- Added executable examples to each function.
- First official CRAN release! 🎉
- New prototype implementation of exhaustive search of qpAdm models through rotation of sources and outgroups (
- New tutorial describing the
- Fixed broken regex parsing of qpWave log files in some cases.
- Added type checking to a couple of functions (not all wrappers type checked yet).
- qpAdm proportions table now shows p-values and SNP counts (both for overall analysis but also for the target sample).
- Satisfying further checks on various testing servers.
- Cleaned up build procedure.
- Expanded some tests.
- Prepared for the first CRAN submission.
- Fixed a minor bug in single-source qpAdm analyses. The package now detects this and correctly handles missing “subsets/patterns” information.
Finally resumed the development of admixr! Apologies to everyone for having to wait so long. Thank you for your patience and feedback since the last release. I hope that things will start moving a little bit faster and we will reach version 1.0 in the next couple of months.
New features and improvements:
- New function
loginfo() which operates on any output object from an admixr wrapper and shows the full log output (the “log file” in ADMIXTOOLS jargon) associated with the analysis. It also has options for saving the log file to a permanent location and to only show a log file for a target sample of interest (relevant for qpAdm analyses with multiple targets at once).
admixr can now (hopefully) detect broken/truncated output files generated by ADMIXTOOLS whenever there is some fatal issue with an analysis. The package should now detect this and inform the user.
- Removed qpAdm argument
details - the user will always want to see the full analysis summary so this option is redundant. It is still kept in qpWave - for now, until I figure out how useful the full output actually is. Given the implementation of
loginfo() above, we might remove the argument from qpWave too at some point soon.
- The function
download_data() now fetches data from a more stable location.
- Parse admixture proportions from the “best coefficients” line of the qpAdm output file.
- Fixed handling of warnings when dealing with duplicated populations.
- 9s are now replaced with NAs in
write_geno(), which makes it more convenient to write custom analytic code working on data.frames.
qpAdm() output elements and changed its function signature.
transversions_only(). The old function is now deprecated.
print.EIGENSTRAT() now uses a pre-calculated numbers of removed/remaining sites, instead of calculating them each and every time.
- Printing EIGENSTRAT objects now also shows “group” and “exclude” modifiers only if present.
read_output() made public.
- Fixed issues with parsing of mis-formatted ind files (tabs at the ends of lines etc.).
It turned out that dragging along Rcpp and Boost dependencies just for the VCF -> EIGENSTRAT conversion function causes unnecessary complications in the installation process. It’s not worth having it in the package if it would be used only by a small fraction of potential users.
This function has been removed and the
vcf2eigenstrat program is maintained in its own repository.
- Conversion of VCF to EIGENSTRAT format is now implemented in C++ and should be approximately infinitely faster than the old conversion function written in pure R.
- Conversion of EIGENSTRAT into VCF has been removed.
- Added full implementations of
filter_bed() now implemented simply by calling
bedtools in the background. This turned out to be way faster and memory efficient than the previous data.table-based solution.
- Fixed missing
- Removed the huge built-in data set. Implemented
download_data() function that fetches the example data set from the web.
- The package now has a tutorial vignette describing the main functionality.
- Simple SNP dataset is now included with the package.
- The API of many utility functions has been simplified and their internals re-written.
filter_sites is now implemented using
data.table and allows overlap with an arbitrary BED file.
- All wrappers have been given simpler names (
- F4 statistic can now be calculated using a separate
f4() function (
f4mode parameter remains in the
d() function though, as
- All tests are performed on Travis CI using installed and compiled ADMIXTOOLS software.
- D statistic as qpDstat,
- f4 statistic as qpDstat(f4mode = TRUE),
- f4-ratio statistic as qpF4Ratio,
- f3 statistic as qp3Pop.