Scholarlink

<p>Alzheimer’s disease (AD) is a complicated disease that attacks the brain’s neurons. Numerous efforts have been directed towards identifying interactions of single-nucleotide polymorphisms (SNPs). Nevertheless, the large volume of SNP data leads to an explosion of high-order SNP combinations, which substantially limits the effectiveness of interaction detection. Consequently, investigating SNP-SNP interactions is crucial in precision medicine (PM). This paper presents two frameworks for identifying and visualizing SNP-SNP interactions associated with AD risk. The first framework aims to integrate ensemble learning techniques and multifactor dimensionality reduction (MDR). The promising outcomes of this framework presented significant risk genes and SNP-SNP interactions with high accuracy. The achieved classification accuracy of 5-way interaction models was 0.874. The accuracy of the 2-way, 3-way, and 4-way models was 0.6648, 0.7169, and 0.7878, respectively. In the second framework, a deep neural network (DNN) is employed with SHapley Additive exPlanations (SHAP) to identify the most highly ranked SNPs that suggest significant SNP-SNP interactions that could aid in interpreting AD risk. The classification accuracy of 5-way interaction models was 0.8. The classification accuracy of the pairwise, 3-way, and 4-way models was 0.65, 0.7, and 0.77, respectively. This study identifies potential risk genes and SNP-SNP interactions associated with AD risk. This work shows that LOC105374292, NSUN7, LOC101929507, and LINC01482 are four novel genes associated with AD by both proposed frameworks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Computational Frameworks for Identifying and Visualizing SNP-SNP Interactions in Alzheimer’s Disease Risk

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月28日

查看全文
  • Marwa M. Abd El Hamid,
  • Mai S. Mabrouk

摘要

Alzheimer’s disease (AD) is a complicated disease that attacks the brain’s neurons. Numerous efforts have been directed towards identifying interactions of single-nucleotide polymorphisms (SNPs). Nevertheless, the large volume of SNP data leads to an explosion of high-order SNP combinations, which substantially limits the effectiveness of interaction detection. Consequently, investigating SNP-SNP interactions is crucial in precision medicine (PM). This paper presents two frameworks for identifying and visualizing SNP-SNP interactions associated with AD risk. The first framework aims to integrate ensemble learning techniques and multifactor dimensionality reduction (MDR). The promising outcomes of this framework presented significant risk genes and SNP-SNP interactions with high accuracy. The achieved classification accuracy of 5-way interaction models was 0.874. The accuracy of the 2-way, 3-way, and 4-way models was 0.6648, 0.7169, and 0.7878, respectively. In the second framework, a deep neural network (DNN) is employed with SHapley Additive exPlanations (SHAP) to identify the most highly ranked SNPs that suggest significant SNP-SNP interactions that could aid in interpreting AD risk. The classification accuracy of 5-way interaction models was 0.8. The classification accuracy of the pairwise, 3-way, and 4-way models was 0.65, 0.7, and 0.77, respectively. This study identifies potential risk genes and SNP-SNP interactions associated with AD risk. This work shows that LOC105374292, NSUN7, LOC101929507, and LINC01482 are four novel genes associated with AD by both proposed frameworks.