Skip to content

Latest commit

 

History

History
156 lines (129 loc) · 5.18 KB

CHANGELOG.org

File metadata and controls

156 lines (129 loc) · 5.18 KB
  • v0.4.9
    • use C+17 standard
    • breaking: --verbose takes integer value as the level of verbose
    • new short option: -f
  • v0.4.8
    • HWE test accounting for population structure via --inbreed 1
      • BEAGLE file can work with in-core mode
      • PLINK file can work with both in-core and out-of-core mode
    • new option --USV as the prefix of pcaone .eigvecs/.eigvals/.loadigns/.mbim files
    • simplied usage because both --project and --inbreed can work together with --USV
    • the output .eigvecs2 stores the eigenvectors of the covariance matrix, which also works for BEALGE file with genotype likelihood input.
    • the .eigvecs,.eigvals,.loadings stores the U,S,V matrix for reconstructing the PI matrix
  • v0.4.7
    • breaking: rename option --ld-bim as --match-bim
    • projection support via --project
      • 1: projection by multiplying the loadings with mean imputation for missing genotypes
      • 2: projection by solving the least squares system $Vx=g$. sites with missingness are skipped like smartPCA.
    • new options: --read-U, --read-V, --read-S
    • can do LD for subset plink file
  • v0.4.6
    • give warnings when sample standard deviation is zero.
    • checks unknow options in CLI
    • fix parsing bim file ending with empty line
    • adjust memory estimator for LD
  • v0.4.5
    • can do LD prunning and clumpint out-of-core
    • add --ld-bims, --print-r2 options
  • v0.4.4
    • add --clump-names option
  • v0.4.3
    • new output eigvecs2 for plink input
    • fix ld clumping and sort snps by pvalues
  • v0.4.2
    • fix slowness due to the ld logics
    • add ld-clump functions
  • v0.4.1
    • fix ld windows
    • keep sites with higher maf in prune.in
    • support standard pruning using --svd 0 when PCAone is compiled with DEBUG=1
    • add --ld-snps to output pairwise r2 for given SNPs
  • v0.4.0
    • add ld pruning
    • fix --maf being af not maf
  • v0.3.9
    • pre-lease of ld feature
    • verbose cli check
  • v0.3.4
    • fix the order of SNP loadings for plink input
  • v0.3.3
    • breaking change option --svd for choosing different SVD methods. default is PCAone algorithm2.
    • add permuting BGEN file with multithreading.
    • add Full SVD support by --svd 3
    • add log transformation support for bulk RNA-seq data by --scale 1
    • add binary file support
    • recover the original order for SNP loadings.
    • recover the eigenvalues for diploid genotype (0,1,2) data.
  • v0.3.2
    • use tab as separator for output files
  • v0.3.1
    • fix makefile for bioconda can’t find zlib
  • v0.3.0
    • version after first manuscript revision corresponding to the third version manuscript on biorxiv
    • fix algorithm2 to use one more omega updates between epoch.
    • parameter --band changed to --windows
    • support CXXSTD compiling option to use c++17 standard. default CXXSTD=c++11
    • add PCAoneR, which implement the idea of PCAone in R but without out-of-core
    • colorful warning and error message. more checks and warnings.
  • v0.2.1
    • bug fix for PCAngsd
  • v0.2.0
    • add --maf option for SNPs filtering
    • faster parser for beagle file
    • optimization for PCAngsd
    • fix printing loadings for PCAoneA blockwise mode
  • v0.1.9
    • command line options are re-designed
    • upgrade to zstd v1.5.2
    • recover the original order of SNPs loading for fancy batch mode
    • add --no-shuffle option, remove --shuffle option
    • default PCAoneF (fancy RSVD) algorithm is chose.
  • v0.1.8
    • add -a, –tmp options
    • optimize makefile for conda
  • v0.1.7
    • two releases x64 and avx2 for linux
    • add conda install -c bioconda pcaone
    • upgrade Spectra to v1.0.1
    • change oversamples default as max(10, k)
    • add logger
  • v0.1.6
    • add –cpmed option to support raw counts for scRNAs
    • output elapsed I/O time and total wall time
    • remove pgenlib
  • v0.1.5
    • publish PCAone on mac with libiomp5 support
    • update documentation
    • add –shuffle option
    • optimization of FileCsv
  • v0.1.4
    • add CSV format support for scRNAs data
    • add –printv to print out another eigen vectors
    • add -N and -M options
  • v0.1.3
    • refactor Halko implementation
    • use Arnoldi as default method
    • add external/zlib
    • upgrade to Eigen 3.4.0
    • add structured permutation
  • v0.1.2
    • use double instead of float to improve numerical accuracy
    • add –no-shuffle, –oversamples, –buffer options
    • disable checking padding region of plink bed
    • fix denominator too small
    • set band range as 4,8,16,32,64
  • v0.1.1
    • supports bfile, bgen as input for both batch and blockwise mode
    • implement super power iteration for Halko
    • automatically permute plink bed file for fast halko
    • port jeremymcrae/bgen as bgen parser.
    • remove external BGEN v1.3 lib dependence
    • use libiomp5 instead of libgomp as multithreading RTL
    • fix bugs for PCAngsd algorithm
  • v0.1.0
    • supports bfile as input for both batch and blockwise mode
    • supports bgen as input for batch mode
    • supports both EMU and PCAngsd algorithm
    • supports beagle as input for PCAngsd
    • external dependecy: Eigen v3.3.8, Spectra v0.9.0, BGEN v1.3
    • release two pre-compiled binary for Linux and Mac OSX (libiomp5 required).