NEWS

---

digitalDLSorteR 1.1.1

Problems to be solved:

• saveRDS thingy
• Documentation: remove all linkS4classes

digitalDLSorteR 1.1.0 (2024-09-13)

• Included two parameters to control python and tensorflow versions in the installTFpython function.
• The installTFpython function now installs python 3.8 by default.
• Included interGradientsDL.R file with functions to interpret neural networks using Vanilla Gradient (from SpatialDDLS).
• Functions renamed: deconvDigitalDLSorter > deconvDDLSObj and deconvDigitalDLSorterObj > deconvDDLSPretrained.
• Vignettes updated using new deconvolution models/data.

digitalDLSorteR 1.0.1 (2024-02-07)

• Some name functions have been changed: createDDLSobject and trainDDLSModel.
• Loading data (createDDLSobject) has been changed: now, only 2,000 genes are used for deconvolution and bulk RNA-seq must be provided at the beginning of the workflow (see documentation).
• Fixed bugs: standardization of features before training.
• Vignette modified according to the new functions.
• HDF5Array update: use.for parameter has been deleted.

digitalDLSorteR 0.3.2

• Changed the way dense matrices (matrix) and data.frame objects are transformed into dgCMatrix objects.

digitalDLSorteR 0.3.1 (2022-10-05)

• The Matrix.utils dependency has been removed: instead of using the aggregate.Matrix function, it is used functions implemented in the base package.

digitalDLSorteR 0.3.0 (2022-05-24)

• The splatter dependency has been removed: instead of using splatter as a wrapper, zinbFit from the zinbwave package is used directly via the .zinbWaveModel function + the ZinbParametersModel class. This change affects some functions in terms of classes/objects. Previous pre-trained models (digitalDLSorteRmodels package) may not work properly. They will be generated soon.
• The edgeR dependency has been removed: CPM-related calculations have been implemented (.cpmCalculate function). Now, results may be slightly different from those obtained with edgeR.

digitalDLSorteR 0.2.0 (2022-03-10)

• Implemented different ways to generate pseudo-bulk samples: MeanCPM, AddCPM, and AddRawCount (simBulkProfiles function).
• Implemented different ways to scale data before training: standarization and rescaling (trainDigitalDLSorterModel function).
• Vignettes updated.

digitalDLSorteR 0.1.1 (2021-10-19)

• Added Solaris OS as one of the options in switch function (BiocParallel environment in estimateZinbwaveParams function).
• Some changes related to information about the package: messages during execution, README.

digitalDLSorteR 0.1.0 (2021-10-08)

• Added a NEWS.md file to track changes to the package.