In bodkan/spannr: A Simulation Framework for Spatiotemporal Population Genetics

Quick installation

slendr is available on the CRAN R package repository. As such, you can install it simply by executing install.packages("slendr") in your R console.

If you want (or need) to get its development version, you can install it directly from GitHub by executing devtools::install_github("bodkan/slendr") via the R package devtools (you can gen devtools by running install.packages("devtools")). In fact, if you decide to try slendr, please make sure to update it regularly and keep an eye on the changelog on a regular basis! This is where you can find information about latest bugfixes and potential breaking changes.

Once you install slendr, calling library(slendr) will check that all software dependencies are available. If they are not, the R package will provide a brief helpful guide on how to resolve potential issues. The rest of this vignette talks about the necessary software dependencies in a bit more detail.

Please note that slendr has been extensively tested only on macOS and Linux at the moment. However, there is an experimental support for runnig both SLiM and msprime simulations on Windows and for analyzing tree-sequence outputs using its tskit interface on this platform as well. Feedback on issues with using slendr on Windows is highly appreciated.

Software dependencies

slendr relies on three main software dependencies:

1. geospatial data analysis R package sf (for encoding spatial slendr models and analysing spatial tree-sequence data),

2. forward population genetic simulator SLiM (for forward simulations),

3. Python modules tskit, msprime, and pyslim (for coalescent simulations and tree-sequence analysis), and also pandas used internally by the simulation back ends.

All three are widely used in their respective fields and, as such, are easily obtainable on all major operating systems (see below for more information on how to troubleshoot potential problems).

Note that depending on your use case, not all three sets of dependencies will be necessarily needed. If you're not going to be running forward spatial simulations, you don't need SLiM or geospatial R packages sf, stars, and rnaturalearth. This applies also to the animation of spatial models using the gganimate R package. This is why slendr does not install those as dependencies. If you need this functionality, you will have to install the respective R packages manually.

In this vignette, I will briefly explain how to get all slendr's software dependencies installed. That said, note that under normal circumstances (with the exception of SLiM), no manual installation of individual dependencies is required.

sf, stars, rnaturalearth

The R package sf is at the heart of geospatial data analysis in R. It is available on CRAN and can be installed for all major platforms by executing install.packages("sf") in your R session. The same applies for stars and rnaturalearth.

When you first load slendr via library(slendr), if you're missing any of the three geospatial R packages, you will be notified and instructed how you can easily obtain them from CRAN using a single call to install.packages().

That said, sf itself depends on a number of geospatial libraries and depending on the exact setup of your Linux or macOS machine, some of those libraries could be missing. Luckily, all of them are very easy to install via Homebrew (on macOS) or via the appropriate package manager of your Linux distribution (Ubuntu, Fedora, etc.). Detailed instructions on how to do this for your operating system can be found here.

If you're having problems with the installation of any of these three packages, look for help here.

macOS

One user who recently installed slendr on a fresh macOS system reported that they needed to install libgit2 in order to be able to install the package devtools for the devtools::install_github("bodkan/slendr") step described on top of this page.

Additionally, they had to install a couple of C/C++ libraries as well (all dependencies of the sf package). In the end, they were able to successfully install slendr after running:

brew install libgit2 udunits gdal proj

Note that this assumes that you have the Homebrew package manager already setup on your Mac. If you're a beginning computational scientist using a Mac, I strongly encourage you to install Homebrew. Sooner or later you will need some specific Linux/unix program anyway, and Homebrew is the way to get it (Mac is a unix machine, but without Homebrew a very poor one by default).

Linux

Testing slendr installation on a fresh, pristine Debian installation with no dependencies previously installed, I had to run the following:

sudo apt-get install libudunits2-dev libssl-dev libgdal-dev libgsl-dev libgit2-dev libfontconfig1-dev libharfbuzz-dev libfribidi-dev

It's very unlikely you would need all of the above (and you might need other packages on non-Debian distributions), but this is what got slendr and all of its dependencies running on a completely clean system. Might be a good start in case you have trouble on your Linux machine.

Windows

No special treatment should be necessary to get slendr running. If you install slendr via install.packages("slendr"), you will get the binary version of the package and all its dependencies without any need for compiling sources.

SLiM

The forward population genetic software SLiM is available on all major software platforms. Its complete installation instructions can be found here. On a Mac, I recommend installing SLiM via the pkg installer available for direct download from its website. On Linux, you can either install SLiM via the appropriate package manager for your Linux distribution (see SLiM manual here for more information), or you can easily compile your own. To install SLiM on Windows, please follow the instructions in section 2.3.1 of the SLiM manual. Note that although SLiM is also available on conda, it comes without SLiMgui!

Note that slendr requires SLiM 4.0 and will not work with an earlier version. Again, running library(slendr) will inform you of any potential issues with your SLiM installation.

Adding SLiM to the $PATH

In order for slendr to be able to find your SLiM installation, R must be able to find the slim binary (or slim.exe in case of Windows) in the so called $PATH variable. The easiest way to verify that this is true is to call Sys.which("slim") (or Sys.which("slim.exe") on Windows) in your R session. For instance, on my own Mac, I get this:

> Sys.which("slim")
                 slim 
"/usr/local/bin/slim" 

On Windows, I get this (I followed the section 2.3.1 of the SLiM manual describing the most "official" way to install SLiM and I recommend you follow this as well):

> Sys.which("slim.exe")
                            slim.exe 
"C:\\msys64\\mingw64\\bin\\slim.exe" 

If, on the other hand, you might get something like this — an empty string:

> Sys.which("slim")
slim 
  "" 

This means that your $PATH R is not configured properly and R (and slendr) won't be able to find SLiM in the $PATH.

