Hybrid Attention Capsule Network for the Detection of Alzheimer's Disease-Associated Gene Expression Profiles
ID：12
View protection：Participant Only
Updated time：2026-07-28 21:49:17
Views：26
Online
Abstract
The goal of gene signature analysis in gene expression data is to correlate variations in the degree of expression of certain genes with the presence or absence of specific gene expressions. However, there are challenges because microarray databases are high dimensional and the sample count is limited. The study aims to investigate deep learning (DL) techniques, particularly neural networks, to predict Alzheimer's disease using gene expression data. First, the input gene expression data is pre-processed employing Z-score normalization and Missing value imputation methods. The most effective characteristics are then selected using the Chebyshev Pufferfish Algorithm (CPA), which reduces dimensionality. The proposed model introduces a novel Depthwise Separable Convolutional Attention-based Chi-Squared Capsule Network (DSCA-CSCN) for analyzing AD gene expressions using gene expression data. The proposed framework presents an integrated architecture that combines several effective techniques to enable simultaneous enhancement of gene expression discrimination and reduction of computational complexity in high-dimensional gene expression data. Furthermore, the Explainable AI method (XAI) analyzes the interactions between various attributes that can be conveniently captured and displayed using SHapley Additive Explanations (SHAP). The proposed framework attains accuracy (98%), precision (97.41%), recall (97.75%), f1-score (96.17%) and specificity (97.25%).
Keywords
Gene Expression Data,Alzheimer;s disease,pre-processing,Chebyshev Pufferfish Algorithm (CPA),Depthwise Separable Convolutional Attention based Chi-Squared Capsule Network (DSCA-CSCN),SHapley Additive exPlanations (SHAP)
Post comments