SignatureSets: An R Package for RNA-Based Immuno-Oncology Signatures

Introduction

The SignatureSets package provides access to a curated and extensive compendium of RNA-based Immuno-Oncology (IO) signatures. All signatures included in this package are:

• Published in peer-reviewed literature.
• Publicly available through trusted repositories and resources.

References to the original publications and resources for each signature are included in the package documentation, refer to the vignettes or accessible via the web application predictio.ca.

Setup

The latest version of SignatureSets repository can be found on the SignatureSets GitHub repository. To set up the repository, please download this folder locally:

git clone https://github.com/bhklab/SignatureSets
cd SignatureSets

Signature Association Types

In total, 55 Immuno-Oncology (IO) gene signatures have been curated in the SignatureSets package. These signatures were extracted from the literature and manually annotated using GENCODE version 40, with HUGO Gene Symbols as the primary identifiers linked to Entrez Gene IDs and Ensembl Gene IDs. The signatures are categorized based on their association with IO therapy outcomes:

• 65% (36 signatures): Associated with sensitivity to IO therapy, indicating potential positive responses such as immune activation or enhanced checkpoint inhibitor efficacy.

• 35% (19 signatures): Associated with resistance to IO therapy, highlighting mechanisms like immune evasion, suppressive tumor microenvironments, or other resistance pathways.

Methods for Computing IO Signature Scores

Signature scores are computed using standardized methods tailored to the characteristics of each signature, as described in their original publications.

• Unweighted Signatures: Scores are computed using Gene Set Variation Analysis (GSVA) or Single Sample Gene Set Enrichment Analysis(ssGSEA) to assess pathway enrichment. GSVA calculates enrichment scores for gene sets without weighting individual genes.

• Weighted Signatures: Scores are computed as a weighted mean expression, where weights are assigned as follows: +1 for increased expression and -1 for decreased expression.

• Specific Algorithm: Certain signature scores are computed based on their respective original publications, e.g., the PredictIO signature.

More details about signature score computations and their applications can be found on the PredictioR GitHub repository

Citation

If the data from the SignatureSets package are used in your publication, please cite the following paper(s):

• Bareche, Y., Kelly, D., Abbas-Aghababazadeh, F. et al., Annals of Oncology 2022.