How to add SLiM to the $PATH then? Probably the most convenient way to do this is by editing the .Renviron file. The precise location of this file depends on your operation system but you can automatically get it open in a text editor by using this command (you might have to do install.packages("usethis") first):

usethis::edit_r_environ()

Then, you can either add the following (note that there's no $ in this line!):

PATH="<path to directory with SLiM binaries>"

Alternatively, should you already have some $PATH contents specified for your R session, you can get it by calling Sys.getenv("PATH") in your R console, grab the entire string you get in this way, and append the path where you installed SLiM using an appropriate delimiter (: on Linux/macOS, ; on Windows) to this string. In this case, the edit might look something like this:

PATH="<original Sys.getenv("PATH") content><delimiter><path to directory with SLiM binaries>"

For instance, on my Mac, the .Renviron file contains this line (note the last item /usr/local/bin matches the path to slim I showed above):

PATH="/opt/homebrew/bin:/opt/homebrew/sbin:/Users/mp/.my_local/bin:/Users/mp/.my_local/AdmixTools/bin:/Library/TeX/texbin:/opt/homebrew/opt/gnu-sed/libexec/gnubin:/opt/homebrew/opt/coreutils/libexec/gnubin:/usr/local/bin"

On my testing Windows machine, I have this in the .Renviron file (again, look for the path C:/msys64/mingw64/bin;$PATH" which matches the slim.exe binary shown above):

PATH="C:\\msys64\\usr\\bin;C:\\rtools43\\x86_64-w64-mingw32.static.posix\\bin;C:\\rtools43\\usr\\bin;C:\\Program Files (x86)\\R\\R-4.3.2\\bin\\x64;C:/msys64/mingw64/bin"

The most important check that your R can find SLiM is using the Sys.which() command described above. Before this gives you a positive result, you won't be able to use slendr's spatial SLiM simulations. If you're struggling with this, search for advice related to .Renviron and $PATH online.

Fallback options

If you don't want to deal with editing the .Renviron file, you should be able to set the path to SLiM using the command Sys.setenv(PATH = "<path to directory with SLiM binaries>"). You will have to do this at the beginning of each of your slendr R scripts though.

Alternatively, the slim() function has an argument slim_path= where you can specify the full path to the slim or slim.exe binaries directly (not the directory as it's the case with the $PATH! the full path to the slim/slim.exe binary itself).

Python

In order to be able to run coalescent simulations and process tree-sequence files, slendr needs Python modules tskit, msprime, and pyslim (it also needs the pandas library).

Setting up an isolated Python environment with specific version of Python packages (which is very important to avoid clashes among different Python programs needed by your system) can be a bit of a hassle for some users. This is especially true for R users who might not use Python in their daily work.

In order to make sure that the R package has the most appropriate version of Python available, with the correct versions of all of its Python module dependencies, slendr provides a dedicated function setup_env() which automatically downloads a completely separate Python distribution and installs the required versions of tskit, msprime, and pyslim modules into a dedicated virtual environment. Moreover, this Python installation and virtual environment are entirely isolated from other Python configurations that are already present on the user's system, avoiding potential conflicts with the versions of Python and Python modules required by slendr.

Next time you call library(slendr), you will need to activate this environment automatically by calling init_env(). If you're not comfortable with Python you don't need to worry beyond calling setup_env() and init_env(), no interaction with Python is necessary for working with slendr in R.

Failing conda?

In order to support Windows, slendr uses conda to download a Python interpreter as explained above. Given this fact, when you run setup_env(), slendr tries to leverage conda being present to install its Python dependencies (msprime, tskit, pyslim, pandas) via conda itself.

Unfortunately, conda can break for frustratingly many random reasons which completely trips up setup_env(). If you run into this issue, there is a fallback option to install Python dependencies msprime, tskit, pyslim, and pandas via pip which, unlike conda, works practically every time. If your setup_env() fails for some conda-related reason, first clear the broken environment by calling clear_env(), restart your R session, and then call setup_env(pip = TRUE) instead of the default setup_env(). Note that this might require you to install the GSL numerical library, but that's a trivial issue on macOS (brew install gsl) and Linux (on Ubuntu, for instance, sudo apt-get install libgsl-dev).

Information for Python experts

In case you are wondering how does slendr accomplish the above: slendr's Python interface is implemented using the R package reticulate. This embeds a Python interpreter inside an R session, enabling high-performance interoperability between both languages without any need for user intervention.

ModuleNotFoundError: No module named 'tskit' error

If you're running into this error, it means that slendr was prevented from activating its internal Python virtual environment. Most likely you either didn't run setup_env() as described above, or you forgot to run init_env() before attempting to simulate tree sequence data, or the Python environment got somehow corrupted. All of these cause slendr's internal machinery to fail to pick up it's Python dependencies for tree-sequence simulation and analysis, leading to the No module named 'tskit' error.

Before reporting this error on GitHub, please carefully read this writeup which I put together when a user reported the same error. It contains all the required information on how does slendr's Python environment activation works and why, and why you're running into the error. Because this issue is almost always caused by a problem outside of slendr's influence, it's important that you understand what's going on before reporting what is often assumed to be a slendr bug when in reality it isn't.

bodkan/spannr documentation built on Dec. 19, 2024, 11:43 p.m.