- Fixed a bug where empty haplotypes were treated as valid candidates and then reported as part of the diplotype
- Revised inferred chain logic to remove erroneous inferrences and add missing inferrences
- Fixed a bug where a poorly mapping read in an HLA reigon would fail to match any alleles in the database and cause a panic
- The underlying methodology has been significantly altered to improve CYP2D6 diplotyping in targeted sequencing
- The core CYP2D6 regions have been shrunk to contain just the regions containing variants that define the allele
- Additional regions have been added solely for the purpose of linking CYP2D6 alleles: REP6, REP7, spacer, and "link_region" (region between CYP2D6 and CYP2D7 typically)
- The chaining algorithm has been altered to account for the additional regions above. A "normal" haplotype chain is expected to have the following order of regions in GRCh38: REP6 -> CYP2D6 -> link_region -> REP7 -> spacer -> CYP2D7
- The scoring of diplotype chain pairs is now based on a unified scoring scheme that accounts for: 1) edit distance of observations to the chain pair, 2) likelihood of the chain pair based on allele coverage and multinomial, 3) lasso penalty for duplications, and 4) penalty for unexpected chain events (see above "normal" chain)
- The debug BAM will now output a haplotype block for each identified region (e.g., REP6 will have its own block)
- The CLI has been modified such that the VCF file is now optional. If a VCF file is not provided, all variant-based diplotyping will be skipped and those genes will be absent from all output files. If no VCF or BAM files are provided, pb-StarPhase will generate a user error message.
- Two new options control the genes that are diplotyped:
--include-setand--exclude-set. Only one of these options can be specified at a time. Both accept a plain text file with one gene name per line. If--include-setis specified, then only the genes in the given file will be diplotyped. If--exclude-setis specified, then all genes will be diplotyped except the ones in the given file. - Underlying CYP2D6 consensus algorithm was updated for greater compute efficiency
- Updated the crate for CYP2D6 generation, improving alignment offset seeding and reducing over-splitting for shorter reads (e.g., targeted sequencing)
- Added a secondary penalty for using a non-unique chain pair during the CYP2D6 chaining step. This tends to reduce errors caused by ambiguous chains in WGS datasets.
- Fixed an issue with CYP2D6 where an "UNKNOWN" cluster could erroneously generate an extra allele when two defined alleles came from the same HPC-cluster
- Fixed an issue with CYP2D6 where alleles excluding from normalization could form their own unpenalized haplotype (e.g., a *68 allele by itself)
- Fixed a panic caused by typing an empty CYP2D6 sequence
- Fixed a panic caused by no usable reads in the CYP2D6 search region
- Added support for calling CYP2D6 from targeted sequencing data
- In general, accuracy for targeted datasets is less than that of WGS. This is largely due to difficulties with capture that lead to decreased coverage of hybrid or duplicated alleles.
- We recommend using two additional parameters when using targeted sequencing data:
--infer-connections --normalize-d6-only
- Added two new CLI options to support targeted sequencing datasets:
--infer-connections- If set, pb-StarPhase will infer allele connections that are not observed in the dataset but common in the population. For example, *4 and *68 are commonly found together, as are *10 and *36. This option is recommended when reads are too short to directly span from one allele to the next.--normalize-d6-only- If set, pb-StarPhase will only normalize the copy numbers using the CYP2D6 alleles (i.e., excluding any discovered CYP2D7 alleles). This option is recommended when coverage of the CYP2D7 alleles is inconsistent relative to the CYP2D6 alleles.
- Fixed a reporting issue in the PharmCAT TSV where brackets were missing from combination alleles
- The CLI settings log output has been updated for easier human readability
- Exposed three new CLI options that influence how the CYP2D6 algorithm works:
--min-consensus-fraction,--min-consensus-count, and--dual-max-ed-delta. Most users should not need to modify the defaults.
- Improved the allele chaining algorithm to reduce errors with hybrid CYP2D6 allele copy-number counts
- Fixed the CYP2D6 star allele output to be in coding order (i.e. reverse relative to before)
- Fixed the PharmCAT TSV output to use simple CYP2D6 representation
- Added the ability to store CYP2D6 variant information from PharmVar in our database
- Released an updated database with CYP2D6 information
- Added the ability to call CYP2D6 using the updated database and an aligned BAM file for WGS datasets
- Added a CLI option to output realigned CYP2D6 segments to a small BAM file for IGV viewing and debugging
- The version of pb-StarPhase is now recorded as the
pbstarphase_versionin the output JSON
- Fixed a bug where failure to align an HLA allele to a read would cause an unhandled panic
- Fixed a bug where the "*" was not pre-pended to HLA alleles in the output files
- Added the ability to store HLA haplotype sequences from IMGTHLA in our database
- Added the ability to read and write from a gzip-compressed (.gz) database file
- Released an updated database with HLA haplotype sequences
- Added the ability to call HLA-A and HLA-B using an HLA-aware database file and an aligned BAM file
- Added several CLI parameters to support HLA calling
- Fixed a bug where the previous binary would segfault when trying to run
pbstarphase build - Pre-compiled static binary build system has changed to be Docker-based
Initial release.