Hickory

Bayesian analysis of inbreeding and F-statistics

View the Project on GitHub kholsinger/Hickory

Hickory

Bayesian estimates of Wright’s F statistics

This package estimates the within population inbreeding coefficient, , and Wright’s , , using a Bayesian approach. Right now, it is limited to two alleles per locus for co-dominant markers. It’s easiest to use if your data are in a CSV file, but it wouldn’t be too difficult to convert data from other formats to the one that required by analyze_codominant() and analyze_dominant(). If you look at the Roadmap below, you’ll see that I plan to add an interface to adegenet, which should allow you to work with data in most of the widely used formats.

A note on requirements

You will need to have a C/C++ compiler installed in order to use this package (for now).1 The RStan Getting Started page has helpful information on getting your system configured. I’ll help if I can, but I’ve been using Macs for nearly 10 years. I can probably help with Mac or Linux problems, but I may not be much help with Windows. If you’ve managed to install other R packages from source, you’ll probably be just fine.

I also specified minimum versions for all of the libraries that Hickory depends on. It’s possible I could get away with earlier versions, but I don’t have a way to test that. That means you might need to upgrade some of your libraries before you can install this one.

Installation

If you don’t already have the devtools package installed, first you’ll need to install it.

install.packages("devtools")

Once that’s installed, then you can install Hickory like this:

install.packages(c("bayesplot", "rstan", "tidyverse"))
devtools::install_github("kholsinger/Hickory")

If you also want to install vignettes illustrating how to use Hickory in more detail, you’ll want to change that second line to

devtools::install_github("kholsinger/Hickory", build_vignettes = TRUE)

The installation will take longer, but you’ll be able to run vignette("Hickory") to get an overview, and browseVignettes("Hickory") to see all of the vignettes that are available. If you’d prefer, you can simply look at the vignettes in the doc directory here:

https://github.com/kholsinger/Hickory/tree/master/doc

You’ll probably be interested only in the HTML files in this directory. To view them, you’ll need to hit the “Raw” button on the upper right of where the HTML code is displayed, save the file on you computer, and open it from your computer.

The Rmd file is the rmarkdown file that produced the HTML. The R file is R code that’s extracted from the Rmd file.

A note on dominant markers

Foll et al.2 identified biases associated with the method originally implemented in Hickory for dominant markers. Several users also reported that estimates of f seemed unreasonably high when they used a large number of markers.

The C++ version of Hickory used Metropolis-Hastings sampling to approximate the posterior. Sampling was slow and estimates of showed high autocorrelation. The Hamiltonian Monte Carlo algorithm used in Stan doesn’t suffer from either of those limitations, but I haven’t checked the results from this sampler against the biases that Foll et al. reported. Use caution interpreting estimates of until I’ve checked that out. Fortunately, estimates of aren’t too sensitive to , so those estimates are likely to be reliable.

If you plan to use Hickory to analyze dominant markers, I encourage you to read the vignette, vignette("dominant").

A note on the Github repository

If you visit the repository you’ll see that the “main” branch is still called “master”. I plan to rename it after Github releases tools making it easier. I am not good at git, and I need all of the help I can get.

Roadmap

There are a lot of improvements that still need to be made. In addition to the items listed below, I’ll be working on improved documentation. Please don’t hesitate to email me if you have a question, a comment, or suggestions for future improvements.

  1. Implement and models with model comparison using loo to provide ways to evaluate whether there is evidence for inbreeding and whether there is evidence for allele frequency differences among populations.

    DONE

  2. Implement population- and locus-specific effects on , including identification of potential outlier loci and populations.

    DONE

  3. Implement interface with adegenet.

  4. Build in some internal tests to ensure accuracy and consistency.

  5. Implement posterior predictive checks.

  6. Investigate possible biases in dominant marker estimates.

  7. Implement multiallele version of analyze_codominant().

  8. Implement locus and population selection

1Once I feel comfortable enough with this package to release it to CRAN, I believe that the build system there will produce the binaries for different platforms. I don’t have the ability to do that myself.

2Foll M, Beaumont MA, Gaggiotti O. 2008. An Approximate Bayesian Computation Approach to Overcome Biases That Arise When Using Amplified Fragment Length Polymorphism Markers to Study Population Structure. Genetics 179:927-939.