New features

• More options to tune sensitivity using --sd (span-distance, used for insertion clustering) and --search-depth options
• Ability to filter on mapq when using dysgu-filter
• --cohort-update option for dysgu merge, allowing regenotyped calls to be updated in a cohort file
• More efficient merging of large cohorts - option for limiting number of local SV comparisons using dysgu-merge via --max-comparisons
• Improved somatic filtering
• Slightly improved runtime and memory footprint
• Support for python 3.12
• --verbose option for dysgu test
• Added MAPQS to format line (mapq supplementary-reads)
• Added NMB to format line (Edit distance, ignoring large gaps and soft-clips)

Bug Fixes

• Missing usage of bam file tell when using dysgu-call (improves sensitivity slightly)
• Minor clustering bug (improves sensitivity slightly)
• #81 #83

Minor release

• Added option to use a file list for filtering normals using dysgu filter input.vcf list_of_normals.txt > output.vcf
• Bug fix: using -o option with a / in file name, and combined with bam file stdin, now works again

Improvements / changes

• Added --symbolic-sv-size option to switch between symbolic or sequence representation
• Added parallel processing for merge command
• Added --ignore-sample-sites option to ignore the input sample when using --sites. This is set at True by default and addresses https://github.com/kcleal/dysgu/issues/37
• Minimum python version now 3.8

Bug fixes

• Fixed https://github.com/kcleal/dysgu/issues/69 https://github.com/kcleal/dysgu/issues/70
• Fixed minor issue resulting in dropping of paired-end INS SVs

Improvements / changes

• Added 'auto' option for setting --min-support, this is set as default for long-reads
• Long-reads default mapq lowered to 1
• Improved sensitivity for large-insertions using long-reads
• Small numbers of BND events also now reported for long-reads
• Improved precision due to better SV merging and lower numbers of duplicated true-positive calls
• Coverage values now include reads < mapq threshold
• Around 10 % of insertions are now re-classified as DUP; occurs when split-read support is available

Bug fixes

• Segfault for high coverage sample
• Exponential runtime for very large component in graph

v1.5.0

• Significantly improved precision for long read SV calling
• Long-reads with poorly mapped ends are now ignored, controlled using --divergence option
• Nearby indels on the same read are now merged, controlled the --length-extend option
• CLI change: filter-normal command has been renamed filter and supports more options for filtering
• Improvements in runtime for long-reads

Improvements

• Better handling of supplementary reads - slight improvement in SV calling across the board, especially complex events and insertions (+1-2% F1 score)
• Improved 'filter-normal' command for long reads (tumor-normal SV calling), plus updated CLI

Fixes

• Avoid over counting supplementary reads
• More error handling/checking

Improvements

• Added filter-normal command for somatic SV filtering.
• Support for Python 3.11 (linux)
• Run/call command added to vcf header

Bug fixes

• Fixed over counting of within-read SVs from same read
• Prevent clustering of same-read events
• Fixed supplementary clustering issue

Minor release with features:

• dysgu merge now has a default --collapse-nearby True option for better merging of nearby SVs
• added support for building using conda htslib using python setup.py install --conda-prefix $PREFIX or setting PREFIX as an environment variable

Bug fix:

• py_call_code error

Minor release with bug fixes:

• compatibility for numpy >= 1.22
• index error for adding --sites
• type error during graph construction https://github.com/kcleal/dysgu/issues/51

Download source code using the dysgu.tar.bz2 link below, otherwise htslib will be missing.

Minor bug fixes:

• GC type error when merging at population level
• --regions-only True now works as expected
• Contigs are now kept in file when merging

Download source code using the dysgu.tar.bz2 link below, otherwise htslib will be missing.