esvee v1.0-rc.4

Functional:

• Prep write more detail on discordant stats and write max remote group frags for discordant-only junctions
• Assembly adjusts min fragment rates to load junctions: +1 for split junctions, +3 or discordant-only junctions per discordant rate percentage

Technical:

• optionally specify 'esvee_prep_dir' for inputs for assembly and caller
• removed config 'fragment_length_file' config

esvee v1.0-rc.3

Description of release esvee v1.0-rc.3

Full Changelog: esvee-v1.0-rc.2...esvee-v1.0-rc.3

esvee v1.0-rc.2

Config:

• optionally set 'frag_length_file' for Assembly and Caller, otherwise continues to assume file is named TUMOR_ID.esvee.prep.fragment_length.tsv

sage v4.0-rc.2

Technical:

• expand BAM sampling for read length to test all panel regions with buffer

teal v1.3.2

Technical:

• uses BamMetrics summary file

virus-interpreter v1.7-rc.2

Technical:

• Re-trigger failed docker image build

virus-interpreter v1.7-rc.1

Technical

• uses BamMetrics summary file

v-chord v1.0

Initial release

sage v4.0-rc.1

BQR

• Round consensus BQ to canonical quals and 1
• Increase min mapping quality for BQR from 10 to 50, and change AD/AF site thresholds
• Handle contexts with 0 identified errors
• BQR by consensus type
• Update error rate calculation to be robust to different qual distributions per-base

Read context & evidence counting

• New read context algorithm
• Changes to read context annotation
• New read counting conventions
• Approximate ref matching logic for long inserts
• Allow SNV/MNV support in soft-clips
• Fragment sync prefers aligned over non-aligned base quals

Qual model

• Qual filter dynamically scales with depth and splits map and base qual tests
• Explicit germline mode which incorporates empirical VAF implausibility into model
• Novel indel qual is conditionally boosted
• Changes to high depth mode
• Modify edge penalty more in line with realised error rates, and facilitates partial core qual contribution
• Use min base qual rather than mean base qual for MNVs
• Indel qual for MSI repeats is based on microsatellite empirical qual
• Jitter penalty model is overhauled

Filtering

• MED based on distance to soft clip instead of distance to end of read
• MED uses max per-read edge distance for fragments with overlap
• MED uses max edge distance across depth-supporting reads, not just ref
• MAX_READ_EDGE_DISTANCE_PERC split between panel and non-panel regions
• Fix dedupMnv bug
• Filter if ALT average fragment length << REF average fragment length
• Filter if portion of realigned read support for short indel is unusually high
• Filter at candidate phase if only one distinct fragment provides support
• Min fragment coordinate check is stepped and applies all the time
• Tumor VAF filter now uses recalibrated qual

Config:

• msi jitter files mandatory unless 'skip_msi_jitter' specified. Uses BAM path if 'jitter_param_dir' isn't specified

redux v1.0-rc.1

Functional

• write MSI jitter files for Sage
• unmap secondaries and drop from BAM
• use median base qual from selected consensus base
• no dual fragment marking if not duplex UMIs
• drop supplementaries with alignment score < 30 (BWA default)

Technical

• handle hard-clipped reads
• minor memory and performance optimisations

Bugs
-unmap read in unmap region if map becomes unmapped and vice versa

• rare repeated writing of supplementaries