A Robust Tool for Performing Transcriptome Profiling on Cancer Data • RNAsum

RNAsum

RNAsum is an R package that can post-process, summarise and visualise outputs primarily from DRAGEN RNA pipelines. Its main application is to complement whole-genome based findings and to provide additional evidence for detected alterations.

DOCS: https://umccr.github.io/RNAsum

Installation

R package can be installed directly from the GitHub source:

remotes::install_github("umccr/RNAsum") # latest main commit
remotes::install_github("umccr/[email protected]") # version 0.0.X
remotes::install_github("umccr/RNAsum@abcde") # commit abcde
remotes::install_github("umccr/RNAsum#123") # PR 123

Conda package is available from the Anaconda umccr channel:

conda install r-rnasum==X.X.X -c umccr -c conda-forge -c bioconda

Docker image is available from the GitHub Container Registy:

docker pull ghcr.io/umccr/rnasum:latest

Workflow

The pipeline consists of five main components illustrated and briefly described below. For more details, see workflow.md.

Collect patient WTS data from the DRAGEN RNA pipeline including per-gene read counts and gene fusions.
Add expression data from reference cohorts to get an idea about the expression levels of genes of interest in other cancer patient cohorts. The read counts are normalised, transformed and converted into a scale that allows to present the patient’s expression measurements in the context of the reference cohorts.
Supply genome-based findings from whole-genome sequencing (WGS) data to focus on genes of interest and to provide additional evidence for dysregulation of mutated genes, or genes located within detected structural variants (SVs) or copy-number (CN) altered regions. RNAsum is designed to be compatible with WGS patient outputs generated from umccrise.
Collate results with knowledge derived from in-house resources and public databases to provide additional sources of evidence for clinical significance of altered genes e.g. to flag variants with clinical significance or potential druggable targets.
The final product is an interactive HTML report with searchable tables and plots presenting expression levels of the genes of interest. The report consists of several sections described here.

Reference data

The reference expression data are available for 33 cancer types and were derived from external (TCGA) and internal (UMCCR) resources.

External reference cohorts

In order to explore expression changes in the patient, we have built a high-quality pancreatic cancer reference cohort.

Depending on the tissue from which the patient’s sample was taken, one of 33 cancer datasets from TCGA can be used as a reference cohort for comparing expression changes in genes of interest of the patient. Additionally, 10 samples from each of the 33 TCGA datasets were combined to create the Pan-Cancer dataset, and for some cohorts extended sets are also available. All available datasets are listed in the TCGA projects summary table. These datasets have been processed using methods described in the TCGA-data-harmonization repository. The dataset of interest can be specified by using one of the TCGA project IDs for the RNAsum --dataset argument (see Examples).

Internal reference cohort

The publicly available TCGA datasets are expected to demonstrate prominent batch effects when compared to the in-house WTS data due to differences in applied experimental procedures and analytical pipelines. Moreover, TCGA data may include samples from tissue material of lower quality and cellularity compared to samples processed using local protocols. To address these issues, we have built a high-quality internal reference cohort processed using the same pipelines as input data (see data pre-processing).

This internal reference set of 40 pancreatic cancer samples is based on WTS data generated at UMCCR and processed with the bcbio-nextgen RNA-seq pipeline to minimise potential batch effects between investigated samples and the reference cohort and to make sure the data are comparable. The internal reference cohort assembly is summarised in the Pancreatic-data-harmonization repository.

Note

There are two rationales for using the internal reference cohort:

In case of pancreatic cancer samples this cohort is used:
- in batch effects correction
- as a reference point for comparing per-gene expression levels observed in the data of the patient of interest and data from other pancreatic cancer patients.
In case of samples from any cancer type the data from the internal reference cohort is used in the batch effects correction procedure performed to minimise technical-related variation in the data.

Input data

RNAsum accepts WTS data processed by the state-of-the-art bioinformatic tools such as kallisto and salmon for quantification and Arriba for fusion calling. RNAsum can aso process and combine fusion output from Illumina’s Dragen pipeline. Additionally, the WTS data can be integrated with WGS-based data processed using the tools discussed in the section WGS.

In the latter case, the genome-based findings from the corresponding sample are incorporated into the report and are used as a primary source for expression profile prioritisation.

WTS

The only required WTS input data are read counts provided in a quantification file.

