Agora Data and Methods
This page provides information about the data and methods used for the experimental results presented in Agora.
RNA Expression Data and Methods
RNA-seq data used for these analyses was generated from over 2100 samples from post-mortem brains of more than 1100 individuals. In total there are nine distinct brain regions from three human cohort studies:
Cohort Study | Brain Region(s) |
|---|---|
Religious Orders Study and Memory and Aging Project (ROSMAP) | Anterior Cingulate Cortex (ACC), Dorsolateral Prefrontal Cortex (DLPFC), Posterior Cingulate Cortex (PCC) |
Cerebellum (CBE), Temporal Cortex (TCX) | |
Frontal Pole (FP), Inferior Frontal Gyrus (IFG), Parahippocampal Gyrus (PHG), Superior Temporal Gyrus (STG) |
RNA-seq data from these studies was processed through a harmonized computational pipeline to reveal overall expression levels and to measure differential expression between AD cases and controls. The gene coexpression network is based on the original harmonized computational pipeline.
The data and methodology is publicly available through these links:
Statistical Models
Differential RNA expression data can be viewed for each of the following models:
Model | Description |
|---|---|
AD Diagnosis (males and females) | Shows relative changes in gene expression between AD patients and controls. |
AD Diagnosis x AOD (males and females) | Shows relative changes in gene expression between AD patients and controls, and whether Age of Death (AOD) has an impact. |
AD Diagnosis x Sex (females only) | Shows relative changes in gene expression between female AD patients and female controls. |
AD Diagnosis x Sex (males only) | Shows relative changes in gene expression between male AD patients and male controls. |
Proteomics Data and Methods
Protein abundance has been quantified using two methods; liquid-free quantification (LFQ) and tandem mass tagged (TMT) spectrometry:
LFQ Data and Methods
LFQ data was generated from post-mortem brains of more than 500 individuals. Samples were taken from four human cohort studies, representing four different brain regions:
Cohort Study | Brain Region(s) |
|---|---|
Dorsolateral Prefrontal Cortex (DLPFC) | |
Middle Frontal Gyrus (MFG) | |
Temporal Cortex (TCX) | |
Anterior Prefrontal Cortex (AntPFC) |
Samples were analyzed using liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS). Protein abundance was quantified using liquid-free quantification (LFQ). Note that not all proteins are detected in all brain regions; for these proteins, the plots will show fewer than four brain regions.
Samples were harmonized within study by batch correction with COMBAT and regressing out Age, Sex, and Post Mortem Interval (PMI). Differential expression of AD versus controls was determined via ANOVA with an FDR of 0.05.
The processed data is publicly available in Synapse through this link: LFQ Proteomics for Agora.
TMT Data and Methods
TMT data was generated from post-mortem brains of 400 individuals from the ROSMAP cohort:
Cohort Study | Brain Region |
|---|---|
Religious Orders Study and Memory and Aging Project (ROSMAP) | Dorsolateral Prefrontal Cortex (DLPFC) |
Samples were analyzed using isobaric tandem mass tag (TMT) peptide labeling as described in the Round 1 Methods section of Proteomics (TMT quantitation). Batch correction, relative abundance calculation, and log2 transformation was performed as described here.
Differential expression of AD versus controls was determined using linear regression, adjusted for sex and post-mortem interval. Output consists of log2 fold change for AD cases versus controls (with controls as the reference group), 95% confidence intervals, and FDR-adjusted p-values.
The full differential expression results are publicly available in Synapse through this link: TMT Proteomics for Agora.
Metabolomics Data and Methods
Metabolic data for Agora was analyzed using a multi-step process. Metabolite data was obtained from approximately 1,400 samples of the Alzheimer’s Disease Neuroimaging (ADNI) cohorts. Metabolite concentrations from 157 metabolites were obtained in clinical lab tests performed by participating ADNI centers. Sampled metabolites include: 140 from the Biocrates AbsoluteIDQ® P180 platform, 15 from the Biocrates® Bile Acids Kit, as well as concentrations of total triglycerides and total cholesterol.
Metabolite levels were then associated with individual genes by performing a genome- wide association study using metabolite concentrations as phenotype (mGWAS). This mGWAS analysis yielded 15,401 significant gene-metabolite pairs. For these pairs, we report the Ensembl GeneID, the Associated Gene Name, ID and full name of the metabolite with the strongest genetic association, the P-value of the association, as well as the gene-wide significance threshold derived from 1000 Genomes Project variants.
For all 157 metabolic readouts, we then used linear regression to compare medication- and covariate-adjusted, centered, and scaled metabolite residuals from cognitively normal elderly controls (CN) with those of participants with clinical Alzheimer’s disease (AD). In this analysis, 26 out of the 157 metabolites had an adjusted P-value <= 0.05.
This content has been provided by the Duke AMP-AD Team led by Rima Kaddurah-Daouk. Learn more about the team.