tissueResolver Docs

tissueResolver is a package for converting bulk RNA-seq datasets into virtual tissues using information from similar single cell datasets by assigning weights to true single cells, maintaining their molecular integrity. Virtual tissues can be analyzed in a similar way as conventional single cell datasets.

For theoretical details and interpretation of our results, see our preprint bioarxiv.

In this vignette we want to give a general overview over the most important functions used when building a practical tissueResolver pipeline and provide reproducibility of the results of our paper. To round off this vignette we also provide several scripts allowing to reproduce all the results of our paper not covered by the vignette, see Additional Scripts for an overview.

For detailed explanation and examples of the main package functions of tissueResolver we refer to the corresponding R package and its documentation available under Spang-Lab GitHub.

Data Accessibility

We provide all the necessary data for download via the Zenodo DOI: 10.5281/zenodo.10568550.

The case study of our paper is based on the following raw data:

• bulk RNA-seq: From the publication Genetics and pathogenesis of diffuse large B-cell lymphoma by [Schmitz et al., 2018] accessible under GDC
• single cell RNA-seq: Here we combined two datasets without further harmonisation:
  • Dissecting intratumour heterogeneity of nodal B-cell lymphomas at the transcriptional, genetic and drug-response levels by [Roider et al., 2020] accessible under heiDATA ID VRJUNV or EGAS00001004335
  • The landscape of tumor cell states and ecosystems in diffuse large B cell lymphoma by [Steen et al., 2021], accessibly under GSE182436 and GSE182434

We note that the data comprised in bulks.rds and sc.rds is the result of the gene filtering procedure explained in supplemental section Gene filtering of our paper.

Building the .Rmd File

In order to render the .html file from tissue_resolver_vignette.Rmd invoke

rmarkdown::render("tissue_resolver_vignette.Rmd")

If you want to run the code as a script you can simply extract it from the `.Rmd`` file via

knitr::purl("tissue_resolver_vignette.Rmd")

We intentionally set some computationally heavy R chunks to eval = FALSE and provided the output plots for a quick build.

Setting eval = TRUE and only storing bulks.rds and sc.rds in the data folder will compute everything from scratch.

If you run the tissue_resolver_vignette.Rmd for the first time, we provided all the data under Zenodo DOI: 10.5281/zenodo.10568550. For compiling tissue_resolver_vignette.Rmd you only need sc.rds and bulks.rds to be stored in the data folder. The more additional data you provide, the less will be computed by tissueResolver.

Additional Scripts

We provide several scripts to complement the vignette in order to allow for full reproducibility of the results of our paper:

• simulations_paper.R reproduces the benchmark simulation comparing BayesPrism with tissueResolver, see section Simulations and in particular the pseudo algorithm we provide in the supplementary material of our publication.
• bar_heatmap_DGE.R reproduces the bar/heatmap plot for differentially expressed genes possessing good quality scores, see supplementary section Differentially expressed genes in virtual tissues
• bar_heatmap_stromal.R reproduces the bar/heatmap plot for genes belonging to the stromal signature, see section The micro-environment of diffuse large B-cell lymphomas
• qc_all.R reproduces the quality control plots for the whole set of 1000 analyzed genes, see supplementary section Quality control of virtual tissues
• qc_stromal.R reproduces the quality control plots for the genes contained in the stromal signature and their relation to cluster 17, see supplementary section Quality control of virtual tissues
• map_feature_names.R is a helper script converting ENSEMBL gene ids to HGNC symbols for a more convenient depiction of gene names