sage v3.4

Functional

• sync overlapping fragments on by default
• sync overlapping fragments logic: favor longer INDELs for CIGAR mismatch, forms consensus no matter how many bases are mismatching, ignore 3’ side of overlap where insert size < read length
• VCF writes AMQ (average map qual), ANM (average events per read) and MED (max edged distance)
• soft-filter average map qual for ref vs alt reads if different > 15
• average base qual filter lowered from 28 to 25
• new modMAPQ logic supports calling low mapq regions (default behaviour unchanged)
• added read strand bias filter
• added maxEdgeDistance filter
• AF filter can pass with lower AF if p-value condition met
• new deduplication logic for INDELs logic handles 1:M deduplication

Targeted panel mode, enabled with config 'high_depth_mode':

• ignore all reads with AD<30
• ignore all soft clip support
• added jitter AF filter
• BQR ignores overlapping bases

Bugs

• fragment strand bias calculated incorrectly

HTML Visualisations added for variants:

• creates a HTML for a subset of variants based on config. Shows read context and fragment support
• config 'vis_variants' to specify variants to produce visualisations, format: 'chromosome:position:ref:alt' separated by ';'. Will only run sage for +/- 200 base region around specified variants.
• config 'vis_pass_only' to generate files for passing variants, must be called with 'specific_regions'
• config 'vis_output_dir' default 'vis' if not specified

Config:

• sync_fragments removed since now on by default. Use 'no_fragment_sync' to disable.
• output -> output_vcf

purple v4.0

Functional:

• don’t fit short arms on 13,14,15,21,22
• Mask IG/TCR regions in fit
• Mask regions <2MB from centromere in fit
• Where ambiguous (low purity), BAF is now fit to minimise major allele CN
• chromosome X CN amplifications are called at 1.5x ploidy for males
• don’t smooth large germline deletions in diploid normalisation logic
• fit in 0.5% intervals for purity <20% and lower min to 7%
• write list of reportable transcripts to somatic VCF when alt-transcripts (eg CDKN2A) exist for a driver gene
• add LOH percent to QC output file

Technical:

• allow extra sample IDs in VCF and any order
• throw exception and exit on charting error

Bugs:

• ploidy was not calculated accurately in somatic mode

Visualisations

• only show diploid regions in input.png
• tumor is always in blue in input.png

Panel:

• add deviation penalty adjustment for GC ratio (config: gc_ratio_exponent, deviation_penalty_gc_min_adjust)
• set targeted panel default values:
  -ploidy_penalty_standard_deviation 0.10
  -ploidy_penalty_min 0.20
  -ploidy_penalty_sub_one_major_allele_multiplier 3.00
  -deviation_penalty_gc_min_adjust 0.25
  -gc_ratio_exponent 3.0
  -min_diploid_tumor_ratio_count 3
  -min_diploid_tumor_ratio_count_centromere 3

mark-dups v1.1

Functional:

• use unmap regions file to unmap problematic reads
• set NM attribute on consensus reads

Technical:

• various performance improvements for BAM writing

Config:

• unmap_regions - a TSV of regions to unmap, see HMF pipeline resources (v5.34) for current files
• bam_file -> input_bam
• multi_bam, samtools and sambamba paths for post-run sort, merge and index

lilac v1.6

Technical:

• tumor-only mode write fragment into tumor coverage columns instead of ref

Bugs:

• protect against deletion in read near end of coding region

Config:

• in tumor-only mode specify the tumor BAM file with -tumor_bam instead of -reference_bam
• removed 'write_all_files', instead use 'write_types selecting 1 or more from SUMMARY, FRAGMENTS, READS, REF_COUNTS

gripss v2.4

Technical:

• handle Gridss VCF with sample IDs in reversed order

Panel:

• added panel soft filters: qual-per-AD and modified AF
• set panel config by default if in target regions mode
  -hard_min_tumor_qual 200
  -min_qual_break_point 1000
  -min_qual_break_end 1000
  -filter_sgls
  -qual_per_ad 30
  -modified_af 0.03
  -modified_af_hotspot 0.005

gene-utils v1.1

Functional:

• expand Sage coverage to all driver genes
• added known fusion ref file generation routine
• added BED file validation and fix routine
• added general purpose liftover and probe creation routines
• coding region files can now be built from a gene name input file (config: gene_id_file)

Technical:

• improved region overlap check

Bugs:

• fixed rare phasing issues for proteome writer

cuppa v2.0

Major overhaul of CUPPA:

• Core classifier architecture of CUPPA is now layers of layers of logistic regressions. This is implemented in Python primarily using the scikit-learn library. In CUPPA v1, this was a series of formulas.
• Improved visualization
• New Java routine (CuppaDataPrep) to extract features from VCFs and TSVs that are passed to the Python component

cobalt v1.16

Functional:

• add referenceGcContent and tumorGcContent outputs
• use mean read depths instead of read counts, and change GC normalisation to use median instead of interpolated median

cider v1.0.3

Technical:

• handle unpaired reads

amber v4.0

Functional:

• default min map quality now 50 from 1
• disable min/max depth checks in target-regions mode
• min tumor default (config: tumor_min_depth) defaults to 25 for tumor/normal, 8 for tumor-only
• restrict sites to target regions in panel mode

Config:

• target_regions_bed - provide this file in targeted panel mode