This repository contains code and test data for the protocol "Protocol for interpretable and context-specific single-cell informed deconvolution of bulk RNA-seq data", currently under review for STAR Protocols.
- protocol_code.R: Contains the lines of code included in the manuscript (excluding the processing from FASTQ to TPM).
- aux_functions.R: Contains all the R functions necessary for executing the protocol.
- align_fastq.sh: Automates the process of downloading FASTQ files and aligning paired-end RNA-Seq data using the STAR aligner.
- kidney_atlas_matrix.rds: Contains the single-cell signatures atlas built in "A transfer learning framework to elucidate the clinical relevance of altered proximal tubule cell states in kidney disease" (Legouis et al., 2024).
- kidney_atlas_info.xlsx: Contains descriptions of the signatures included in the single-cell signatures atlas built in Legouis et al., 2024.
- DKD_tpm.rds: Contains a processed version (TPM) of the dataset GSE142025, also used in Legouis et al., 2024.
- DKD_clin.rds: Contains clinical information (fibrosis) regarding the dataset GSE142025.
- genelength.txt: Contains genes length (to be used in data processing).
The code in protocol_code.R can be fully executed using the test data provided in this repository. Users who might want to skip the training phase (that takes approximately 9 hours) and test a pre-trained model can find the KCLIER model here, together with other intermediate files produced during the execution. We share these file separately since, given their large size, they cannot fit on GitHub.