RNA

The table below lists all input data accepted in RNAsum:

Input file	Tool	Example	Required
Quantified transcript abundances	salmon (description)	*.quant.sf	Yes
Quantified gene abundances	salmon (description)	*.quant.gene.sf	Yes
Fusion gene list	Arriba	fusions.tsv	No
Fusion gene list	DRAGEN RNA	*.fusion_candidates.final	No

WGS

RNAsum is designed to be compatible with WGS outputs.

The table below lists all input data accepted in RNAsum:

Input file	Tool	Example	Required
SNVs/Indels	PCGR	pcgr.snvs_indels.tiers.tsv	No
CNVs	PURPLE	purple.cnv.gene.tsv	No
SVs	Manta	sv-prioritize-manta.tsv	No

Usage

rnasum_cli=$(Rscript -e 'x = system.file("cli", package = "RNAsum"); cat(x, "\n")' | xargs)
export PATH="${rnasum_cli}:${PATH}"

$ rnasum.R --version
1.1.5 

$ rnasum.R --help
Usage
=====
 
/Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/library/RNAsum/cli/rnasum.R [options]


Options
=======
--arriba_dir=ARRIBA_DIR
        Directory path to Arriba results containing fusions.pdf and fusions.tsv.

--arriba_pdf=ARRIBA_PDF
        File path of Arriba PDF output.

--arriba_tsv=ARRIBA_TSV
        File path of Arriba TSV output.

--batch_rm
        Remove batch-associated effects between datasets.

--cn_gain=CN_GAIN
        CN threshold value to classify genes within gained regions. [def: 95]

--cn_loss=CN_LOSS
        CN threshold value to classify genes within lost regions. [def: 5]

--dataset=DATASET
        Dataset to be used as external reference cohort. [def: PANCAN]

--dataset_name_incl
        Include dataset in report name.

--dragen_fusions=DRAGEN_FUSIONS
        File path to DRAGEN RNA-seq 'fusion_candidates.final' output.

--dragen_mapping_metrics=DRAGEN_MAPPING_METRICS
        File path to DRAGEN RNA-seq 'mapping_metrics.csv' output.

--dragen_wts_dir=DRAGEN_WTS_DIR
        Directory path to DRAGEN RNA-seq results.

--drugs
        Include drug matching section in report.

--filter
        Filter out low expressed genes.

--immunogram
        Include immunogram in report.

--log
        Log2 transform data before normalisation.

--norm=NORM
        Normalisation method.

--pcgr_splice_vars
        Include non-coding splice region variants reported in PCGR.

--pcgr_tier=PCGR_TIER
        Tier threshold for reporting variants reported in PCGR. [def: 4]

--pcgr_tiers_tsv=PCGR_TIERS_TSV
        File path to PCGR 'snvs_indels.tiers.tsv' output.

--project=PROJECT
        Project name, used for annotation purposes only.

--purple_gene_tsv=PURPLE_GENE_TSV
        File path to PURPLE 'purple.cnv.gene.tsv' output.

--report_dir=REPORT_DIR
        Directory path to output report.

--salmon=SALMON
        File path to salmon 'quant.genes.sf' output.

--sample_name=SAMPLE_NAME
        Sample name to be presented in report.

--sample_source=SAMPLE_SOURCE
        Type of investigated sample. [def: -]

--save_tables
        Save interactive summary tables as HTML.

--scaling=SCALING
        Scaling for z-score transformation (gene-wise or group-wise). [def: gene-wise]

--subject_id=SUBJECT_ID
        Subject ID.

--sv_tsv=SV_TSV
        File path to text file with genes related to structural variation.

--top_genes=TOP_GENES
        Number of top ranked genes to be presented in report.

--transform=TRANSFORM
        Transformation method to be used when converting read counts. [def: CPM]

--umccrise=UMCCRISE
        Directory path of the corresponding WGS-related umccrise data.

--version, -v
        Print RNAsum version and exit.

--help, -h
        Show this help message and exit

Note

Human reference genome GRCh38 (Ensembl based annotation version 105) is used for gene annotation by default. GRCh37 is no longer supported.

Examples

Below are RNAsum CLI commands for generating HTML reports under different data availability scenarios:

Note

Example data is provided in the /inst/rawdata/test_data folder of the GitHub repo.
The RNAsum runtime should be less than 15 minutes using 16GB RAM memory and 1 CPU.

