The DigitalDLSorter object is the core of digitalDLSorteR. This object stores different intermediate data resulting from the creation of new context-specific deconvolution models from single-cell data. It is only used in the case of building new deconvolution models. To deconvolute bulk samples using pre-trained models, see deconvDDLSPretrained function and the package digitalDLSorteRdata.


This object uses other classes to store the different types of data produced during the process:

  • SingleCellExperiment class for single-cell RNA-Seq data, using sparse matrix from the Matrix package (dgCMatrix class) or HDF5Array class in the case of using HDF5 files as back-end (see below for more information).

  • ZinbModel class with estimated parameters for the simulation of new single-cell profiles.

  • SummarizedExperiment class for large bulk RNA-Seq data storage.

  • ProbMatrixCellTypes class for the compositional cell matrices constructed during the process. See ?ProbMatrixCellTypes for details.

  • DigitalDLSorterDNN class to store the information related to Deep Neural Network models. This step is performed using keras. See ?DigitalDLSorterDNN for details.

digitalDLSorteR can be used in two ways: to build new deconvolution models from single-cell RNA-Seq data or to deconvolute bulk RNA-Seq samples using pre-trained models available at digitalDLSorteRdata package. If you want to build new models, see createDDLSobject function. On the other hand, if you want to use pre-trained models, see deconvDDLSPretrained function.

In order to provide a way to work with large amounts of data on RAM-constrained machines, we provide the possibility to use HDF5 files as back-end to store count matrices of both real/simulated single-cell and bulk RNA-Seq profiles. For this, the package uses the HDF5Array and DelayedArray classes from the homonymous packages.

Once the Deep Neural Network model has been trained trained, it is possible to save it as RDS or HDF5 files. Please see DigitalDLSorterDNN for more details.



Real single-cell data stored in a SingleCellExperiment object. The count matrix is stored as dgCMatrix or HDF5Array objects.

List of SummarizedExperiment objects where it is possible to store new bulk RNA-Seq experiments for deconvolution. The name of the entries corresponds to the name of the data provided. See trainDDLSModel for details.


ZinbModel object with estimated parameters for the simulation of new single-cell expression profiles.


Simulated single-cell expression profiles from the ZINB-WaVE model.


ProbMatrixCellTypes class with cell composition matrices built for the simulation of pseudo-bulk RNA-Seq profiles with known cell composition.


A list of simulated train and test bulk RNA-Seq samples. Each entry is a SummarizedExperiment object. The count matrices can be stored as HDF5Array files using HDF5 files as back-end in case of RAM limitations.


DigitalDLSorterDNN object with all the information related to the trained model. See ?DigitalDLSorterDNN for more details.


Slot containing the deconvolution results of applying the deconvolution model to the data present in the slot. It is a list in which the names corresponds to the data from which they come.


Name of the project.


Version of DigitalDLSorteR this object was built under.