v1.1.6

24AUG2017 - v1.1.6 Modified behavior of --cm flag based on user feedback. When this flag is set ROH lengths and length classifications will be reported in cM but ROH boundaries will be reported in bps.

v1.1.5

v1.1.5 Introduced --ld-subsample argument. If set > 0, will use only a subset of individuals for computing LD statistics when calculating wLOD scores.

GARLIC v1.1.4

18MAY2017 - v1.1.4 If a freq file is provided, GARLIC now ensures that internal coding of alleles is consistent with the allele specified in the file. Improved handling of loci with 100% missing data. Improved reading of gzipped freq files. Improved lod score cutoff discovery. Removed default setting for --winsize, which must now be specified. Now allows --auto-winsize to be used with --winsize-multi, will choose best from the list.

GARLIC v1.1.3

12MAY2017 - v1.1.3 adds a new option for --gl-type called GQ, which is a phred-scaled probability that the genotype is wrong. If p = 0.999 is the probability that the genotype is correct, then GQ = -10*log10(1-p) = 30. The default for --gl-type is now set to none, so if you provide a TGLS file you will be required to choose between GQ, PL, and GL.

GARLIC v1.1.2b

08MAY2017 - v1.1.2b fixes a bug that causes a crash when --overlap-frac is set to 0. Setting to zero now chooses the smallest sensible overlap fraction (1/winsize). If you wish to let garlic automatically guess overlap fraction, use the new command line option: --auto-overlap-frac. Also includes an interation counter to monitor progress of GMM portion of the code.

GARLIC v1.1.2

07MAY2017 - v1.1.2 fixes a bug that causes a crash when a tgls file indicates a genotype has 0 likelihood. Add progress bars for LD calculations and LOD score calculations. Improve efficiency of tgls file loading.

GARLIC v 1.1.1a

06MAY2017 - v1.1.1a fixes a bug introduced with 1.1.1 that caused crashes when --resample is used.

GARLIC v1.1.1

05MAY2017 - Update to version 1.1.1. For some large datasets (e.g. WGS), KDE can take extremely long to complete, even when only a subset of individuals are considered. Here we introduce some changes to speed this calculation up. Instead of passing all LOD score windows to the KDE function, we now pass only non-overlapping windows. This reduces the number of points sent to the function by a factor of WINSIZE. If you wish to retain all LOD scores, you may set the --no-kde-thinning flag. We also change the default number of random individuals for KDE (--kde-subsample) to 20, which still may be set to 0 to use all individuals.
We also introduce a lower bound on the size of ROH reported. Now, an ROH will not be called unless it contains at least WINSIZE*OVERLAP_FRAC number of SNPs. For example, for a WINSIZE (--winsize) of 100 and an OVERLAP_FRAC (--overlap-frac) of 0.25 the smallest possible ROH that will be reported will be 25 SNPs long.
Finally, if --overlap-frac is set to 0, GARLIC will attempt to guess a good choice.

GARLIC v1.1.0a

02MAY2017 - Update to version 1.1.0a. Bug fixed that caused crashes when using TGLS file.

GARLIC v1.1.0 - weighted LOD score calculation

Update to version 1.1.0. Includes several updates, including a method for weighting LOD scores by gaps between SNPs, the option to report ROH lengths in cM, and the ability to provide per-genotype error rates.