CHANGES IN VERSION 1.34.1
-------------------------
o fix the compiler warning: sprintf is deprecated
o show "64-bit" correctly when run on Windows
CHANGES IN VERSION 1.34.0
-------------------------
o fix the compiler issue on Mac M1 chip
o fix the C stack error in RStudio
CHANGES IN VERSION 1.32.0
-------------------------
o fix the issue on Win32 because of using deprecated
tbb::task_scheduler_init
CHANGES IN VERSION 1.30.2
-------------------------
o require GCC >= v8.0 for compiling the AVX-512VPOPCNTDQ intrinsics
o fix `hlaGDS2Geno()` when loading a SeqArray GDS file
CHANGES IN VERSION 1.30.0
-------------------------
o add the support of Intel AVX-512VPOPCNTDQ intrinsics (faster than
AVX512BW)
CHANGES IN VERSION 1.28.0
-------------------------
o `hlaPredict()` returns the dosage of HLA alleles when
type="response+dosage", and `hlaPredict()` returns the best guess and
dosages by default
o a new option "Pos+Allele" in `hlaPredict()`, `hlaGenoCombine()`,
`hlaGenoSwitchStrand()`, `hlaSNPID()` and `hlaCheckSNPs()` for matching
genotypes by positions, reference and alternative alleles; it is
particularly useful when the training and test set are both matched to
the same reference genome, e.g., 1000 Genomes Project
o `hlaGDS2Geno()` supports SeqArray GDS files
o a new option 'maf' in `hlaAttrBagging()` and `hlaParallelAttrBagging()`
o 'pos.start' and 'pos.end' are replaced by 'pos.mid' in `hlaFlankingSNP()`
and `hlaGenoSubsetFlank()`
o new function `hlaAlleleToVCF()` for converting the imputed HLA classical
alleles to a VCF file
CHANGES IN VERSION 1.26.1
-------------------------
o the hlaAttrBagging object can be removed in garbage collection without
calling `hlaClose()`
o enable internal GPU API
o improved multithreaded performance compared with v1.26.0
CHANGES IN VERSION 1.26.0
-------------------------
o users can interrupt the model building in an interactive R session
o remove `hlaErrMsg()` since it is never used
o a new option 'nthread' in `hlaAttrBagging()` as a complement to
`hlaParallelAttrBagging()`
o kernel version 1.5: generates the same training model as v1.4,
but 2-6x faster, by taking advantage of Intel AVX, AVX2 and AVX512
intrinsics
o new function `hlaSetKernelTarget()` to automatically select the CPU
target the algorithm is optimized for
CHANGES IN VERSION 1.24.0
-------------------------
o fix a Win32 compiler issue
CHANGES IN VERSION 1.22.0
-------------------------
o change the file name "vignettes/HIBAG_Tutorial.Rmd" to
"vignettes/HIBAG.Rmd", so `vignette("HIBAG")` can work directly
CHANGES IN VERSION 1.18.0
-------------------------
o new function `hlaDistance()`
o `match.type="Position"` is the default option in `hlaPredict()`
CHANGES IN VERSION 1.16.0
-------------------------
o KIR information in `hlaLociInfo()`
o new functions `hlaGenoSubsetFlank()` and `hlaLDMatrix()`
CHANGES IN VERSION 1.14.0
-------------------------
o modify the kernel to support the GPU extension
o add matching proportion to measure the similarity of SNP haplotypes
between training and test datasets
o new function `hlaReportPlot()`
o the argument 'cl' in `predict.hlaAttrBagClass()`, `hlaPredict()` and
`hlaParallelAttrBagging()` allows a numeric value for the number of cores
CHANGES IN VERSION 1.12.0
-------------------------
o change "hg20" to "hg38" according to the UCSC Genome Browser datasets
and documentation
o add "DRB3" and "DRB4" to the HLA gene list
CHANGES IN VERSION 1.10.0
-------------------------
o the version number was bumped for the Bioconductor release version 3.4
CHANGES IN VERSION 1.9.0-1.9.3
-------------------------
o the development version
CHANGES IN VERSION 1.8.0-1.8.3
-------------------------
o the version number was bumped for the Bioconductor release version 3.3
o new arguments 'pos.start' and 'pos.end' in `hlaFlankingSNP()`
CHANGES IN VERSION 1.7.0-1.7.7
-------------------------
o new functions `hlaCheckAllele()`, `hlaAssocTest()`,
`hlaConvSequence()` and `summary.hlaAASeqClass()`
CHANGES IN VERSION 1.6.0
-------------------------
o the version number was bumped for the Bioconductor release version 3.2
CHANGES IN VERSION 1.5.0-1.5.1
-------------------------
o increase the test coverage
