The present invention demonstrates how to use multi-omics information to mine disease-associated signals, and for the first time uses the concept of subcommunity structure in microbiome-host genome association studies. It can not only improve the generalization ability of AD phenotype prediction models using fewer species characteristics, but also verify causal relationships by combining genomes, integrate multi-omics data of microbiome-host genomes, and mine microbial disease markers for AD, which have both microbial significance and can be explained from the perspective of host genetics, and can construct disease prediction models with better generalization capabilities; it compares two ideas and can expand the exploration of the contribution of multi-kingdom (fungi-bacteria, etc.) characteristics to the accuracy of disease prediction models for AD cohorts; the causal relationship between the subcommunity structure mined by the topic model and the disease phenotype can be inferred through host genomics data. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] The present invention will be further described below in conjunction with the accompanying drawings.
[0027] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION
[0028] 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.
[0029] See also Figure 1 As shown, the present invention is a method for identifying Alzheimer's disease markers based on the interaction between microorganisms and hosts, comprising the following steps:
[0030] S1. Microbiome data quality inspection: After quality inspection of metagenomic sequencing samples collected from feces, species annotation is performed.
[0031] S2. Host genome data quality control: The whole genome sequencing data of blood samples is compared with the reference gene sequence to obtain SNP variation information, which requires quality control before association analysis.
[0032] S3. Discovery set and test set: The microbiome-host genome matching samples that have passed quality inspection are used as training sets for downstream analysis. The high-quality AD cohort public data from NCBI is used as a validation set to characterize the generalization ability of the disease prediction models trained by the two schemes.
[0033] S4. From the perspective of microbial phenotype: The abundance data at the bacterial genus level were used using topic models to mine subcommunity structures that were significantly associated with the AD phenotype.
[0034] S5. Interpretation of subcommunity characteristics in genetic variation: The subcommunity structure screened out in S4 was associated with the host genomics data, and the significantly associated SNP set was further annotated for gene enrichment analysis to reveal the potential pathogenesis of AD.
[0035] S6. Causal inference between sub-community characteristics and phenotypes: Use existing data to perform single-sample Mendelian randomization to infer the causal relationship between this structural feature and AD. At the same time, collect GWAS summary data from Biobank Japan of the same ethnic group, use two-sample Mendelian randomization to infer the causal relationship between this community structure and other phenotypes, and explore similar microbial change structures between different diseases and traits.
[0036] S7. From the perspective of gene-phenotype: conduct association analysis between the SNPs covered by the candidate risk gene set associated with AD and single bacteria, extract significantly associated microbial features for inter-group difference analysis, and test whether the associated feature set is significantly enriched in the differential species of AD. Finally, the feature set is used to build a disease prediction model.
[0037] S8. The fungal data set will repeat the analysis process of S4-S7 to compare the relative contribution of fungi and bacteria to the construction of disease prediction models. At the same time, the present invention uses the association with risk genes to construct a bacteria-fungus interaction network and compares the impact of combining the two types of data on the accuracy of the prediction model.
[0038] S9. Selection of the optimal prediction model feature set: Compare the classification accuracy of the microbial feature sets obtained from the two starting points for model construction, and combine the classification accuracy on the independent validation set to select the optimal model and further explain the features with higher contribution.
[0039] Specifically, an AD prediction model was established using the microbial abundance involved in the subcommunity structure and externally verified on AD public data. Similarly, a disease prediction model was established using a set of candidate key microbial species that were significantly associated with AD risk genes, and the prediction accuracy and generalization ability of the two were compared.
[0040] In a preferred case of this embodiment, in step S1, the original sequence of the metagenomic sequencing sample is filtered using Trimmomatic and Bowtie2, and species annotation is performed using MetaPhlAn3 to obtain a relative abundance table from kingdom to species level.
[0041] In another preferred embodiment of the present invention, in step S2, the vcf file storing the variants after aligning the reference gene sequence is used, and the samples and SNPs are quality checked with the help of the plink tool, and finally the SNPs that meet the conditions are retained for downstream association analysis: (1) missing rate <2%; (2) MAF ≥ 5%; (3) Hardy-Weinberg equilibrium P-values > 1e-06.
[0042] In another preferred embodiment of the present invention, in step S4, the Latent Dirichlet Allocation (LDA) topic model is used to identify the sub-community structure, and Dirichlet regression is used to perform difference analysis between component data groups.
[0043] In another preferred situation of the present embodiment, in step S5, plink associates the community component data significantly associated with the AD phenotype with the host SNP, and position mapping of the significantly associated SNP (p<1e-05) covers the gene in its 250kb window, and performs enrichment analysis containing GWAS catalog geneset, GO, and KEGG with the help of FUMA online platform.
