Please read Dr. Wei’s article in Genome Research titled, “Complex hierarchical structures in single-cell genomics data unveiled by deep hyperbolic manifold learning.“
Single-cell RNA-seq (scRNA-seq) has provided plenty of new opportunities for exploring cell development and differentiation. Computation methods to accurately reveal and display the cell development process from large single-cell data have grown tremendously in recent years. The progression of cells in continuous trajectories is like a hierarchical tree, with multiple branches typically, such as in Waddington’s classic epigenetic landscape. Methods for analyzing these complex structures in the single-cell data have been published, including visualization, clustering, and pseudotime inference. Visualizing large-scale single-cell data in low dimensions will effectively reveal high-level structural information, which often provides interesting insights for downstream analyses. Despite the compelling potential of scRNA-seq, we note that scRNA-seq data are highly noisy, full of zeros (the dropout phenomenon), and highly dimensional, which makes dimensionality reduction a daunting task. An ideal dimensionality-reduction method is desired to address all these challenges to effectively reveal biological structural patterns in the data. To read the full article.
Complex hierarchical structures in single-cell genomics data unveiled by deep hyperbolic manifold learning. Tian T, Zhong C, Lin X, Wei Z, Hakonarson H. Genome Res. 2023 Feb;33(2):232-246. PMID: 36849204 PMCID: PMC10069463 DOI: 1101/gr.277068.122 Epub 2023 Feb 27.