jgeofil/mycorrhiza-algorithm

Combining phylogenetic networks and Random Forests for prediction of ancestry from multilocus genotype data

Mycorrhiza algorithm

Published in -Mycorrhiza: Combining phylogenetic networks and Random Forests for prediction of ancestry from multilocus genotype data

Installing Mycorrhiza on Ubuntu 16.04

1. Make sure you have the latest version of Python 3.x

   python3 --version

2. Install pip3, Java and the tkinter library

   sudo apt-get install python3-pip python3-tk default-jre

3. Install Mycorrhiza

   pip3 install --upgrade mycorrhiza

4. Install SplitsTree

   Follow the instructions in the GUI installer, leaving all settings to default.

   wget http://ab.inf.uni-tuebingen.de/data/software/splitstree4/download/splitstree4_unix_4_14_6.sh
   chmod +x splitstree4_unix_4_14_6.sh
   ./splitstree4_unix_4_14_6.sh

   If the link above is not available - find the most recent version of the SplitsTree: http://ab.inf.uni-tuebingen.de/data/software/splitstree4/download

Installing Mycorrhiza on Mac OS X Sierra 10.12

1. If you don't already have the package manager HomeBrew, install it before proceeding.

   ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
   

2. Install Python 3.x

   brew install python

3. Install Mycorrhiza

   sudo -H pip3 install --upgrade mycorrhiza

4. Install SplitsTree

   The package can be found here. Follow the installer instructions, leaving all settings to default.

   If the link above is not available - find the most recent version of the SplitsTree: http://ab.inf.uni-tuebingen.de/data/software/splitstree4/download

Running an analysis from command line

1. Run an analysis.

   Run a 5-fold crossvalidated analysis.

   crossvalidate -i gipsy.myc -o out/ -s 5

   Run a analysis with a training set and a prediction set. Samples with a learing flag = 1 will be used for training and predictions will be made on samples with a learning flag = 0.

   supervised -i gipsy.myc -o out/

   To see all available parameters:

   crossvalidate -h

Running an analysis in a script

1. Import the necessary modules.

   from mycorrhiza.dataset import Myco
   from mycorrhiza.analysis import CrossValidate
   from mycorrhiza.plotting.plotting import mixture_plot

2. (Optional) By default Mycorrhiza will look for SplitStree in your home folder. I you wish to specify a different path for the SplitsTree executable you can do so in the settings module.

   from mycorrhiza.settings import const
   const['__SPLITSTREE_PATH__'] = '~/splitstree4/SplitsTree'

3. Load some data. Here data is loaded in the Mycorrhiza format from the Gipsy moth sample data file. Example data can be found here.

   myco = Myco(file_path='data/gipsy.myc')
   myco.load()

4. Run an analysis. Here a simple 5-fold cross-validation analysis is executed on all available loci, without partitioning.

   cv = CrossValidate(dataset=myco, out_path='data/')
   cv.run(n_partitions=1, n_loci=0, n_splits=5, n_estimators=60, n_cores=1)

5. Plot the results.

   mixture_plot(cv)

Documentation

https://jgeofil.github.io/mycorrhiza/

File formats

For microsatellite loci set the is_str flag to True.

```python
data = Myco(file_path='data/myco.myc', is_str=True)
data = Structure(file_path='data/myco.str', is_str=True)
```

Myco

Diploid genotypes occupy 2 rows (the sample identifier must be identical).

Column(s)	Content	Type
1	Sample identifier	string
2	Population	string or integer
3	Learning flag	{0,1}
4 to M+3	SNP Loci	{A, T, G, C, N}
4 to M+3	STR Loci	any or 000

STRUCTURE

Diploid genotypes occupy 2 rows (the sample identifier must be identical).

Column(s)	Content	Type
1	Sample identifier	string
2	Population	integer
3	Learning flag	{0,1}
4 to O+3	Optional (Ignored)
O+3 to M+O+3	SNP Loci	integer or -9
O+3 to M+O+3	STR Loci	any or -9