bixverse SingleCells class

This is the bixverse-based SingleCells class. Under the hood it uses a DuckDB for obs and vars storing, and a Rust-based binarised file format to store the raw and normalised counts. In both cases, the idea is not to hold any data that is not needed at a given point of time in memory, but leverage speedy on-disk computations and streaming engines powered by Rust and DuckDB to run the analysis. This version is specifically designed for single cell RNAseq. If you want to use multi-modal data, please refer to SingleCellsMultiModal() - this class can store multiple layers of 'omics.

Usage

SingleCells(dir_data)

Arguments

dir_data

String. This is the directory in which the experimental files will be

Value

Returns the SingleCells class for further operations.

Properties

db_connection

This contains an R6 class with DuckDB pointers and wrappers to interact with the table-like data for this experiment.

count_connection

This contains an R6-like environment that points to Rust functions that can work on the counts more specifically.

dir_data

Path to the directory in which the data will be saved on disk.

sc_cache

Class with cached data. Contains less memory-heavy objects such as embeddings, kNN information or sNN graphs.

sc_map

Class containing various mapping information such as HVG indices, cells to keep, etc.

dims

Dimensions of the original data.

Examples

# demo_single_cells() wraps the construction and the ingestion
sc <- demo_single_cells(prepped = FALSE)
sc
#> Single cell experiment (Single Cells).
#>   No cells (original): 500
#>    To keep n: 500
#>   No genes: 50
#>   HVG calculated: FALSE
#>   PCA calculated: FALSE
#>   Other embeddings: none
#>   KNN generated: FALSE
#>   SNN generated: FALSE
#>   MAGIC imputed: none
#>   Residual model: none
#>   Stale artefacts: none

unlink(sc@dir_data, recursive = TRUE, force = TRUE)