Polysat

polysat is an R package for polyploid microsatellite analysis in ecological genetics. Version 1.3-2 is on CRAN as of September 2013.

Screenshot of polysat from RGui

Version 1.3 handles information about ploidy differently than earlier versions. Due to this, there are minor changes to the source code of most functions with respect to earlier versions. If you think you have found a bug, please feel free to let me know about it! If I fix the bug but am not immediately ready to release a new version of polysat, I will send you the corrected source code so that you can proceed with your analysis in the meantime.

What polysat does

• Assumes allele copy number ambiguity in partial heterozygotes.
• Handles data of any ploidy, including mixed ploidy samples.
• Stores genotype data in a simple format that can be easily manipulated to exclude or add samples and loci.
• Imports and exports data in ABI GeneMapper Genotypes Table, GenoDive, Structure, SPAGeDi, ATetra, Tetrasat/Tetra, POPDIST, and binary presence/absence formats. Import is also available for STRand format, and export is available for adegenet presence/absence format.
• Calculates pairwise distances between individuals using a stepwise mutation model or infinite alleles model.
• Calculates Shannon and Simpson indexes of genotype diversity, and can calculate a confidence interval for the Simpson index. Counts alleles for measures of allelic diversity.
• Counts alleles to assist user in estimating ploidy.
• Estimates allele frequencies in autopolyploids using either an iterative or non-iterative algorithm. Calculates pairwise F_ST based on these estimates. Mixed ploidy population size is measured in genomes rather than individuals.
• Exports allele frequencies in SPAGeDi and adegenet formats.
• Easily extensible; ordinary users can write new functions to interface with the package.

Author and Maintainer

User:Lindsay V. Clark

Note that I have changed institutions since the paper was published! I no longer check email at my UC Davis address. If you replace "ucdavis.edu" with "illinois.edu", that is my new email address.

Obtaining polysat

If you don't already have R, download it from CRAN and install it.

At the prompt in the R console, type:

install.packages("combinat")

install.packages("polysat")

library(polysat)

Documentation

Tutorial manual: Most users will want to read this first to get a general idea of how to use the package. It starts with a broad tutorial to familiarize users with the package, then goes into more detail about how data are stored in polysat and which analyses are appropriate for autopolyploid and allopolyploid data.

Reference manual: This is an alphabetized collection of all of the help files provided with the package. It contains more details about each function, as well as additional examples.

Example files for data import

Example files referred to in the tutorial manual are linked below:

• Media: GeneMapperCBA15.txt
• Media: GeneMapperCBA23.txt
• Media: GeneMapperCBA28.txt
• Media: structureExample.txt
• Media: ATetraExample.txt
• Media: genodiveExample.txt
• Media: spagediExample.txt
• Media: GeneMapperExample.txt
• Media: dominantExample.txt
• Media: POPDISTexample1.txt
• Media: POPDISTexample2.txt
• Media: tetrasatExample.txt
• Media: STRandExample.txt

How to cite polysat

Clark, LV and Jasieniuk, M, 2011. POLYSAT: an R package for polyploid microsatellite analysis. Molecular Ecology Resources 11(3): 562-566. DOI: 10.1111/j.1755-0998.2011.02985.x

Wish List

This section lists additional functionality that I'm thinking of adding to polysat. If you have any additional requests (please be specific), or would like to "vote" for one of the items below to be a top priority, just send me an email! If you have created your own functions to interface with the package and would like to be added as a contributor, I am open to that as well.

• For allopolyploids, assign alleles to one genome or the other based on what genotypes are found in the population. Use these allele assignments to re-code allopolyploid data into autopolyploid data by splitting each locus into two or more loci. (I've got some ideas on how to do this, but for now I can only work on it in my free time.)
• Related to the above functionality, use the distribution of genotypes to give a probabalistic estimate of whether a locus is polysomic or disomic.
• Improve computational efficiency on inter-individual distance measures. (I tried vectorizing some of the code for Bruvo.distance and actually made it slower, but if I ever learn C maybe a compiled version would be faster.)
• Given probabilities of unambiguous genotypes (genotypeProbs function), randomly generate an unambiguous dataset. This could then be passed to software such as adegenet that allows for polyploidy but not allele copy number ambiguity.
• More population statistics (Weir and Cockerham 1984, etc.).
• Parentage analysis
• Options for handling data where allele copy number is known.
• Estimate selfing rate under polysomic inheritance, based on observed and expected frequencies of fully heterozygous genotypes. I wrote a function to do this, but the results were imprecise due to stochastic effects in simulated datasets. I can email you the source code and documentation if you would like to tinker with it.
• Some relatedness coefficients for unambiguous genotypes, to be used with meandistance.matrix2.

Frequently asked questions

If you have never used R before, particularly if you find command-line software to be intimidating, you may need to spend a day or two just learning R before you even touch polysat. (Look for the An Introduction to R manual on the CRAN website.) I have tried to make polysat as user-friendly as possible, but that cannot substitute for a basic understanding of how R works. Trust me, learning R is worth it! R is very powerful and efficient software for data analysis, and if you take the time to learn it for the sake of using polysat, you may find yourself using R in other areas of your research. If you are not sure how something works, try experimenting to see if it does what you think it does.

