报告详情
A Survey on Explainable Deep Learning Architectures for Genomic Medicine: Towards Trustworthy AI in Biomedical Informatics
编号：52
访问权限：仅限参会人
更新：2026-11-10 16:32:14 浏览：9次
Online
报告开始：2026年11月21日 14:30（Asia/Kolkata）
报告时间：15min
所在会场：[S4] Artificial Intelligence and Intelligent Computing [S4-9] Artificial Intelligence and Intelligent Computing
暂无文件
摘要
Deep learning models (DL) can capture complex biological signals in high-dimensional omics fields and Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) or even transformers deliver strong performance interpreting e.g., transcription-factor binding, chromatin accessibility, gene expression modelling and variants. At the same time, interpretability represents a restricting factor in their uptake within clinical practice, because we can barely grasp their inner mechanisms. This review attempts to explain XAI methods such as SHAP, DeepLIFt, Layer-wise Relevance Propagation (LRP), Integrated Gradients, attention methods, saliency or even domain aware designs, to explore interpretability for genomic medicine. The methodology starts with screening studies conducted between 2015 and 2025 and 20 are left over of 312 available. A systematization framework is then proposed, which connects data prep, deep model design, explanation, biological validation of XAI and clinical experiment, followed by quantitative evaluation. On average, the authors report an AUC of 0.94 with explainability deep models and some interpretability scores as high as 85%, which is in align with known JASPAR motifs. When combining transformers, SHAP and biological validation, transparency and discovery potential may be boosted while preserving prediction performance. The greatest limitations include lack of biological alignment, no standard interpretability metrics, no experimental cross validation, computational burden and integrating into a clinical workflow at large. In search for more accurate genomic medicine using XAI, solutions could therefore consist of biological-inspired XAI methods, benchmark datasets, wet-lab validation and clinical applications for their use and reports.
关键词
Explainable artificial intelligence, deep learning, genomic medicine, interpretability, SHAP, DeepLIFT, Layer-wise Relevance Propagation, Integrated Gradients, attention mechanisms, biological validation, precision medicine.
报告人
Y Prasada Reddy
Research Scholar Jawaharlal Nehru Technological University Anantapuramu稿件作者
全部评论
重要日期
-
会议日期
11月20日
2026
至11月22日
2026
-
10月31日 2026
初稿截稿日期
-
11月07日 2026
提前注册日期
-
11月22日 2026
注册截止日期
主办单位
United Societies of Science
承办单位
Pondicherry University
协办单位
IEEE Section
IEEE Madras Section
IEEE Madras Section
联系方式
- ICon INDIA 2026
- ic******@usssociety.org
发表评论