o fix an issue when `predict.hlaAttrBagClass(..., allele.check=FALSE)`
CHANGES IN VERSION 1.4.0
-------------------------
o the version number was bumped for the Bioconductor release version 3.1
CHANGES IN VERSION 1.3.0-1.3.2
-------------------------
NEW FEATURES
o support the human genome "hg20"/"hg38"
o add a new function `hlaGDS2Geno()` to support SNP GDS files (in the
R/Bioconductor package SNPRelate)
o `hlaReport()` outputs text with markdown format
SIGNIFICANT USER-VISIBLE CHANGES
o optimize the calculation of hamming distance using SSE2 and hardware
POPCNT instructions if available
o hardware POPCNT: 2.4x speedup for large-scale data, compared to
the implementation in v1.2.4
o SSE2 popcount implementation: 1.5x speedup for large-scale data,
compared to the implementation in v1.2.4
BUG FIXES
o bug fixes on big-endian machines (like Solaris SPARC, Apple PowerPC)
o minor fix on random sampling from discrete uniform distribution
o bug fix if `requireNamespace("HIBAG")` instead of `require(HIBAG)` is
called from other packages
CHANGES IN VERSION 1.2.1-1.2.4
-------------------------
NEW FEATURES
o add a new function `hlaOutOfBag()` offering out-of-bag estimation of
overall accuracy, per-allele sensitivity, etc
o improve the display of the function `hlaReport()`
SIGNIFICANT USER-VISIBLE CHANGES
o improve the display of `summary.hlaAlleleClass()`
o fix warnings in 'R-3.1.1 CMD check'
o avoid missing values in `hlaFlankingSNP()`
o add more gene information in `hlaLociInfo()`
BUG FIXES
o make the C function 'HIBAG_SortAlleleStr' compatible with the
clang++ compiler on Mac Mavericks
CHANGES IN VERSION 1.2.0
-------------------------
NEW FEATURES
o add a new function `hlaPredMerge()` to merge predictions from multiple
models via voting
o add assembly information (like hg19) to the objects of
'hlaAttrBagClass' and 'hlaAttrBagObj'
o add a new argument 'stop.cluster' to `hlaParallelAttrBagging()`
o allow matching SNPs by positions only
o allow imputing HLA alleles in parallel at a single locus
SIGNIFICANT USER-VISIBLE CHANGES
o optimize the C code: replace some intensive 'exp' and 'log'
calculations by precalculated values, roughly 2.5x ~ 4x speedup
o optimize the C code of the hamming distance calculation between
genotypes and two haplotypes by SSE2 intrinsics, ~50% faster
for large datasets
o the code optimization guarantees the same training and predictive
accuracies as version 1.1.0
o remove unused row names in the function `hlaCompareAllele()`
o improve the function `hlaPublish()`
o improve vignettes and documents
CHANGES IN VERSION 1.1.0
-------------------------
NEW FEATURES
o add genome assembly information (such like hg18, hg19)
o allow class majority voting in `predict.hlaAttrBagClass()`
o add an autosave function to `hlaParallelAttrBagging()`
o add a new function `hlaPublish()`
o allow additional information in the objects of 'hlaAttrBagClass' and
'hlaAttrBagObj'
SIGNIFICANT USER-VISIBLE CHANGES
o update citation information
o change `hlaModelfromObj()` to `hlaModelFromObj()` (a typo)
o improvement in mismatching alleles for allelic strands
CHANGES IN VERSION 1.0.0
-------------------------
SIGNIFICANT USER-VISIBLE CHANGES
o increase the limit of the possible number of SNPs in a single
classifier up to 256
BUG FIXES
o use `useDynLib()` in NAMESPACE instead of 'library.dynam()' since
R-Ext strongly suggests to use `useDynLib()`
o update citation information
o add a vignette
CHANGES IN VERSION 0.9.2-0.9.4
-------------------------
NEW FEATURES
o add a load balancing function to `hlaParallelAttrBagging()`
o add `hlaUniqueAllele()`
o add an argument 'match.pos' to the functions `predict.hlaAttrBagClass()`,
`hlaGenoCombine()`, `hlaGenoSwitchStrand()`
o add an argument 'with.pos' to the function `hlaSNPID()`
SIGNIFICANT USER-VISIBLE CHANGES
o give more warnings about missing SNPs in `predict.hlaAttrBagClass()`
o a warning is given for strand mismatching when calling
`hlaGenoSwitchStrand()`
BUG FIXES
o bug fix: passing only one individual to `predict.hlaAttrBagClass()`
o avoid the segmentation fault in `predict.hlaAttrBagClass()`
when all SNPs are missing.
CHANGES IN VERSION 0.9.1
-------------------------
o first release of HIBAG in July, 2012