Zhang Mingxuan, Lai Yinglei
School of Mathematical Sciences, University of Science and Technology of China, Hefei, 230026 Anhui, China.
Department of Statistics, The George Washington University, Washington, DC 20052, United States.
Brief Bioinform. 2025 May 1;26(3). doi: 10.1093/bib/bbaf247.
Single-cell RNA sequencing has significantly advanced our understanding of cell heterogeneity and gene regulation. Batch-effect correction is essential for achieving robust data integration. Multiple methods have been developed to address this issue, particularly procedural approaches involving components such as anchoring or deep learning, which have achieved notable successes. However, order preservation, as an important feature, has been largely overlooked in procedural methods. Based on a monotonic deep learning network, we developed a correction method with order-preserving feature. By comparing with existing methods, we demonstrated that our method effectively improved clustering performance, better retained original inter-gene correlation and differential expression information.
单细胞RNA测序极大地推进了我们对细胞异质性和基因调控的理解。批次效应校正对于实现稳健的数据整合至关重要。已经开发了多种方法来解决这个问题，特别是涉及锚定或深度学习等组件的程序方法，这些方法取得了显著成功。然而，顺序保留作为一个重要特征，在程序方法中很大程度上被忽视了。基于单调深度学习网络，我们开发了一种具有顺序保留特征的校正方法。通过与现有方法比较，我们证明了我们的方法有效地提高了聚类性能，更好地保留了原始基因间相关性和差异表达信息。