1. WTS and WGS data

This is the most frequent and preferred case, in which the WGS-based findings will be used as a primary source for expression profile prioritisation. The genome-based results can be incorporated into the report by specifying the location of the corresponding output files (including results from PCGR, PURPLE, and Manta). The Mutated genes, Structural variants and CN altered genes report sections will contain information about expression levels of the mutated genes, genes located within detected SVs and CN altered regions, respectively. The results in the Fusion genes section will be ordered based on the evidence from genome-based data. A subset of the TCGA pancreatic adenocarcinoma dataset is used as reference cohort (--dataset TEST).

rnasum.R \
  --sample_name test_sample_WTS \
  --dataset TEST \
  --dragen_wts_dir inst/rawdata/test_data/dragen \
  --report_dir inst/rawdata/test_data/dragen/RNAsum \
  --umccrise inst/rawdata/test_data/umccrised/test_sample_WGS \
  --save_tables FALSE

The HTML report test_sample_WTS.RNAsum.html will be created in the inst/rawdata/test_data/dragen/RNAsum folder.

2. WTS data only

In this scenario, only WTS data will be used and only expression levels of key Cancer genes, Fusion genes, Immune markers and homologous recombination deficiency genes (HRD genes) will be reported. Moreover, gene fusions reported in the Fusion genes report section will not contain information about evidence from genome-based data. A subset of the TCGA pancreatic adenocarcinoma dataset is used as the reference cohort (--dataset TEST).

rnasum.R \
  --sample_name test_sample_WTS \
  --dataset TEST \
  --dragen_wts_dir inst/rawdata/test_data/dragen \
  --report_dir inst/rawdata/test_data/dragen/RNAsum \
  --save_tables FALSE

The output HTML report test_sample_WTS.RNAsum.html will be created in the inst/rawdata/test_data/dragen/RNAsum folder.

3. WTS WGS and clinical data

For samples derived from subjects, for which clinical information is available, a treatment regimen timeline can be added to the HTML report. This can be added by specifying the location of a relevant excel spreadsheet (see example test_clinical_data.xlsx under inst/rawdata/test_data/test_clinical_data.xlsx) using the --clinical_info argument. In this spreadsheet, at least one of the following columns is expected: NEOADJUVANT REGIMEN, ADJUVANT REGIMEN, FIRST LINE REGIMEN, SECOND LINE REGIMEN or THIRD LINE REGIMEN, along with START and STOP dates of corresponding treatments. A subset of the TCGA pancreatic adenocarcinoma dataset is used as the reference cohort (--dataset TEST).

rnasum.R \
  --sample_name test_sample_WTS \
  --dataset TEST \
  --dragen_wts_dir $(pwd)/../rawdata/test_data/dragen \
  --report_dir $(pwd)/../rawdata/test_data/dragen/RNAsum \
  --umccrise $(pwd)/../rawdata/test_data/umccrised/test_sample_WGS \
  --save_tables FALSE \
  --clinical_info $(pwd)/../rawdata/test_data/test_clinical_data.xlsx \
  --save_tables FALSE

The HTML report test_sample_WTS.RNAsum.html will be created in the ../rawdata/test_data/stratus/test_sample_WTS_dragen_v3.9.3/RNAsum folder.

Output

The pipeline generates a HTML Patient Transcriptome Summary report and a results folder:

|
|____<output>
  |____<SampleName>.<output>.html
  |____results
    |____exprTables
    |____glanceExprPlots
    |____...

Report

The generated HTML report includes searchable tables and interactive plots presenting expression levels of altered genes, as well as links to public resources describing the genes of interest. The report consists of several sections, including:

Input data
Clinical information*
Findings summary
Mutated genes**
Fusion genes
Structural variants**
CN altered genes**
Immune markers
HRD genes
Cancer genes
Drug matching

* if clinical information is available; see --clinical_info argument
** if genome-based results are available; see --umccrise argument

Detailed description of the report structure, including result prioritisation and visualisation is available here.

Results

The results folder contains intermediate files, including plots and tables that are presented in the HTML report.

Code of Conduct

The code of conduct can be accessed here.

Citation

To cite package ‘RNAsum’ in publications use:

Kanwal S, Marzec J, Diakumis P, Hofmann O, Grimmond S (2024). “RNAsum: An R package to comprehensively post-process, summarise and visualise genomics and transcriptomics data.” version 1.1.0, https://umccr.github.io/RNAsum/.

A BibTeX entry for LaTeX users is

  @Unpublished{,
    title = {RNAsum: An R package to comprehensively post-process, summarise and visualise genomics and transcriptomics data},
    author = {Sehrish Kanwal and Jacek Marzec and Peter Diakumis and Oliver Hofmann and Sean Grimmond},
    year = {2024},
    note = {version 1.1.0},
    url = {https://umccr.github.io/RNAsum/},
  }