Moon, K. R. et al. Nat. Biotechnol. 37, 1482–1492 (2019).
High-dimensional biological data conveys rich information but presents major challenges for analysis and visualization. Mapping such data to lower-dimensional spaces for visualization is often accompanied by information loss. Vast sizes of datasets and omnipresent noise further complicate the task. Moon et al. developed a new method, PHATE (Potential of Heat Diffusion for Affinity-based Transition Embedding), for visualizing high-dimensional data. The main idea is to first encode local data structure and then use a potential distance to measure global relationships. Finally, multidimensional scaling (MDS) is performed to embed the data in a lower-dimensional space. By this strategy, both local and global structures of the original data are accounted for. PHATE not only enables better data visualization than existing methods, but also helps identify interesting patterns such as branching or end points. It is robust to noise, has good scalability and can be used for analyzing different data types, such as mass spectrometry, scRNA-seq, Hi-C and gut microbiota data.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Tang, L. High-dimensional data visualization. Nat Methods 17, 129 (2020). https://doi.org/10.1038/s41592-020-0750-y
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41592-020-0750-y
This article is cited by
-
Review of the state of the art in autonomous artificial intelligence
AI and Ethics (2023)
-
Super-forecasting the ‘technological singularity’ risks from artificial intelligence
Evolving Systems (2022)