def plot_variance(main_dir='./', plot_dir='./plots/',
log=True, genus_only=True):
df = sort_by_variance(main_dir=main_dir, genus_only=genus_only)
fig, ax = plt.subplots()
df.reset_index().variance.plot(ax=ax, kind='hist', bins=100)
if log:
ax.set_yscale('log')
plt.title('distribution of variances (log scale)')
plt.xlabel("variance for genera's abundance across all samples")
plt.ylabel("frequency (log scale)")
filepath = \
concat_dir_and_filename(plot_dir,
'distribution_of_sample-wise_variances.pdf')
plt.savefig(filepath)
def plot_pca_results(top_percent=20, genus_only=False,
facet_row=True, uniform_axes=True,
main_dir='./', plot_dir='./plots/',
savefig=True):
# get data that's ready for PCA:
pca_input, data_transformed, variances = \
run_pca(main_dir=main_dir, top_percent=top_percent,
genus_only=genus_only)
# import the sample info needed to map features to sample IDs.
sample_info = colnames_to_sample_info_array(pca_input, main_dir=main_dir)
# prepare axis labels, which also serve as dataframe column names.
x_axis_label = 'principal component 1 ({0:.0%})'.format(variances[0])
y_axis_label = 'principal component 2 ({0:.0%})'.format(variances[1])
# put together the transformed data and sample descriptions
plot_data = pd.concat([pd.DataFrame(
{x_axis_label: data_transformed[:, 0],
y_axis_label: data_transformed[:, 1]}), sample_info], axis=1)
# define a custom color palette using:
# Conditions were seized at week ten, so seven early samples in
# the original condition and four latest samples in an alternative
# condition.
color_palette = build_color_palette(num_items=14 - 4 + 1,
weeks_before_switch=7)
# color_palette = sns.cubehelix_palette(11, start=.5, rot=-.75)
# update matplotlib params for bigger fonts, ticks:
mpl.rcParams.update({
'font.size': 16, 'axes.titlesize': 17, 'axes.labelsize': 15,
'xtick.labelsize': 10, 'ytick.labelsize': 13,
'font.weight': 600,
'axes.labelweight': 600, 'axes.titleweight': 600})
# Plot with Seaborn
if facet_row:
plt.figure(figsize=(4, 8))
else:
plt.figure(figsize=(6, 12))
sns.set(style="ticks")
# prepare the max and min axes values if we are forcing them to same range
pc_colnames = [col for col in plot_data.columns
if 'principal component' in col]
max_value = plot_data[pc_colnames].max(axis=0).max()
min_value = plot_data[pc_colnames].min(axis=0).min()
axis_max = math.ceil(max_value * 100) / 100.0
axis_min = math.floor(min_value * 100) / 100.0
def base_plot(**kwargs):
plot = sns.FacetGrid(plot_data,
hue='week', palette=color_palette,
size=3, aspect=1,
**kwargs)
plot = (plot.map(plt.scatter, x_axis_label, y_axis_label,
edgecolor="w", s=60).add_legend())
return plot
plot_args = {}
if facet_row:
plot_args['row'] = 'oxy'
plot_args['col'] = 'rep'
if uniform_axes:
plot_args['xlim'] = (axis_min, axis_max)
plot_args['ylim'] = (axis_min, axis_max)
print(plot_args)
g = base_plot(**plot_args)
filename = concat_dir_and_filename(
plot_dir, 'pca_of_top_{}_percent--'.format(top_percent))
# prepare a filename, depending on whether all taxonomy or only genus
# is used.
if genus_only:
filename += 'genus_only'
else:
filename += 'all_taxa'
if uniform_axes:
filename += '_unif_axes_'
if facet_row:
filename += '--faceted.pdf'
else:
filename += '.pdf'
if savefig:
g.fig.savefig(filename)
