Annals of Indian Academy of Neurology
This summary was generated by AI from a single paper. It has not been reviewed by a clinician and is not clinical advice. Verify against the source before acting on it.
This study developed a computational model using normalized pointwise mutual information (nPMI) to identify genes associated with stroke and its subtypes from PubMed literature. A prognostic panel of nine genes (CYP4A11, ALOX5P, NOTCH, NINJ2, FGB, MTHFR, PDE4D, HDAC9, ZHFX3) was identified as potential diagnostic markers for stroke risk. The nPMI-based model outperformed gene expression-based methods, suggesting a novel computational approach for stroke gene prediction.
**Background:** Stroke is a complex heterogeneous disorder resulting from the interaction of environmental and genetic risk factors. It is a leading cause of mortality and long-term disability worldwide. While numerous genes have been linked to stroke risk, there are no well-established genetic markers that can discriminate between stroke types and subtypes. Experimental determination of such markers is costly and time-consuming, highlighting the need for novel computational methods. This study aimed to develop a computational algorithm to identify potential key genes and pathways associated with the pathogenesis and prediction of stroke types and their subtypes.
**Methods:** The researchers used the NCBI gene database to extract 28,281 human gene symbols with an 'Active' status. PubMed advanced queries were constructed using stroke-related keywords, and associations between each gene symbol and the queries were calculated using normalized pointwise mutual information (nPMI). Only titles and abstracts from articles published up to August 31, 2020, were searched to reduce false hits. The model's performance was evaluated using receiver operating characteristic (ROC) and precision-recall (PR) curve analysis on a cumulative dataset of human housekeeping genes (negative set) and published stroke-related genes (positive set). The nPMI-based model was compared with an existing gene expression-based method using microarray datasets from the Gene Expression Omnibus (GEO). Pathway analysis was performed using the clusterProfiler R package on KEGG, WikiPathways, and Reactome databases.
**Key Results:** A total of 2,785 (9.4%) genes were found to be linked to stroke risk. Of these, 1,287 (46.2%) were associated with ischemic stroke (IS) and 376 (13.5%) with hemorrhagic stroke (HS). Further stratification of IS based on TOAST classification revealed 86 (6.6%) genes for large artery atherosclerosis (LAA), 131 (10.1%) for small vessel disease (SVD), and 130 (10%) for cardioembolism (CE). For HS subtypes, 292 (77.8%) genes were predicted for intracerebral hemorrhage (ICH) and 132 (35.2%) for subarachnoid hemorrhage (SAH). The nPMI-based model demonstrated better performance than the gene expression-based method, as reflected by comparisons of evaluation metrics. A prognostic panel of nine highly associated genes was concluded: CYP4A11, ALOX5P, NOTCH, NINJ2, FGB, MTHFR, PDE4D, HDAC9, and ZHFX3. Pathway enrichment analysis using 190 manually curated stroke-associated genes identified 53, 32, and 35 unique pathways enriched in Reactome, WikiPathways, and KEGG, respectively.
**Clinical Implications:** This study presents a novel computational approach for identifying stroke-associated genes, which may aid in predicting individuals at risk of developing stroke. The identified nine-gene prognostic panel could serve as a diagnostic marker, though validation through large-scale case-control studies is required. The model's ability to stratify genes by stroke type and subtype may facilitate a better understanding of the underlying pathophysiology and potentially guide personalized prevention and treatment strategies.