This notebook creates a jackknifed PCoA plot based on multiple rarefactions.

In [ ]:
%matplotlib inline
from emperor import Emperor, nbinstall


from skbio.stats.ordination import pcoa
from skbio.diversity import beta_diversity

from biom import load_table

# pydata/scipy
import pandas as pd, numpy as np
from scipy.spatial import procrustes

def load_mf(fn, index='#SampleID'):
    _df = pd.read_csv(fn, sep='\t', dtype=str, keep_default_na=False, na_values=[])
    _df.set_index(index, inplace=True)
    return _df

We are going to load data from Fierer et al. 2010 (the data was retrieved from study 232 in Qiita, remember you need to be logged in to access the study).

In [ ]:
mf = load_mf('keyboard/mapping-file.txt')
bt = load_table('keyboard/otu-table.biom')

We'll create five different distance matrices and compare them.

In [ ]:
ordinations = []
distances = ['jaccard', 'dice', 'russellrao']
rarefied = bt.subsample(1000)

for r in range(len(distances)):
    data = np.array([ for i in rarefied.ids()], dtype='int64')
    res = pcoa(beta_diversity(distances[r], data, rarefied.ids()))

Procrustes plots need a master set of coordinates where there rest of the matrices will be fitted around.

In [ ]:
master = ordinations[0]
fitted_ordinations = []

for i in range(1, len(ordinations)):
    _, matrix, _ = procrustes(master.samples.values, ordinations[i].samples.values)
    samples = pd.DataFrame(index=ordinations[i].samples.index,
    ordinations[i].samples = samples

If you want to share your notebook via GitHub use remote=True and make sure you share your notebook using nbviewer.

In [ ]:
viz = Emperor(master, mf, procrustes=fitted_ordinations, remote=False)

Lastly, we set the name of the distances as the procrustes_names attribute so we can differentiate them in the plot.

In [ ]:
viz.procrustes_names = distances
In [ ]: