digitalDLSorteR: an R package to deconvolute bulk RNA-Seq samples using single-cell RNA-seq data and neural networks

digitalDLSorteR is an R package that allows to deconvolute bulk RNA-seq data using context-specific deconvolution models based on single-cell RNA-seq data and neural Networks. These models are able to make accurate estimates of cell composition of bulk RNA-Seq samples from the same context using the meaningful information provided by scRNA-seq data. See Torroja and Sanchez-Cabo (2019) (doi:10.3389/fgene.2019.00978 ) and Mañanes et al., (2024) (doi:10.1093/bioinformatics/btae072 ) for more details.

Details

The method consists of a workflow that starts from single-cell RNA-seq data and, after a few steps, a neural network model is trained with simulated pseudo-bulk RNA-seq samples whose cell composition is known. These trained models are able to deconvolute new bulk RNA-seq samples from the same biological context. Its main advantage is the possibility to build deconvolution models trained with real data from certain biological environments. This fact tries to overcome the limitation of other methods, since cell types may significantly change their transcriptional profiles depending on tissue and disease context.

The package offers two usage ways: deconvolution of bulk RNA-seq samples using pre-trained models available on the digitalDLSorteRmodels R package, or building new deconvolution models from already identified scRNA-seq data. See vignettes and https://diegommcc.github.io/digitalDLSorteR/ for more details.