Sovereign 8s.is mirror of standardgalactic/Multimorbidity-Trajectory-DietaryResponse
- R 100%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| function | ||
| Analysis.R | ||
| Plots.R | ||
| README.md | ||
| Sex_information_cohorts.xlsx | ||
Longitudinal multimorbidity trajectories shape personalized glycemic patterns
Overview
This repository contains the R source code for a comprehensive bioinformatics study investigating the associations between aging-related multimorbidity trajectories and personalized glycemic patterns. The analysis utilizes Continuous Glucose Monitoring (CGM) derived metrics, clinical phenotypes, personalized glycemic sensitivity index (PGS), and serum proteomics.
Analysis pipeline
Part 0: Aging-related multimorbidity trajectories linked to divergent glycemic patterns
- Correlation analysis: partial Spearman correlations between uniqueness of multimorbidity and uniqueness indices of various CGM traits.
- Slicing window analysis: calculation of the coefficient of variation (CV) of glycemic traits across age groups to evaluate glycemic stability during aging.
Part 1: Disease duration of morbidities and CGM-derived daily glycemic traits
- Linear modeling: associations of disease durations with daily glycemic traits (e.g., eA1C, MAGE, CV).
- Disease network: implementation of Bayesian Network (BN) structure learning and causal inference to map the complex interdependencies among various diseases.
- Interaction analysis: investigation of the difference in longitudinal trajectories of multimorbidity stratified by baseline dyslipidemia status.
Part 2: Systemic multimorbidity index (MMI-system) and daily glycemic traits
- Stratified comparison: comparison of glycemic variability across different levels of multimorbidity severity by using multi-variable linear regression analysis.
Part 3: Systemic multimorbidity index, PGS, and dietary responses
- Meal response analysis: Wilcoxon tests and linear regression to observe how PGS associates with postprandial glycemic responses to standardized meal tests.
- Stratified comparison: comparison of the relationship between MMI-system and PGS stratified by multimorbidity severity.
Part 4: Prediction modeling
- Machine learning: random forest models to predict CGM-measured daily traits and postprandial responses.
- Performance evaluation: comparison of basic models (standard clinical markers) versus integrated models (incorporating the multimorbidity trajectories and related traits).
Part 5: Systemic multimorbidity index and serum proteomics
- Identification of multimorbidity-related protein signatures: applying mixed-linear models to identify multimorbidity-related proteins using longitudinal proteomic data.
- Network topology: analyzing protein-protein correlation networks across different health states using topological features.
- Functional enrichment: GO (Gene Ontology) enrichment analysis for biological pathways associated with metabolic comorbidity.
- Mediation analysis: testing the mediation effect of specific proteins between systemic multimorbidity severity and PGS.
- PGS prediction and SHAP integration: using SHAP (SHapley Additive exPlanations) values to interpret the contribution of individual proteins to the prediction of PGS.
Data visualization
Plots.R serves as the core visualization module, containing scripts for generating all primary figure panels using processed outputs derived from the analysis pipeline.
Environment & Dependencies
R Packages
The following core packages are required (full list available in the /functions folder):
- Statistics:
lmerTest (v3.1.3),ppcor (v1.1),vegan (v2.6.8),bnlearn (v5.1),mediation (v4.5.0) - Machine Learning:
randomForest (v4.7.1.2),iml - Bioinformatics:
clusterProfiler (v4.14.4),org.Hs.eg.db (v3.20.0) - Visualization:
ggplot2 (v3.5.1),ggpubr (v0.6.0),igraph (v2.1.2) - Data Handling:
Hmisc (v5.2.2),compositions (v2.0.8)
Usage
- Set Directories: Define your
rootandworkpathin the R environment. - Sequential Execution: Run scripts from
Part 0toPart 5sequentially to maintain data dependencies. - Outputs: Statistical results and processed data are automatically saved in subfolders (e.g.,
./part1_aging/,./part5_proteomics/).
Contact
[Ke Zhang / Lab of Precision nutrition and computational medicine]
[zhangke@westlake.edu.cn]