Version 3.4.0

Summary of changes:

• Fix bug where MM/ML tags are off in supplementary alignments
• Fix bug where the last base of a read is sometimes used in phasing
• Fix bug in variant calling where in some reads a wrong base is chosen between primary and supplementary alignments
• Fix bug where some alleles may include redundant haplotype names
• Fix bug where it's not possible to lower the minimum variant frequency for variant calling for individual target regions. This change affects opn1lw and ikbkg.
• Improve large deletion calling in reads. This change may affect all targets, particularly ikbkg and pms2.
• Rename haplotypes to label gene1 and gene2 for regions with fusion calling(GBA, CYP2D6, CYP11B1 and CFH/CFHR3)
• For smn1, consider more scenarios for adjusting SMN2 copy number
• For ncf1, adjust copy number for the scenario where three haplotypes are found and all are present at two copies.
• Update two copy haplotypes for CFHclust

Version 3.3.4

Summary of changes

• Fix f-string bug that causes python 3.11 or earlier to fail.

Version 3.3.3

Summary of changes:

• Fix bug that min_variant_frequency cannot be set lower than the default value
• Fix minor bug in ikbkg that causes program to error out
• Sort reads by name first to remove indeterminism in haplotype names
• Do not write VCF when region is clearly not homozygous but no haplotypes are phased (most likely due to low depth)
• Add the phase_region field in JSON output to report the coordinates of the analysis region and the genome build
• Minor improvement on indel detection
• Minor improvement on handling gene1_cn2, a scenario specified in the config asking Paraphase to assume a paralog group to always have two copies of gene1
• Improve documentation
  • Update NEB tutorial to clarify on the order of TRI1/2/3
  • Update README to clarify that the fusions_called field is only reported for four regions
  • Update the targeted data tutorial to include more details on PureTarget

Version 3.3.2

Summary of changes:

• Fix rare scenarios when program errors out in some low-depth regions. No algorithm change.
• Update license.

Version 3.3.1

Bug fix:

• Preserve MM/ML tags through minimap2 realignment instead of parsing MM/ML tags (Version 3.3.0 uses pysam to parse base modifications, but we have noticed some cases where pysam v0.23 crashes when parsing base modifications)

Version 3.3.0

Summary of changes:

• Improve phasing haplotypes into alleles, allowing (n-1)/1 scenario
• Do not adjust total_cn from 2 to 4 based on depth when calling fusions
• Fix rare bugs in depth-related analysis
• Update ALT alleles to . in the VCF for LowQual calls when ALT is equal to REF
• Add Ml/Mm tags in Paraphase bam for base modification information
• In Json output, rename hap_links to haplotype_links and rename linked_haplotypes to raw_alleles
• For targeted data
  • Improve copy number adjustment based on depth
  • Add option to assume a paralog group to always have more than one copy of gene1
  • Update command line options to use frequency-based parameters: --min-variant-frequency and --min-haplotype-frequency
• smn1
  • Enable smn1 analysis for CHM13-mapped data (Note that haplogroup assignment is not available with CHM13)
  • Fix bug with assigning haplogroups to smn2 haplotypes
  • Rename smn2_del78 haplotypes to smn_del78
• hba
  • Do not consider homology haplotypes during allele phasing
  • Better handle hba with targeted data, fixing problems with identifying homology haplotypes
  • Update genotype calls based on phased alleles
  • Report genome coordinates for 3p7 and 4p2 SVs in the sv_called field
• Minor update to strc and ncf1 for copy number adjustment based on depth
• Update f8 to reflect SV types in the haplotype name

Version 3.2.1

Summary of changes:

• Fix problem working with CRAM
• Other minor changes:
  • Add RN tag to output bam. RN stands for region name, indicating reads used to analyze one region (paralog group)
  • Sort some output fields in the JSON, making it consistent from run to run
  • Clean up reported alleles, filtering out cases where all haplotypes are linked into one allele, or not all haplotypes are included when two alleles are reported
  • Fix getting sample ID from bam header when there are blank spaces
  • Fix logic for writing homozygous haplotypes in VCF (no haplotypes phased -> no haplotypes phased and no heterozygous variant sites)
  • Update two_cp_haplotypes when adjusting total_cn from 2 to 4 based on depth
  • Use all reads instead of unique reads for variant calling at edges of clipped haplotypes

Version 3.2.0

Summary of changes:

1. Updates to better handle targeted data

• Filter reads on rq (>=0.99), if rq is present in input bam
• Add a --targeted option for targeted data to drop the assumption of uniform coverage across the genome
• Add two optional parameters for targeted data
  • --min-read-variant: Partially controls the number of supporting reads for a variant for identifying variants used for phasing. The cutoff for variant-supporting reads is determined by min(this number, max(5, depth*0.11)). Default is 20. At standard WGS depth, the default value is overwritten by max(5, depth*0.11).
    • Use cases: 1) Set this number low for low-coverage data or to increase sensitivity. 2) For targeted data with high coverage, set this number relatively high to avoid picking up sequencing errors and to reduce run time.
  • --min-read-haplotype: Minimum number of unique supporting reads for a haplotype. Default is 4. For targeted data with high coverage, this cutoff can be increased to reduce errors and to reduce run time.

2. Updates to target regions:

• Update coordinates of some target regions to include full genes whenever possible: pms2,ikbkg,hba,DDT,MBD3L2,DEFA1,PRY,CHRNA7,DHX40,GOLGA8A,IQCK,NXF2,OTOA,PDPK1,POTEI,RGPD1,RGPD3,RSPH10B,SIK1,TMLHE,CBS,KCNE1,CASTOR2,NBPF4,RGPD5,GOLGA8N,POTEB,ANKRD20A1,NSF
• Add TNXB as a region on its own so that the full gene can be genotyped (the RCCX region only includes part of TNXB)

3. Algorithmic changes

• Improve fusion calling in cases of homozygous deletion
• Add some homozygous sites to cover target regions evenly during phasing to improve read assignment to haplotypes and variant calling
• Update a few gene-specific callers
  • hba: Add calling of 4.2 deletion/duplication
  • smn1: If homozygous throughout region, default to CN =2 instead of 1; Drop carrier call if only one SMN1 haplotype is found but the total CN of SERF1A/B (neighboring locus) is larger than the total CN of SMN1/2
  • ikbkg: Improve calling of the 11.7kb deletion; Update the config to genotype the entire gene
  • ncf1: Drop carrier call if only one NCF1 haplotype is found but the total CN of GTF2I (neighboring locus) is larger than the total CN of NCF1 family
  • rccx: Better handle homozygous deletion cases
  • pms2: Update the config to genotype the entire gene

4. Other changes:

• Support cram as input
• Standardize haplotype naming across regions: {gene name}_{haplotype name}

Version 3.1.2

Summary of changes:

• Add --write-nocalls-in-vcf option to write no-call sites in the VCF

Version 3.1.1

Summary of changes:
Minor update. Fix program error in low-depth or no-data regions. Completes analysis even when the input is a small bamlet (result is still a no-call).