README.md

In eriqande/pedvis: Simple drawing of pedigrees with dot

pedvis: an R package for drawing pedigrees

Eric C. Anderson

pedvis is a tiny package to take a data frame that lists a pedigree in three columns (kid, pa, ma) and make a .dot file for a marriage node diagram of the thing. It also will write it out as a pedigree factor graph. And it will run dot on it if you have dot in your system path.

This is sort of a hairball at the moment, and I would do it all differently if I had the time to do it over, but oh well. it works for what I want to do at the moment.

The documentation isn’t complete, but I mostly want to get this thing up on GitHub so Clemento, my homie at the lab, can draw some pedigrees.

Preliminaries

1. You should have dot from the Graphviz package installed. If not, get it from http://www.graphviz.org/ for your system and install it and make sure that dot is in your PATH.

2. You should have epstodpdf on your system. i.e. you ought to have ImageMagick or something of the sort. Thought this is probably not necessary if you are content to use the SVG output.

3. The program will try to open a PDF file with a system call to the “open” command, which might be a Mac thing, and not portable. But you still should get SVG output.

Getting the package and some examples.

The easiest way to get and install this package is with the devtools package from within R:

remotes::install_github("eriqande/pedvis")

Then, if that was successful, here you can see what the input is all about and do some examples

library(pedvis)  # load the package

# here is a simple input pedigree
simple_test_ped()

##    Kid Pa Ma
## 1    a  A  Z
## 2    b  A  Z
## 3    c  A  Y
## 4    d  B  Z
## 5    e  B  X
## 6    f  C  U
## 7    g  D  U
## 8    h  E  V
## 9    A  1 10
## 10   B  2 10
## 11   C  3 11
## 12   U  4 12
## 13   X  5 12

Note that the names of the individuals have to be strings NOT factors or integers. So, use stringsAsFactors = FALSE if reading in a data frame with read.table. (Or, better yet, use the readr package from the tidyverse).

Make a simple, faceless, nameless pedigree

Here we draw that simple pedigree as a marriage node diagram but we don’t label anyone and we don’t shade anyone in as observed:

out1 <- ped2dot(
  simple_test_ped(), 
  outf = "ped2dot_ex1", 
  pfactorNodeStyle = "invis", 
  pfactorEdgeStyle = "invis"
)

Since the outfile name was ped2dot_ex1, this has created files in the current working directory of:

ped2dot_ex1.dot  # the dot input file
ped2dot_ex1.ps   # the Postscript file that dot produced (if you have
           # properly installed dot and it is in your system path)
ped2dot_ex1.svg  # an SVG file of the drawn pedigree
ped2dot_ex1.pdf  # if you have epstopdf on your system

We can create a folder to put ped2dot_ex1.svg into, and then display it in the README:

dir.create("figures", showWarnings = FALSE)
dump <- file.copy("ped2dot_ex1.svg", "figures/", overwrite = TRUE)
knitr::include_graphics("figures/ped2dot_ex1.svg")

Shade some of the individuals as having observed data

out2 <- ped2dot(
  simple_test_ped(), 
  ObsNodes = c("a", "b", "f", "C", "D", "Z", "U", "10"), 
  pfactorNodeStyle = "invis", 
  pfactorEdgeStyle = "invis",
  outf = "ped2dot_ex2"
)

Again, we insert it into this RMarkdown README like this:

dump <- file.copy("ped2dot_ex2.svg", "figures/", overwrite = TRUE)
knitr::include_graphics("figures/ped2dot_ex2.svg")

Draw the same thing, but label the individuals

all_names <- unique(unlist(simple_test_ped()))
out3 <- ped2dot(
  simple_test_ped(), 
  ObsNodes = c("a", "b", "f", "C", "D", "Z", "U", "10"), 
  pfactorNodeStyle = "invis", 
  pfactorEdgeStyle = "invis",
  ShowLabelNodes = all_names,
  outf = "ped2dot_ex3"
)

Again, we insert it into this RMarkdown README like this:

dump <- file.copy("ped2dot_ex3.svg", "figures/", overwrite = TRUE)
knitr::include_graphics("figures/ped2dot_ex3.svg")

Make a full blown pedigree factor graph

This is a pedigree factor graph in the sense of Anderson and Ng (2016). In addition to the marriage nodes, it also explicitly shows (p)-nodes (representing factors having to do with allele frequencies) and (g)-nodes (representing factors having to do with genotypes of the observed individuls). It also shows that you can designate some individuals as “prongs,” which means they are hypothesized individuals. They are drawn in as dashed, by default.

out4 <- ped2dot(
  simple_test_ped(), 
  ObsNodes = c("a", "b", "f", "C", "D", "Z", "U", "10"),
  outf = "ped2dot_ex4",
  ShowLabelNodes = all_names,
  Draw_O_factors = TRUE,
  ProngNodes = c("3", "11")
)

Again, we insert it into this RMarkdown README like this:

dump <- file.copy("ped2dot_ex4.svg", "figures/", overwrite = TRUE)
knitr::include_graphics("figures/ped2dot_ex4.svg")

Note that to get the most out of this package, in terms of customizing the pedigrees, it is good to know the different possible edge and node attributes in dot, etc.

Literature Cited

Anderson, Eric C, and Thomas C Ng. 2016. “Bayesian Pedigree Inference with Small Numbers of Single Nucleotide Polymorphisms via a Factor-Graph Representation.” *Theoretical Population Biology* 107: 39–51.

eriqande/pedvis documentation built on Dec. 10, 2020, 1:15 p.m.