TDCOSR: A Multimodality Fusion Framework for Association Analysis Between Genes and ROIs of Alzheimer’s Disease

Bioinformatics Research and Applications(2023)

引用 0|浏览18
暂无评分
摘要
The complementary multimodality data fusion analysis provides a new perspective for revealing associations between genes and brain regions of interests (ROIs) of Alzheimer’s disease (AD). In this paper, we proposed a multimodality fusion framework of AD gene-ROI association analysis, named the two-stage decision cluster optimal structure reduction (TDCOSR). Specifically, gene-ROI association pairs were constructed as fusion features, where the set of candidate genes was identified from samples with only genetic modality. Second, a series of back propagation neural networks (BPNNs) were trained in parallel and then divided into multiple decision clusters. Finally, the weights of decision clusters were balanced and predict disease states by a linear stacking approach. The weights of decision clusters provide a built-in interpretation mechanism to discover fusion features with strong discriminative power. We tracked the best discriminatory features to identify disease-specific gene-ROI associations. Experiments using the TDCOSR were performed on data from the Alzheimer’s Disease Neuroimaging Initiative database, the results of which showed that the TDCOSR could identify more potential pathogenic genes and ROIs related to AD.
更多
查看译文
关键词
Alzheimer’s disease, Ensemble learning, Optimal feature extraction, Brain imaging genetics
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要