This repository contains a demonstration of the RAPToR tool with multiple branching trajectories. The algorithm is described in this paper. The code is based on the original RAPToR implementation.
The main function is multige_im in the multige_im.R file.
Parameters: X: Expression matrix (genes by samples), expecting log1p(tpm) p: Pheno data, samples as rows. Must at the minimum have the variables used in formulas. lineages: named list of vectors of indices or booleans defining each lineage formulas: the formulas to use for each model. can be given as - a single formula to use the same for all lineages/components - a vector of size length(lineages) to use lineage-specific formulas - a vector of size length(ncs) to use component-specific formulas - a matrix of dim [length(lineages), length(ncs)] to use lineage- and component-specific formulas
Returns: dim_red: whether to use pca/ica nc: the number of components to extract ncs (optional): which components to keeo. defaults to 1:nc.
We also developed a function to infer lineages on top of the usual RAPToR output. This function is the ale function in the ale.R file. It is based on the original ae function.
Parameters: samp: a gene expression matrix (genes by samples). multiref: a list with
- $interpGE, a named list of gene expression matrices (one per lineage), as returned by predict_mgeim()
- $time, a vector of time values (length of ncol(interpGE[[1]]))
In addition to these two you will find several utility functions in the utils.R file.
For this analysis we worked with a single-cell C.elegans dataset Packer et al. (2019). The data is accessible on GEO: GSE126954. The data is also accessible via the VisCello R package (Zhu, Murray, Tan, and Kim 2019) and the expression matrix is directly downloadable from their GitHub repository here.
The whole analysis can be found in the neurons_analysis.R file as well as the neurons_analysis.Rms notebook, compiled with knitr into the neurons_analysis.html file.
Several figures were generated in the fig folder.
A complete analysis of another dataset is also accessible in the multit_traj.R file. In this one we analysed a single-cell profiling data by Hashimshony et al. (2015). The data is accessible on GEO: GSE50548.
Finaly, a slingshot demonstration can be found in the slingshot_analysis.R file.