NEWS

decompTumor2Sig 2.13.1 (2022-05-09)

Fix: corrected recognition of EstimatedParameters (from package pmsignature) objects obtained with the use of a background signature.
Fix: corrected internal scripts for production of external example data.

Fix: changed example for adjustSignaturesForRegionSet() to limit memory usage (previous example produced error during check on Windows for arch 'i386').
Fix: made sure that the extension of genomic regions by half the sequence pattern (needed for the adjustSignaturesForRegionSet() function) does not result in out-of-bounds regions.

Updated readAlexandrovSignatures() to read the COMSIC signature format v3.2 (published in March 2021).

Updated readAlexandrovSignatures() to add the possibility to read COSMIC signatures of version 3.1 directly from an Excel file (as provided on the COSMIC website).
Added possibility to adjust/normalize mutational signatures to specific subsets of the human genome (defined by means of GRanges objects). The adjustment/normalization is performed accoring to the nucleotide frequencies in the specified regions (with respect to nucleotide frequncies in the reference sequences, e.g., the whole genome, for which the signatures were derived in the first place).

Removed dependency of package CRAN vcfR (archived on 2020-07-05), using functions of Bioconductor package VariantAnnotation instead
Improved the mutation filtering so that multiallelic SNVs aren't excluded when loading tumor genomes from a VCF file
Updated citation and affiliation information
Added consistency check for reference genome and genome annotation
Improved error messages

adapted readAlexandrovSignatures to read the file format used by the COSMIC mutational signatures version 3 (May 2019; Single Base Substitution/SBS signatures only)

Changed the procedure which converts lists of individual mutations from MPF files into tumor mutation profiles to reduce the memory footprint and allow for larger sets of tumors to be read from file.
When genomes are read from files: remove genomes without SNVs (no mutation frequencies)
Recognize indels for which either the REF or ALT base is specified as "-"
Fix bugs: o Error when quadprog returned a mutation frequency of minimally larger than 1 (which is theoretically impossible but can probably happen due to the rounding of floating point numbers)

When transcription direction is taken into account: exclude mutations in regions with overlapping genes of opposing transcription directions by default! Previous versions of decompTumor2Sig took the approach of pmsignature, using the transcription direction of the first gene encountered in the transcript database (which is rather arbitrary); excluding these mutations appears more appropriate. The old approach can still be used with an additional function parameter.
Removed pmsignature from the suggested packages (not allowed in Bioconductor)
Reduced the number of tumor genome examples in extdata for faster processing and smaller package size (six out of 21 tumors from PMID:22608084)

Implemented own signature plotting to remove code dependency from pmsignature
Decoupled the following data conversions functions from pmsignature code: o getGenomesFromMutFeatData() o getSignaturesFromEstParam()
Moved some common code to a new internal function for ease of maintenance.
Fixed bugs: o Error when some sequence names where not found in the reference genome, e.g., due to different names of decoy sequences. o Minor problems with unlikely signatures composed of only the mutated base (no flanking bases)

Fixed a minor bug in mapSignatureSets() that occurred when the two signature sets had the same size.
Added functions to verify the format of signatures, genomes and exposures: o isAlexandrovSet(), isShiraishiSet(), isSignatureSet() o sameSignatureFormat() o isExposureSet()
Moved description of deprecated BiocInstaller in vignette to BiocManager.

Updated DESCRIPTION file o replaced Author and Maintainer by Authors@R o added URL for BugReports
Changed from manual NAMESPACE and *Rd files to a creation by roxygen2
Changed function names for reading data from "load..." to "read..."
Shortened the longest function names o getGenomesFromMutationFeatureData -> getGenomesFromMutFeatData o getSignatureListFromEstimatedParameters -> getSignaturesFromEstParam
Used robust sequence creation with seq, seq_len, seq_along instead of 1:N etc
Consistent use of native class checking such as is.numeric()
Implemented wrapper functions for external classes which do not provide the necessary accessor functions
Replaced instances of sapply() and unlist(lapply()) by vapply()
Updated vignette and made most code chunks runnable
Added greedy search option to plotExplainedVariance (significant speed-up)