arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2008.03322q-bio.QM
收藏

scGNN:通过诱导的分层细胞相似性图进行单细胞RNA测序缺失插补

scGNN: scRNA-seq Dropout Imputation via Induced Hierarchical Cell Similarity Graph

Kexin Huang

摘要:

单细胞RNA测序为理解生物系统提供了巨大的见解。然而,来自dropout的噪声可能破坏下游生物分析。因此,最好能准确地估算辍学人数。在这项工作中,我们提出了一个简单而强大的辍学插补方法(scGNN),通过应用一个单校验图卷积神经网络诱导分层细胞相似性图。我们表明,scGNN在三个数据集上与最先进的基线相比具有竞争力的性能,并且可以改善下游分析。

英文摘要:

Single-cell RNA sequencing provides tremendous insights to understand biological systems. However, the noise from dropout can corrupt the downstream biological analysis. Hence, it is desirable to impute the dropouts accurately. In this work, we propose a simple and powerful dropout imputation method (scGNN) by applying a bottlenecked Graph Convolutional Neural Network on an induced hierarchical cell similarity graph. We show scGNN has competitive performance against state-of-the-art baselines across three datasets and can improve downstream analysis.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