Scholarlink

This paper presents a novel hybrid framework that uniquely integrates a genetic algorithm (GA) for high-dimensional feature selection with a convolutional neural network (CNN) for interaction-aware classification in Genome-Wide Association Study (GWAS) data. Our approach addresses the critical challenges of high dimensionality and computational inefficiency in traditional GWAS methods. Validated on the DIAGRAM consortium’s Type 2 Diabetes dataset comprising 10,000 samples and 500,000 SNPs, our method achieves 90.1% prediction accuracy with 0.9 AUC-ROC, significantly outperforming conventional approaches. The framework reduces feature dimensionality by 90% while maintaining superior classification performance, demonstrating its potential as a scalable solution for large-scale GWAS analysis in medical research. \(\dots \)

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

GA-CNN Framework for GWAS-Based T2D Prediction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ahmed Miloudi,
  • Mohamed Chikri,
  • Said Boujraf,
  • Youssef Bakadir

摘要

This paper presents a novel hybrid framework that uniquely integrates a genetic algorithm (GA) for high-dimensional feature selection with a convolutional neural network (CNN) for interaction-aware classification in Genome-Wide Association Study (GWAS) data. Our approach addresses the critical challenges of high dimensionality and computational inefficiency in traditional GWAS methods. Validated on the DIAGRAM consortium’s Type 2 Diabetes dataset comprising 10,000 samples and 500,000 SNPs, our method achieves 90.1% prediction accuracy with 0.9 AUC-ROC, significantly outperforming conventional approaches. The framework reduces feature dimensionality by 90% while maintaining superior classification performance, demonstrating its potential as a scalable solution for large-scale GWAS analysis in medical research. \(\dots \)