NEWS.md
tidyr, readr, using future, carrier.assignment_ngs
pkgdown
assignment_ngs
fst_WC84: work fasterassigner with SeqArray and GDS object/filefst_WC84: work with radiator v.1.0assigner and now lives exclusively in package grur
assignment_mixture generated by purrr::df replaced recently by purrr:dfr. Changed DESCRIPTION field accordingly.subsample argument in assignment_ngs and assignment_mixture can now automatically detect the smallest sample size in the data’s grouping. So you can use subsample = "min" to let the function decide (if your not sure).pbmcapply package.pbmcapply package.dlr: simplified arguments, faster function and now creates the Dlr plotsSNPRelate are removed until the bugs with Fst calculation are resolved.assignment_ngs introduced in last commit that was suppose to be fix. Problem introduced by stackr::change_pop_names.assigner as a logofst_NEI87
subsample and iteration.subsample in fst_NEI87 and fst_WC84
SNPRelate bias issue is resolved the option is unavailablepbmcapply for Windowsassignment_ngs and assignment_mixture code cleaning to prep for CRAN and make them easier to debug.assigner now works in parallel with Windows
write_gsi_sim where the file was not created properly from an internal module.assigner::fst_WC84 can now use SNPRelate to compute Fst. The confidence intervals are not implemented, yet. The speed increase left me speechless, dataset with 30K snp are computed in less than 15 sec!assigner::fst_WC84 is 40% faster!assignment_ngs during imputations, the imputation module could not recognise that REF/ALT alleles are not necessary or usefull for assignment analysis. *enhancement to assignment_ngs and assignment_mixture so that when marker.number include "all" the iteration.method is set automatically to 1 when conducting the assignment with all the markers. Iterations at this point is useless and a waist of time.assignment_mixture: with assignment.analysis = "gsi_sim the unknown/mixture samples are compared with baseline populations using common markers between the pair. Now, the tables include the number of markers used. The summary provides the mean number of markers. This number will change each time randomness is used.fst_NEI87: very fast function that can compute: the overall and pairwise Nei’s (1987) fst and f’st (prime). Bootstrap resampling of markers is avalaible to build Confidence Intervals. The estimates are available as a data frame and a matrix with upper diagonal filled with Fst values and lower diagonal filled with the confidence intervals. Jost’s D is also given ;)fst_WC84: bug fix, the function was not properly configured for multi-allelic markers (e.g. microsatellite, and haplotype format from STACKS). Thanks to Craig McDougall for catching this.assignment_mixture: added a check to throw an error when pop.levels != the pop.id in strataassignment_mixture:
stackr.fst_WC84
tidyr::spread and tidyr::gather for data.table::dcast.data.table and data.table::melt.data.table to make the code faster, I forgot to split genotype into alleles for gsi_sim.you need to update [stackr] (https://github.com/thierrygosselin/stackr) to v.0.2.7 to appreciate this new version of assigner.
updated assignment_ngs with the separate stackr modules to simplify the function.
new data file available for assignment_ngs: genepop and genind object.
assignment_ngs now accept any vcf input file! i.e. it’s no longer limited to stacks vcf.
new arguments in assignment_ngs. The assignment using dapc can now use the optimized alpha score adegenet.dapc.opt == "optim.a.score" or the cross-validation adegenet.dapc.opt == "xval". This is useful for fine tuning the trade-off between power of discrimination and over-fitting (for stability of group membership probabilities). Cross validation with adegenet.dapc.opt == "xval" doesn’t work with missing data, so it’s only available with imputed data (i.e. imputation.method == "rf" or "max"). With non imputed data or the default: the optimized alpha-score is used (adegenet.dapc.opt == "optim.a.score"). When using adegenet.dapc.opt == "xval", 2 new arguments are available:
(1) adegenet.n.rep and (2) adegenet.training. See documentation for details.
removed arguments in assignment_ngs. Removed the pop.id.start and pop.id.end arguments that were confusing people. For those used to these arguments, they are now recycled in the new function individuals2strata in [stackr] (https://github.com/thierrygosselin/stackr). The strata file created by this function can be used with the strata argument in assignment_ngs.
2 modified arguments in assignment_ngs: (1) gsi_sim.filename is now filename; and
(2) if you didn’t use the imputation argument, replace imputation.method = FALSE to imputation.method = NULL or leave the argument missing.
simplified sections of codes in assignment_ngs that dealt with strata, pop.levels and pop.labels.
new function: write_gsi_sim. Write a gsi_sim file from a data frame (wide or long/tidy). Used internally in [assigner] (https://github.com/thierrygosselin/assigner) and might be of interest for users.
NEWS.md file to track changes to the package.fst_WC84 is now a separate and very fast function that can compute: the overall and pairwise Weir and Cockerham 1984 Theta/Fst. Bootstrap resampling of markers is avalaible to build Confidence Intervals (For Louis Bernatchez and his students;). The estimates are available as a data frame and a matrix with upper diagonal filled with Fst values and lower diagonal filled with the confidence intervals.assignment_ngs + assignment.analysis = "adegenet" + sampling.method = "ranked". A line at the beginning of a gsi_sim code section was deleted makig the assignment with adegenet go through that chunk of code and causing 100% assignment! if (assignment.analysis = “gsi_sim”) {code} prevent this problem…import_subsamples_fst to import the fst ranking results from all the subsample runs inside an assignment folder.assignment_mixture with sampling.method = "ranked" and assignment.analysis = "adegenet".imputations is now impute.method.impute with 2 options: impute = "genotype" or impute = "allele".data and covers the three types of files the function can use: VCF file, PLINK tped/tfam or data frame of genotypes file.tfam file will be used for the strata argument, unless a new one is provided. Columns 1, 3 and 4 of the tped are discarded. The remaining columns correspond to the genotype in the format 01/04 where A = 01, C = 02, G = 03 and T = 04. For A/T format, use PLINK or bash to convert. Use [VCFTOOLS] (http://vcftools.sourceforge.net/) with --plink-tped to convert very large VCF file. For .ped file conversion to .tped use [PLINK] (http://pngu.mgh.harvard.edu/~purcell/plink/) with --recode transpose.GBS_assignment to assignment_ngs. Stands for assignment with next-generation sequencing data.df.file if you don’t have a VCF file. See documentation.strata if you don’t have population id or other metadata info in the individual name. See documentation.THL to thl and snp.LD to snp.ld to follow convention.iterations.subsample changed to iteration.subsample.iterations changed to iteration.method to avoid confusion with other iteration arguments.baseline and mixture arguments from the function GBS_assignment. These options will be re-introduce later in a separate function.marker.number higher than the number of markers in the data set was causing problems. This could arise when using arguments that removed markers from the dataset (e.g. snp.ld, common.markers, and maf filters).sudo rm /usr/local/bin/gsisim