Artificial neural network-based analysis of ferroptosis-associated molecular subtypes and immunological profiles in abdominal aortic aneurysm

TL;DR

This study explores how ferroptosis, a type of cell death, contributes to abdominal aortic aneurysms and identifies subtypes linked to immune responses using artificial neural networks.

Contribution

The study introduces a novel ANN-based model to classify AAA subtypes based on ferroptosis-related genes and their immune profiles.

Findings

Nine differentially expressed ferroptosis-related genes were identified, with three key genes achieving high diagnostic accuracy (AUC = 0.988).

AAA samples were classified into two subtypes with distinct immune profiles, with C1 showing higher immune infiltration.

The NeuraAAA score correlated with immune-cell infiltration and was validated with immunofluorescence in AAA specimens.

Abstract

Abdominal aortic aneurysm (AAA) is a severe vascular disease that can lead to rupture and life-threatening hemorrhage. The role of ferroptosis in AAA pathogenesis remains insufficiently understood. This study aims to investigate the role of ferroptosis in AAA by identifying ferroptosis-associated molecular subtypes and examining their relationship with immunological characteristics using an artificial neural network (ANN) model.

We analyzed three publicly available datasets (GSE7084, GSE47472, and GSE57691) to identify differentially expressed ferroptosis-related genes (FRGs) and employed consensus clustering to classify AAA samples into two subtypes. Immune infiltration was assessed with the CIBERSORT algorithm, and a diagnostic artificial neural network (ANN) model based on subtype-specific genes was developed to discriminate ferroptosis-associated molecular subtypes and derive the NeuraAAA score.

Nine differentially expressed FRGs were identified, and the model incorporated three key genes (oncostatin M, heme oxygenase-1, and interleukin-6), achieving high diagnostic accuracy (AUC = 0.988). Consensus clustering stratified AAA samples into two ferroptosis-associated subtypes with distinct immune profiles, with the C1 subtype showing higher immune infiltration and immune scores than C2. The derived NeuraAAA score was elevated in the immune-enriched subtype and correlated with immune-cell infiltration, and a nomogram integrating NeuraAAA and immune score showed good calibration. Immunofluorescence confirmed increased expression of all three genes in AAA specimens.

Our study reveals the heterogeneous role of ferroptosis in AAA pathogenesis, demonstrating that ferroptosis-associated subtypes are linked to variations in the immune microenvironment. These findings provide new insights into AAA pathophysiology and suggest potential targets for subtype-specific therapeutic strategies, contributing to advances in precision medicine for AAA treatment.

Figures