RayleighSelection is an R package for feature selection in topological spaces. Features are defined as differential forms on a simplicial complex and their significance is assessed through Rayleigh quotients. Further details can be found in:
- K. W. Govek*, V. S. Yamajala*, and P. G. Cámara. Clustering-independent analysis of genomic data using spectral simplicial Theory. PLOS Computational Biology 15 (2019) 11. DOI: 10.1371/journal.pcbi.1007509. [*authors contributed equally].
and its Supplementary Note 1. Spectral simplicial theory for feature selection.
library(devtools)
install_github("CamaraLab/RayleighSelection")
Note for Windows users: This package uses the mclapply function for parallelization, which is not supported for Windows. You can either run with num_cores = 1 or use our Docker image.
Please use our Docker image camaralab/rayleigh-selection to run an RStudio server (v3.4 or v3.6) with RayleighSelection already installed:
docker run -d --rm -p 8787:8787 -v "<dir_path>:/home/rstudio/<dir_name>" -e USER=rstudio -e PASSWORD=<password> camaralab/rayleigh-selection
After running the above command, RStudio should be available at localhost:8787 in your browser with the local directory at <dir_path> mounted in Home.
Given an open cover and a feature on points, compute the Combinatorial Laplacian scores of that feature on the nerve complex of the cover.
Vietoris-Rips on cyclic scRNA-seq data
Given the PCA results of mouse embryonic cells in two differentiation protocols and an ordering on the cells, create a Vietoris-Rips complex. Compute the Combinatorial Laplacian score of gene expression on either just the 0-forms (fast) or both 0-forms and 1-forms (slow).
Nerve complex on Mapper representation of MNIST
Run Mapper on the MNIST dataset to compute an open cover on the handwriting samples, then compute the Combinatorial Laplacian score of the pixel intensity.