StarSignDNA algorithm for mutational signature analysis which offers efficient refitting and de novo mutational signature extraction. StarSignDNA is capable of deciphering well-differentiated signatures linked to known mutagenic mechanisms and demonstrates strong associations with patient clinical features. The package offers a user- friendly interface and data visualization routines.

• Free software: MIT license
• Documentation: Link to be provided

Link to be provided

To install StarSign,

you can install the package directly from PyPi:

   pip install starsigndna

Alternatively,  you can install it via terminal by running this command::

   1. Download StarSign from https://github.com/uio-bmi/StarSignDNA
   2. Unzip StarSignDNA-master.zip
   3. cd StarSigndna-master/
   4. pip install -e .

To obtain help:

starsigndna --help

Usage: starsigndna [OPTIONS] COMMAND [ARGS]...

• --install-completion: Install completion for the current shell.
• --show-completion: Show completion for the current shell, to copy it or customize the installation.
• --help: Show this message and exit.

• count-mutation: Count mutation types in VCF file.
• denovo: Performs denovo Mutational Signatures analysis.
• refit: Mutational Signatures Refit Parameters

The refitting algorithm takes as input a mutational catalog and a COSMIC mutational signature file. The user can also specify signatures to be considered instead of using the full COSMIC matrix or a subset matrix:

starsigndna refit --help

Arguments

• matrix_file (TEXT): Tab separated matrix file [default: None] [required]
• signature_file (TEXT): Tab separated matrix file [default: None] [required]

Options

• --ref_genome (TEXT): Path to the reference genome [default: None]
• --n_bootstraps (INTEGER): Number of bootstraps [default: 200]
• --opportunity_file (TEXT): Path to the opportunity file [default: None]
• --numeric_chromosomes: If set, chromosome names are numeric [default: no-numeric-chromosomes]
• --no_numeric_chromosomes: If set, chromosome names are not numeric [default: no-numeric-chromosomes]
• --genotyped: If set, VCF file has genotype information for many samples [default: genotyped]
• --no_genotyped: If set, VCF file does not have genotype information for many samples [default: genotyped]
• --output_folder (TEXT): Path to the output folder [default: output/]
• --signature_names (TEXT): Comma separated list of signature names [default: None]
• --n_iterations (INTEGER): Number of iterations [default: 1000]
• --help: Show this message and exit

Running StarSignDNA refit:

starsigndna refit example_data/M_catalogue.txt example_data/COSMICv34.txt --output-folder /test_result --signature-names SBS40c,SBS2,SBS94
starsigndna refit example_data/tcga_coad_single.vcf example_data/sig_cosmic_v3_2019.txt --output-folder /output --signature-names SBS40c,SBS2,SBS94 --ref-genome

When the --signature-names option is used, the default number of signature is 3, but we recommend minimum of 5 signatures. The test data is provided in the example_data folder. To convert *.vcf to a matrix, the user must provide the path to the reference genome using the option --ref-genome.

The user can also provide the distribution of triplets in a reference genome/exome or normal tissue in the same patient (Opportunity matrix) using the option --opportunity-file human-genome/human-exome.

The de novo algorithm takes as input a mutational catalog and infers the exposure matrix and mutational signature matrix. The COSMIC mutational signature file is provided to compute the cosine similarity:

starsigndna denovo --help

Arguments

• matrix_file (TEXT): Tab separated matrix file [default: None] [required]
• n_signatures (INTEGER): Number of signatures to identify [default: None] [required]

Options

• --lambd (FLOAT): Regularization parameter [default: 0.7]
• --opportunity-file (TEXT): The distribution of triplets in a reference 'human-genome' or 'human-exome' or normal tissue [default: None]
• --cosmic-file (TEXT): Tab separated cosmic file [default: None]
• --numeric-chromosomes: If set, chromosome names are numeric [default: no-numeric-chromosomes]
• --no-numeric-chromosomes: If set, chromosome names are not numeric [default: no-numeric-chromosomes]
• --genotyped: If set, VCF file has genotype information for many samples [default: genotyped]
• --no-genotyped: If set, VCF file does not have genotype information for many samples [default: genotyped]
• --max-em-iterations (INTEGER): Maximum EM iterations [default: 100]
• --max-gd-iterations (INTEGER): Maximum GD iterations [default: 50]
• --file-extension (TEXT): File extension [default: None]
• --ref-genome (TEXT): Path to the reference genome [default: None]
• --output-folder (TEXT): Path to the output folder [default: output/]
• --help: Show this message and exit

Step 1: Grid Search: The grid uses cross-validation to find the optimal pairwise (k and λ) by going to the snakemake folder and opening the running file (Snakefile) to check all the paths and input files:

cd snakemake
vi Snakefile

Step 2: In the Snakefile, provide the range of the number of signatures k and λ for the grid search to determine the optimal k and λ:

localrules: all
ks = list(range(2, 10)): default range of the number of signatures
lambdas = [0, 0.01, 0.05, 0.1, 0.2]: default range of λ

Input mutational catalogue needs to be provided in the dataset folder:

rule test_train_split:
    input: "results/{dataset}/M_catalogue.txt"

Running the grid search:

snakemake -j num_cpu

To check manually the optimal k and λ from the output:

sort -k3n,3 results/data/all.csv

Run denovo using optimal k=4 and λ=0.1:

starsigndna denovo snakemake/results/data/M_catalogue.txt 4 0.1 --cosmic-file example_data/COSMICv34.txt --output-folder /test_result

Maintainer Name - chrisbop@uio.no or christianbope@gmail.com