| 摘 要: 空间转录组的聚类分析对解析肿瘤微环境的异质性和细胞互作至关重要。现有的聚类方法依赖单一维度信息,导致空间信息利用不足,空间域识别的准确性也不理想。针对上述问题,提出基于动态细胞图谱和多分辨率图融合的时空聚类算法 MRGF。在人脑背外侧前额叶皮层(DLPFC)和阿尔茨海默病(AD)数据集上的实验结果明,MRGF的调整兰德指数(ARI)和归一化互信息(NMI)比现有方法平均提升了 12.7% 和 9.3%,验证了MRGF在空间转录组数据分析中具有更高的准确性和鲁棒性。 |
| 关键词: 空间转录组聚类 多视图学习 注意力机制 多视图融合 |
|
中图分类号: TP391
文献标识码: A
|
|
| Research on Spatiotemporal Clustering Algorithm Based on Dynamic Cellular Atlas and Multi-Resolution Graph Fusion |
|
LI Junnan1,ZHANG Wei2,DAI Qi1
|
(1.College of Life Sciences and Medicine, Zhejiang Sc-i Tech University, Hangzhou 310018, China;
2.School of Computer Science and Technology, Zhejiang Sc-i Tech University, Hangzhou 310018, China)
l_junnan@163.com; zhangweicse@zstu.edu.cn; daiqi@zstu.edu.cn
|
| Abstract: Clustering analysis of spatial transcriptomics is crucial for deciphering the heterogeneity of the tumor microenvironment and cellular interactions. Existing clustering methods rely on single-dimensional information, leading to insufficient utilization of spatial information and suboptimal accuracy in spatial domain identification. To address these issues, this paper proposes a spatiotemporal clustering algorithm called MRGF, based on dynamic cellular atlas and mult-i resolution graph fusion. Experimental results on the human brain DorsoLateral PreFrontal Cortex (DLPFC)and Alzheimer’s disease (AD) datasets demonstrate that the proposed MRGF method achieves an average improvement of 12.7% in the Adjusted Rand Index (ARI) and 9.3% in the Normalized Mutual Information (NMI) compared toexisting methods, proving that MRGF exhibits higher accuracy and robustness in spatial transcriptomics data analysis. |
| Keywords: spatial transcriptomics clustering mult-i view learning attention mechanism mult-i view fusion |
