PCA is the most commonly used clustering method. Through machine learning some new methods were also reported.
Title image from http://satijalab.org/seurat/get_started_v1_2.html.
Methods for dimensionality reductions
- Linear: PCA
- Non-linear: t-SNE (t-Distributed Stochastic Neighbor Embedding)
- shared k-nearest neighbor (KNN), e.g., SNN-Clip
- Centered Pearson’s correlation (e.g., SINCERA)
- cell similarity matrix (e.g., SIMLR)
- zero-inflated factor analysis (ZIFA)
- Discriminant Analysis
- Neural networks (e.g., Using neural networks for reducing the dimensions of single-cell RNA-Seq data)
Clustering methods (examples here)
- Hierarchical clustering
- scRNAseqDB (human single cell)