writeH5AD: Write liger object to H5AD files

In rliger: Linked Inference of Genomic Experimental Relationships

Write liger object to H5AD files

Description

Create an H5AD file from a liger object. This function writes only raw counts to adata.X, while normalized and scaled expression data will not be written, because LIGER use different normalization and scaling strategy than most of the other tools utilizing H5AD format.

Supports for single sparse matrices or internal ligerDataset objects are also provided if there is a need to convert single datasets.

Usage

writeH5AD(object, ...)

## S3 method for class 'dgCMatrix'
writeH5AD(
  object,
  filename,
  obs = NULL,
  var = NULL,
  overwrite = FALSE,
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

## S3 method for class 'ligerDataset'
writeH5AD(
  object,
  filename,
  obs = NULL,
  overwrite = FALSE,
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

## S3 method for class 'liger'
writeH5AD(
  object,
  filename,
  overwrite = FALSE,
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

Arguments

object	One of liger, ligerDataset or dgCMatrix-class object.
...	Arguments passed down to S3 methods
filename	A character string, the path to the H5AD file to be written
obs	External data.frame that contains metadata of the cells but does not embed inside the object. Rownames must be identical to the colnames of object.
var	External data.frame that contains metadata of the features but does not embed inside the object. Rownames must be identical to the rownames of object.
overwrite	Logical, whether to overwrite the file if it exists.
verbose	Logical. Whether to show information of the progress. Default getOption("ligerVerbose") which is TRUE if users have not set.

Value

No return value, an H5AD file is written to disk with the following specification, assuming the file is loaded to adata in Python:

• adata.X - Raw count CSR matrix, outer joined with all datasets

• adata.obs - Cell metadata, with exactly same content of cellMeta(object)

• adata.var - Feature metadata containing only the feature names as the index of pd.DataFrame.

• adata.obsm['X_inmf_aligned'] - The integrated embedding, aligned cell factor loading matrix, the primary output of LIGER, if available.

• adata.obsm['X_inmf'] - The raw cell factor loading matrix, if available.

• adata.obsm['<dimRedName>'] - The dimensional reduction matrix, such as UMAP or TSNE, if available.

• adata.uns['inmf']['W'] - The shared factor feature loading matrix, if available.

• adata.uns['inmf']['V']['<datasetName>'] - The dataset-specific factor feature loading matrix, if available.

• adata.uns['inmf']['features'] - The variable features being used for factorization, supposed to match to the second shape of W and V, if available.

• adata.uns['inmf']['lambda'] - The hyperparameter lambda used, the regularization parameter for the factorization, if available.

• adata.uns['inmf']['k'] - The number of factors used for the factorization, if available.

Examples

print("The example below works, but causes PDF manual rendering issue for some reason")
## Not run: 
writeH5AD(pbmc, filename = tempfile(fileext = ".h5ad"))

## End(Not run)

rliger documentation built on June 8, 2025, 1:56 p.m.