README.md

In eriqande/SNPcontam: Detecting contaminated samples from SNP genotypes

SNPcontam

This is an R package with some code implementing methods under development to detect contaminated samples from SNP genotyes sampled in molecular ecology contexts.

Currently, the directory not-package contains things that will not go into the final R package. This includes the manuscript that we will be preparing and several readings and resources.

To typeset the manuscript (which is really just a stub at this point) cd into the not-package/manuscript directory and issue this command (note that you have to have a TeX system installed on your system, as well as R, with the knitr package installed):

Rscript -e "knitr::knit2pdf('snp-contam-main.Rnw')"

The resulting PDF file, snp-contam-main.pdf should then be complete.

Compiling manuscript with TeXShop

To compile the manuscript in TeXShop you have to add a knitr engine and do a few things to try to make syncing work better:

1. First, copy the file in the repo ./not-package/extras/Knitr.engine to ~/Library/TeXShop/Engines/.
2. Then make the directory ~/Library/TeXShop/Rscripts and copy the file ./not-package/extras/patchKnitrSynctex.R from the repo to ~/Library/TeXShop/Rscripts. Note that this file is from https://github.com/jan-glx/patchKnitrSynctex.
3. Once you have done that you will have to restart TeXShop. Then you ought to choose TeXShop Preferences -> Typesetting -> Sync Method -> Both PDF sync and Search.

Note that synctex and knitr and TeXShop don't play very well with multiple Rnw files in a project. At some point that might get figured out. Until then we just use \input{} to input the child .Rnw files and don't have any R code chunks in the child files. Instead, we do all the R analysis in separate scripts and will save the output in an .rda file and just slurp the results out via, perhaps, a single .Rnw file that we input via knitr's child functions, but which does not create any text output.

Preparing Data for the Markov Chain Monte Carlo (MCMC) Method

The MCMC function requires that the input data be a matrix with columns equal to number individuals and rows equal to number of loci where each element is the genotype of an individual at a particular locus. Genotypes are given by the number of "1" alleles and should be 0s, 1s, 2s, or NA. Functions from the package fullsniplings can be used to turn a data frame of SNP alleles into values of 0, 1, 2, or NA using the code below:

library(fullsniplings)
snp_genos <- get_snp_genos(tmp_data)
data <- snp_genos$mat

Any columns or rows contain non-SNP information (i.e. information regarding the origin of the individual tmp_data should only be removed before using the function get_snp_genos.

Running the Code for the Markov Chain Monte Carlo (MCMC) Method

The functions below are used to run the MCMC for a set of data called data.

MCMC <- contam_MCMC(data)
means <- analyze_MCMC(MCMC)

The functioncontam_MCMC performs the MCMC method on the data, returning the all the values of the allele frequencies, proportion of contaminated samples, and z values for each sweep. If the input into contam_MCMC is only data, the function will use default values of inters = 1000, alpha = 0.5, beta=0.5, and lambda=0.5.

The function analyze_MCMC computes the posterior means of the allele frequencies, contamination proportion, and z values. If the MCMC results are the only input data, analyze_MCMC will use a default value of burnin=100.

Running the simulations

The first set of simulations can be run like so:

nohup Rscript --vanilla ./not-package/simulations/01_simulation_1.R 

This is changing. Now, you can just source the file: compile-analyses-and-documents.R from the directory that it is in (not-package)..

Terms

As a work partially of the United States Government, this package is in the public domain within the United States.

Of course, it would be awfully lame of anyone to take anything found within here and use it as their own before we tried publishing any of this, should we choose to do that.

See TERMS.md for more information.

eriqande/SNPcontam documentation built on May 16, 2019, 8:44 a.m.