# Integrated bioinformatic analysis and machine learning strategies to identify new potential immune biomarkers for Alzheimer’s disease and their targeting prediction with geniposide

## TL;DR

This study identifies immune biomarkers for Alzheimer's disease and explores geniposide's potential as a treatment using bioinformatics and machine learning.

## Contribution

A novel integrated approach combining bioinformatics and machine learning to discover immune biomarkers and predict geniposide's therapeutic potential in Alzheimer's disease.

## Key findings

- Identified 18 immune-related differential genes in Alzheimer's disease, including GFAP, VGF, NPY, CCK, and NFKBIA.
- Molecular docking and simulations showed geniposide binds well with key Alzheimer's-related proteins.
- Immune cell infiltration analysis linked multiple immune cells to the identified biomarkers.

## Abstract

To analyze the immune biomarkers, pathogenesis, level of immune infiltration, and anti-Alzheimer’s disease (AD) potential of geniposide in immune-related AD. The expression profiles of the GSE132903 dataset were downloaded from the gene expression omnibus (GEO) database to obtain differentially expressed genes (DEGs) in AD, while immune-related genes (IRGs) were obtained from the ImmPortal database, and these genes were intersected to obtain immune differential genes. These genes were intersected to obtain immune differential genes, which were subsequently enriched for further analysis. With the help of protein-protein interaction (PPI) network and cytoHubba analysis, the key immune differential genes were screened out, and the characteristic biomarkers were further identified and screened by the least absolute shrinkage and selection operator (LASSO) regression model and SVM-RFE algorithm. The (receiver operating characteristic) ROC curve was validated in the validation group of GSE5281 microarray and the area under the ROC curve value was used to evaluate the diagnostic and therapeutic values. The CIBERSORT algorithm was used to analyze the pattern of immune cell infiltration and the association between immune cells and characteristic biomarkers. Finally, geniposide was subjected to molecular docking and molecular dynamic simulations with core characterized genes to predict its anti-AD potential. In total, 345 DEGs were identified and 18 AD immune-related differential genes were identified by intersecting immune-related genes, which were involved in multiple signaling pathways, cellular components, molecular functions, and pathways. Five characterized genes were identified using integrated machine learning, including glial fibrillary acidic protein (GFAP), VGF Nerve Growth Factor Inducible (VGF), Neuropeptide Y (NPY), Cholecystokinin (CCK), and NFKB Inhibitor Alpha (NFKBIA). The ROC curve validation results were as expected. Immune cell infiltration analysis revealed that multiple immune cells were associated with the characterized genes. Molecular docking and molecular dynamic simulations showed good binding activity and stability between geniposide and the key characterized targets. Characteristic biomarkers of AD were screened using various methods, and the biological processes and signaling pathways related to AD were identified by enrichment analysis, which elucidated immune-related mechanisms. In addition, geniposide may have binding affinity for key target proteins involved in the pathogenesis of AD, suggesting its potential as a candidate worthy of further investigation. And this study provides a new approach to the pathogenesis and targeted therapy for AD.