popInfer

popInfer is a gene regulatory network inference tool for single-cell multiome data implemented in R.

Citation

If you use popInfer in your research, please cite:

MK Rommelfanger, M Behrends, Y Chen, J Martinez, M Bens, L Xiong, KL Rudolph, AL MacLean (2023). Gene regulatory network inference with popInfer reveals dynamic regulation of hematopoietic stem cell quiescence upon diet restriction and aging. bioRxiv, 10.1101/2023.04.18.537360.

Contents

• data contains single-cell gene expression values, single-cell gene accessibility scores, and pseudotime values for five differentiation trajectories across four samples: HSC -> Multipotent transition in yAL, yDR, oAL, oDR, and the HSC -> GMP transition in oAL.
• R-files contains all of the popInfer R functions.
• outputs contains the files returned by popInfer: a gene expression pseudocell matrix, a gene accessibility score pseudocell matrix, and a gene-gene relationship weight matrix.
• Run-popInfer.R is a script to demonstrating how to run popInfer using the samples found in data.

popInfer inputs and parameters

• rnaData: gene x cell matrix containing the normalized expression values of the cells/genes to run popInfer on.
• geneAccessData: gene x cell matrix containing the gene accessibility of the cells/genes to run popInfer on (must be the same order of features and cells as rnaData). To run popInfer on RNA data only, set geneAccessData to be rnaData.
• pseudotimeData: a dataframe with two columns containing (1) cell barcodes in the same order as the columns of rnaData/geneAccessData, and (2) the corresponding pseudotime values for each cell.
• alpha: a sequence of alpha values (all in the range [0,1]) which popInfer will be run over. Alpha values closer to zero will produce more sparse networks while alpha values closer to one will produce more dense networks.
• outputDir: path to the directory to write the results of popInfer.
• printProgress: boolean variable that designates whether or not to print the gene/index that is being evaluated.

popInfer outputs

• pseudocell-bins.csv: matrix of three columns, where the first two are the same as the pseudotimeData input, and the third column gives the pseudocell bin to which each cell was assigned.
• pseudocell-expression-matrix.csv:
• pseudocell-gene-accessibility-matrix.csv:
• popInfer-weight-matrix.csv: gene x gene matrix containing the weights of gene-gene relationships. Rows correspond to regulator genes while columns correspond to target genes. Negative weights indicate an inhibitory relationship. Absolute weight values can be considered to be the "confidence" in a gene pair interaction.

Paper

Dependencies

[glmnet] for LASSO regression