If you get an error message or if the software behaves in an unexpected way, I'm happy to help you figure out what went wrong. (Particularly since several users have helped me to find bugs in this way.) In this case, please send your dataset (or a subset of your data that you don't mind sharing) and code that can be used to duplicate the error.

polysat

• Is missing data allowed in polysat? Yes it is! For the Structure, GenoDive, SPAGeDi, and Tetrasat/Tetra formats, you can code the missing data as you normally would for that format. For the GeneMapper format, you can either delete rows with missing data, or fill in a -9 in the first allele column for that row.
• In read.GeneMapper I got the error "line 2 did not have X elements". Each line of the file needs to have the same number of tab stops. You can add these manually in a text editor, or if you open and save the file in a spreadsheet program it should automatically insert the right number of tab stops.
• Do all populations need to have the same number of individuals? Do the samples have to be ordered by population? No and no. Example code

R base

• I have made my PCoA plot. Can I add a label for each sample? Yes. See ?text.
• How can I assign each population its own color or symbol for plotting? See additional Example code.
• How can I find the percentage of variation represented by the axes in Principal Coordinate Analysis? When you run cmdscale, use eig=TRUE. Then (assuming here you name the output pca) you can do pca$eig/sum(pca$eig). When you plot the points, you will now have to use pca$points[,1] and pca$points[,2] instead of pca[,1] and pca[,2]. See ?cmdscale.
• I tried to do PCoA (cmdscale) but got the error "NA values not allowed in d." If you only have one or two loci, you will need to exclude all individuals with missing data from your analysis. If you have three or more loci and still see this error, you may need to exclude individuals that are missing genotypes at multiple loci. This will prevent missing data (NA) in the distance matrix (d).
• Can I use my distance matrix to make a neighbor-joining tree instead of a PCoA plot? Yes. The R package ape works well for making neighbor-joining trees, and I have gotten it to work on distance matrices produced by polysat.

Known issues

Current version (1.3-2)

• Example files for data import do not install with the package as intended. I have made them available on this web page.
• Locus names should not contain a period. For now this is going to remain a "feature" of polysat.

Older versions

• read.STRand has problems if any one locus never has more than one allele per individual (1.3-1 and earlier).
• Lynch.distance does not give the right answer if a genotype contains duplicated alleles (e.g. AAAB instead of AB; 1.3-1 and earlier).
• There is a problem with "genbinary" objects if a particular locus has only one allele for the whole dataset. (Versions 1.2 and earlier.)
• Rounding errors in R can cause errors when the Simpson or Shannon indices are used. ("p should sum to 1.") (Version 1.2)
• Bruvo.distance: In version 1.2-0 and earlier, if both genotypes are missing, returns 0 rather than NA.
• read.GenoDive: The function does not expect a "Clones" column, and will simply take sample names from whichever column is second. (Version 1.2-0 and earlier)
• If one locus name is a shorter version of another locus name, e.g. "ABC1" and "ABC12", there will be some issues with the "genbinary" class and with the allele frequency functions. (Version 1.2-0 and earlier)
• In version 1.1, write.GeneMapper has problems when genotypes have more alleles than the maximum ploidy in the dataset.
• editGenotypes in version 1.0-0 rearranges the genotypes if the samples and loci are not in alphabetical order.
• In version 0.1, read.SPAGeDi will not work with missing=0, missing=00, etc. This should not be an issue in version 1.0 because of the change in data structure. (In either version, even if the missing data symbol is at the default, -9, the software still knows that zero indicates missing data in a SPAGeDi file.)

Source code

For advanced R users, here is the source code for the functions in the package, so that you may tweak them or create new functions for your own use:

Current version (1.3-2)

• Media: classes_generics_methods_polysat_1-3.R.txt
• Media: class_conversion_polysat_1-3-1.R.txt
• Media: dataimport_polysat_1-3-2.R.txt
• Media: dataexport_polysat_1-3-1.R.txt
• Media: individual_distance_polysat_1-3-2.R.txt
• Media: population_stats_polysat_1-3-2.R.txt

Older versions

• Media: individual_distance_polysat_1-3.R.txt
• Media: population_stats_polysat_1-3-1.R.txt
• Media: dataimport_polysat_1-3-1.R.txt
• Media: classes_generics_methods_polysat_1-2-1.R.txt
• Media: class_conversion_polysat_1-0.R.txt
• Media: dataimport_polysat_1-2-1.R.txt
• Media: dataexport_polysat_1-2.R.txt
• Media: individual_distance_polysat_1-2-1.R.txt
• Media: population_stats_polysat_1-2-1.R.txt
• Media: individual_distance_polysat_1-2.R.txt
• Media: population_stats_polysat_1-2.R.txt
• Media: classes_generics_methods_polysat_1-0-1.R.txt
• Media: dataimport_polysat_1-1.R.txt
• Media: dataexport_polysat_1-1.R.txt
• Media: individual_distance_polysat_1-0.R.txt
• Media: population_stats_polysat_1-1.R.txt
• Media: polysat_0.1_functions.R.txt
• Media: dataimport_polysat_1-0.R.txt
• Media: dataexport_polysat_1-0.R.txt
• Media: population_stats_polysat_1-0a.R.txt

External links

• You can rate and review polysat on Crantastic. (I am of course also open to questions and comments via email.)
• CRAN page with source and binary downloads.