Exploring patterns enriched in a dataset with contrastive principal component analysis

• DOI: 10.1038/s41467-018-04608-8
• Publication Date: 2018-05-24T11:46:01Z
• Authors: Abubakar Abid; Martin J. Zhang; Vivek K. Bagaria; James Zou
• Keywords: Exploratory analysis; Exploratory data analysis; Component (thermodynamics)

Visualization and exploration of high-dimensional data is a ubiquitous challenge across disciplines. Widely used techniques such as principal component analysis (PCA) aim to identify dominant trends in one dataset. However, many settings we have datasets collected under different conditions, e.g., treatment control experiment, are interested visualizing exploring patterns that specific This paper proposes method, contrastive (cPCA), which identifies low-dimensional structures enriched dataset relative comparison data. In wide variety experiments, demonstrate cPCA with background enables us visualize dataset-specific missed by PCA other standard methods. We further provide geometric interpretation strong mathematical guarantees. An implementation publicly available, can be for exploratory applications where currently used.