liger-class: The LIGER Class

In rliger: Linked Inference of Genomic Experimental Relationships

The LIGER Class

Description

The liger object is created from two or more single cell datasets. To construct a liger object, the user needs to provide at least two expression (or another single-cell modality) matrices. The class provides functions for data preprocessing, integrative analysis, and visualization.

Details

The key slots used in the liger object are described below.

Slots

raw.data

List of raw data matrices, one per experiment/dataset (genes by cells)

norm.data

List of normalized matrices (genes by cells)

scale.data

List of scaled matrices (cells by genes)

sample.data

List of sampled matrices (gene by cells)

scale.unshared.data

List of scaled matrices of unshared features

h5file.info

List of HDF5-related information for each input dataset. Paths to raw data, indices, indptr, barcodes, genes and the pipeline through which the HDF5 file is formated (10X, AnnData, etc), type of sampled data (raw, normalized or scaled).

cell.data

Dataframe of cell attributes across all datasets (nrows equal to total number cells across all datasets)

var.genes

Subset of informative genes shared across datasets to be used in matrix factorization

var.unshared.features

Highly variable unshared features selected from each dataset

H

Cell loading factors (one matrix per dataset, dimensions cells by k)

H.norm

Normalized cell loading factors (cells across all datasets combined into single matrix)

W

Shared gene loading factors (k by genes)

V

Dataset-specific gene loading factors (one matrix per dataset, dimensions k by genes)

A

Matrices used for online learning (XH)

B

Matrices used for online learning (HTH)

U

Matrices used for unshared Matrix factorization

tsne.coords

Matrix of 2D coordinates obtained from running t-SNE on H.norm or H matrices

alignment.clusters

Initial joint cluster assignments from shared factor alignment

clusters

Joint cluster assignments for cells

snf

List of values associated with shared nearest factor matrix for use in clustering and alignment (out.summary contains edge weight information between cell combinations)

agg.data

Data aggregated within clusters

parameters

List of parameters used throughout analysis

version

Version of package used to create object

rliger documentation built on Nov. 9, 2023, 1:07 a.m.