Mutational Signatures
Human Cancer v3.5 Experimental v1.0SigProfilerAssignment
About the SigProfilerAssignment tool
SigProfilerAssignment is a novel tool developed by the Alexandrov lab at UC San Diego. The tool is used to assign previously known mutational signatures to individual sequenced samples, utilising a custom implementation of the forward stagewise algorithm and nonnegative least squares.
SigProfilerAssignment outperformed other common approaches in a benchmark using 2,700 synthetic cancer samples. More information can be found in our preprint.
Our guided tour of SigProfilerAssignment demonstrates how to assign reference mutational signatures to individual samples.
Using SigProfilerAssignment
Input
Input data for SigProfilerAssignment can be provided by mutational matrices, mutation calling, and segmentation files.
Standard mutational matrices can be used for any variant class, with rows corresponding to mutational channels and columns to samples. Mutational matrices can be generated using the SigProfiler suite of tools.
Mutation calling files are exclusive for the analysis of single base substitutions (SBS), doublet base substitutions (DBS), and small insertions and deletions (ID). SigProfilerAssignment supports common mutation calling files formats, including the Variant Call Format (VCF), the Mutation Annotation Format (MAF), as well as ICGC and simple text files. Example data can be found here.
Multi-sample segmentation files obtained from ASCAT, ABSOLUTE, Sequenza, FACETS, Battenberg, or PURPLE are supported for the analysis of copy number variations (CN).
Different species (human, mouse, and rat), genome builds (GRCh37/38, mm9/10, and rn6), COSMIC reference signatures versions, and sequencing assays (whole genome sequencing, whole exome sequencing, and targeted sequencing) are supported.
Output
The main output of SigProfilerAssignment includes the activity of each known mutational signature for each of the supplied samples and the reconstruction of the original dataset. Signature activities correspond to the specific numbers of mutations from the original catalog caused by a particular mutational process. Considering these activities, as well as the provided set of COSMIC reference mutational signatures, a reconstruction of the original mutational catalog for each sample is derived. Different accuracy metrics for this reconstruction are outputted by SigProfilerAssignment, including cosine similarity, Kullback–Leibler divergence, Pearson correlation, L1 relative error, and L2 relative error.
The signature assignment results are summarized using an interactive heatmap plot, including the signatures’ activities and the samples’ reconstruction accuracies (based on the cosine similarity metric). Also, an individual reconstruction plot per sample is provided, which includes the mutational profiles for both the original and the reconstructed input sample, different accuracy metrics, and the mutational profiles for each of the known mutational signatures assigned to that sample. In addition, general visualizations of the mutational profiles and tumor mutational burden (TMB) of the input samples are provided. Numerical data files containing activities and reconstruction metrics can also be downloaded.
How do I cite SigProfilerAssignment?
Díaz-Gay et al. 2023 Bioinformatics and Tate et al. 2019 Nucleic Acids Research
How are mutational signatures assigned to individual samples?
Assigning previously known mutational signatures to individual samples, also known as refitting of known signatures, enables identifying the set of operative mutational signatures in a particular sample and quantifying the number of mutations attributed to each signature found in that sample.
This set of signatures corresponds to the different mutational processes that have been active in a particular tumour or somatic tissue during an individual's lifetime, including both endogenous and exogenous processes. Leveraging the vast amount of DNA sequencing data accumulated in recent years, these mutational processes have been characterised, resulting in reference signatures which are recorded in COSMIC Mutational Signatures.
Making use of these reference mutational signatures, SigProfilerAssignment provides an accessible web platform to assign these signatures to cancer samples and elucidate the biological processes responsible for the accumulated somatic mutations.