[0044] In another preferred embodiment of the present invention, in step S6, the single-sample Mendelian randomization is directly implemented using two-stage logistic regression, and the two-sample Mendelian randomization is implemented using the standard process of the TwoSampleMR R package.
[0045] In another preferred case of this embodiment, in step S7, starting from the candidate risk genes associated with AD collected from existing literature, plink expands the SNPs covered by the 20kb window in its gene region and associates them with the abundance of single microorganisms. Before checking the candidate SNP-associated species enrichment analysis, the LinDA R package was used to perform inter-group difference analysis on single bacteria, and species with p<0.05 and |log2(FoldChange)|>1 in each subgroup were selected as differences. In the construction of the disease prediction model, a nested random forest classifier with 5 folds was used three times, and the model AUC and F1 score were used as classification accuracy indicators.
[0046] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.
Claims
1. A method for identifying Alzheimer's disease markers based on the interaction between microorganisms and hosts, characterized in that: The following steps are involved: Fecal and blood samples were collected from patients, and the microbial genome and host genome were extracted and quality checked respectively. The microbial-host genome paired samples that passed the quality check were marked as the discovery set; Use topic models to identify subcommunity structures in microbial genomes that are associated with Alzheimer's disease phenotypes, associate the subcommunity structures with host genomes, and perform gene annotation and enrichment analysis on significantly associated host gene variation sites; Using significantly associated gene variant sites as instrumental variables, single-sample Mendelian randomization and two-sample Mendelian randomization were performed on the subcommunity structure to identify the causal relationship between the subcommunity structure characteristics and the Alzheimer's disease phenotype and the phenotype of existing human diseases; Collect SNPs covered by risk genes associated with Alzheimer's disease, conduct association analysis between the SNPs and the abundance of microbial single bacteria, and extract microbial species with significant differences between the disease and healthy groups as differential species sets. Test whether the candidate key microbial species set significantly associated with disease risk genes is enriched in the differential species of Alzheimer's disease, and mark the microbial features that pass the test as Alzheimer's disease markers; The process of quality inspection of the microbial genome is as follows: Extracting metagenomic sequencing samples of microorganisms in the fecal samples, quality checking the original gene sequences of the metagenomic sequencing samples using Trimmomatic and Bowtie2, and annotating species using MetaPhlAn3 to obtain a relative abundance table from kingdom to species level, and marking the metagenomic sequencing samples that passed the quality inspection as microbial genomes; The process of quality inspection of the host genome is as follows: Extract whole genome sequencing data from blood samples, compare the whole genome sequencing data with the reference gene sequence to obtain SNP variation information, use the plink tool to perform quality inspection on samples and SNPs, and mark valid SNPs that meet the preset conditions as host genomes; The process of determining the causal relationship between the subcommunity structure and the disease phenotype is: Using the existing AD risk gene variant sites, single-sample Mendelian randomization was performed on the subcommunity structure to infer the causal relationship between the subcommunity structure characteristics and AD; GWAS summary data from Biobank Japan was collected, and two-sample Mendelian randomization was used to infer the causal relationship between the subcommunity structure and the existing disease phenotype.
2. The method for identifying Alzheimer's disease markers based on microbial-host interaction according to claim 1, characterized in that: The default conditions for SNP are: Select SNP variation information that meets the conditions of missing rate lower than 2%, minimum allele frequency MAF greater than or equal to 5%, and HWE P value less than 1*10-6.
3. The method for identifying Alzheimer's disease markers based on microbial-host interaction according to claim 1, characterized in that: The two-sample Mendelian randomization inferred subcommunity structures that may be associated with phenotypes is based on the GWAS database of Biobank Japan.
4. The method for identifying Alzheimer's disease markers based on microbial-host interaction according to claim 1, characterized in that: One-sample Mendelian randomization was directly implemented using two-stage logistic regression, and two-sample Mendelian randomization was implemented using the standard process of the TwoSampleMR R package.
5. The method for identifying Alzheimer's disease markers based on microbial-host interaction according to claim 1, characterized in that: Before testing the candidate risk SNP-associated species enrichment analysis, the LinDA R package was used to perform inter-group difference analysis on single bacteria, and species with p < 0.05 and |log2(Foldchange)| > 1 in each subgroup were selected as differentials; in the construction of the disease prediction model, a nested random forest classifier with 3 5-folds was used, and the model AUC and F1 score were used as classification accuracy indicators.
