# Exploring common circulating diagnostic biomarkers for sleep disorders and stroke based on machine learning

## TL;DR

This study identifies ARL2 as a potential blood-based biomarker for sleep disorders and stroke, using machine learning and bioinformatics to uncover shared molecular pathways.

## Contribution

The study introduces ARL2 as a novel, non-invasive circulating biomarker for sleep disorders and stroke.

## Key findings

- ARL2 was identified as a key diagnostic biomarker with high predictive value (AUC = 0.91).
- Shared biological processes like 'response to external stimuli' and 'propionate metabolism' were significantly enriched in both disorders.
- Animal experiments confirmed ARL2 upregulation in the experimental group.

## Abstract

Sleep disorders (SD) and stroke have long been health concerns. Sleep disorders are known to be a risk factor for stroke, and in recent years it has also been shown that the prevalence of sleep disorders is increased in stroke patients. We inferred that there is some inevitable connection between the two. This study aims to identify common molecular biomarkers and pathways connecting SD and stroke by integrating bioinformatics and machine learning approaches.

We analyzed transcriptome data from the GEO dataset to identify differentially expressed genes (DEGs). Key biological processes, as well as metabolic pathways, were highlighted by GO and KGEE enrichment analyses. Co-expression modules were then identified in the SD and stroke datasets by weighted gene co-expression network analysis (WGCNA), respectively, and machine learning algorithms (RandomForest, LASSO, and XGBoost) were performed to identify ARL2 as a key diagnostic biomarker with high predictive value (AUC = 0.91). This was finally complemented by animal experiments to verify that ARL2 was upregulated in the experimental group.

In GO and KEGG enrichment analyses, key biological processes such as ‘response to external stimuli’ and ‘organic metabolic processes’ as well as metabolic pathways such as ‘propionate metabolism’ and ‘oxidative phosphorylation’ were significantly enriched, suggesting their potential roles in the pathogenesis of the two disorders. With WGCNA and machine-learning algorithms analyses, we found that ARL2 is an important common marker for both diseases.

This study provides insights into the common molecular mechanisms of SD and stroke, highlighting the potential of ARL2 as a diagnostic marker and therapeutic target. Unlike previous studies, we used circulating markers rather than tissue markers, improving the clinical translation in terms of non-invasive, rapid identification of patients at risk for sleep disorders. We need to further investigate the functional role of these genes and their potential in developing targeted therapies for SD and stroke patients.