corrections_clc.py

# This script will make corrected figures based on Howard's comments. 

# First we will import the necessary modules.
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.cluster import KMeans as KM
from sklearn.preprocessing import KBinsDiscretizer as Bin
import matplotlib.patches as patches
from statannotations.Annotator import Annotator
from statsmodels.stats.multicomp import pairwise_tukeyhsd
from scipy.stats import f_oneway
import scipy.stats as stat
from scikit_posthocs import posthoc_tukey
from io import StringIO
import os.path

# Define paths.
All = './All/'
No_1 = './No_1Mo/'
PK = './Only_PKO/'
syn = './prism/syn/'
wbh = './prism/wbh/'

# Turn off pandas warnings, not clear on why they appear sometimes and not
# others. In both cases, the desired result is the same. Will need to dig in on
# this.
pd.options.mode.chained_assignment = None 

# We will read in the data
df = pd.read_excel('Concat_with_all.xlsx')

# We will rename the 'Frequency' column to 'Heteroplasmy' and the 'Reference
# Position' column to 'Mitochondrial Position as these are more appropriate
# names.
df = df.rename(columns={'Frequency' : 'Heteroplasmy', 'Reference Position' :
    'Mitochondrial Position'})

# We will also replace 'Syn' with 'Synaptosome'
df = df.replace('Syn', 'Synaptosome')
df2 = df[df.Strain != 'Polg-W402A']
df1 = df[df.Strain != 'WT\n1Mo']
df1 = df[df.Strain != 'PKO\n1Mo']

# First we'll create a count column so that the number of samples can be
# counted and hence be compared to each other.
df['count'] = 0
df1['count'] = 0

# We now create variables only for the strains of interest (WT, Polg, and
# Polg-PKO) and then from these subsets create further subsets, syn and
# homogenate.
W = df1.loc[df1['Strain'] == 'WT']
WS = W.loc[W['Sample_type'] == 'Synaptosome']
WH = W.loc[W['Sample_type'] == 'Homogenate']
P = df1.loc[df1['Strain'] == 'Polg']
PS = P.loc[P['Sample_type'] == 'Synaptosome']
PH = P.loc[P['Sample_type'] == 'Homogenate']
K = df1.loc[df1['Strain'] == 'Polg-PKO']
KS = K.loc[K['Sample_type'] == 'Synaptosome']
KH = K.loc[K['Sample_type'] == 'Homogenate']
W4 = df1.loc[df1['Strain'] == 'Polg-W402A']
W4S = W4.loc[W4['Sample_type'] == 'Synaptosome']
W4H = W4.loc[W4['Sample_type'] == 'Homogenate']

# Perform elbow method on all of the dataframes above to on the Heteroplasmy
# column to see how many clusters they each have. To do so, assign a variable
# to the heteroplasmy column, connvert into an np array, and change the shape
# to (-1,1).
x = WS['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method WT Syn')
plt.savefig('WT_Syn_Het_Elbow.png')

# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
WS_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (WS_bins[:-1] + WS_bins[1:])/2
WS.loc[:,'Heteroplasmy Cluster'] = pd.cut(WS['Heteroplasmy'], WS_bins, labels=center)

# Repeat for Homogenate.
x = WH['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method WT Homogenate')
plt.savefig('WT_WBH_Het_Elbow.png')

# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
WH_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (WH_bins[:-1] + WH_bins[1:])/2
WH.loc[:,'Heteroplasmy Cluster'] = pd.cut(WH['Heteroplasmy'], WH_bins, labels=center)

# Repeat for Polg Syn
x = PS['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method Polg Syn')
plt.savefig('Polg_Syn_Het_Elbow.png')

# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
PS_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (PS_bins[:-1] + PS_bins[1:])/2
PS.loc[:,'Heteroplasmy Cluster'] = pd.cut(PS['Heteroplasmy'], PS_bins, labels=center)

# Repeat for Polg Homogenate
x = PH['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method Polg Homogenate')
plt.savefig('Polg_WBH_Het_Elbow.png')

# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
PH_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (PH_bins[:-1] + PH_bins[1:])/2
PH['Heteroplasmy Cluster'] = pd.cut(PH['Heteroplasmy'], PH_bins, labels=center)

# Repeat for Polg-PKO Syn.
x = KS['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method Polg-PKO Syn')
plt.savefig('Polg-PKO_Syn_Het_Elbow.png')

# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
KS_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (KS_bins[:-1] + KS_bins[1:])/2
KS['Heteroplasmy Cluster'] = pd.cut(KS['Heteroplasmy'], KS_bins, labels=center)

# Repeat for Polg-PKO Homogenate/
x = KH['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method Polg-PKO Homogenate')
plt.savefig('Polg-PKO_WBH_Het_Elbow.png')
# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
KH_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (KH_bins[:-1] + KH_bins[1:])/2
KH['Heteroplasmy Cluster'] = pd.cut(KH['Heteroplasmy'], KH_bins, labels=center)

# Repeat for Polg-W402A Syn.
x = W4S['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method Polg-W402A Syn')
plt.savefig('Polg-W402a_Syn_Het_Elbow.png')

# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
W4S_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (W4S_bins[:-1] + W4S_bins[1:])/2
W4S['Heteroplasmy Cluster'] = pd.cut(W4S['Heteroplasmy'], W4S_bins, labels=center)

# Repeat for Polg-W402A Homogenate/
x = W4H['Heteroplasmy']
x = np.array(x)
x = x.reshape(-1,1)

# Use a range of k to run the elbow method on.
K = range(1,7)

# Create a list of distortions, and use kmeans to fit the data to. Finally,
# append the distortion list with the kmeans data.
distortions = []
for k in K:
    km = KM(n_clusters=k, random_state=42)
    km.fit(x)
    distortions.append(km.inertia_)

# Plot the elbow results and save the figure.
plt.figure(figsize=(16,8))
plt.plot(K, distortions, 'bx-')
plt.ylabel('Distortion')
plt.title('The Elbow Method Polg-W402A Homogenate')
plt.savefig('Polg-W402A_WBH_Het_Elbow.png')

# Each time the kmeans algo is run, an unsorted list of cluster centers
# are outputeed, are in the form of floats, and not sorted or rounded.
# Therefore, they will be sorted and rounded to the nearest integer
km.cluster_centers_ = np.sort(km.cluster_centers_,axis=None).round()

# Now we will create bins using the sorted, rounded kmeans cluseters
W4H_bins = km.cluster_centers_

# Bin centers are defined by taking half the sum of consecutive bins. The
# resulting bin centers will be put into a new column in the dataframe called
# 'Heteroplasmy Cluster'
center = (W4H_bins[:-1] + W4H_bins[1:])/2
W4H['Heteroplasmy Cluster'] = pd.cut(W4H['Heteroplasmy'], W4H_bins, labels=center)

# Create concatenated files Syn and WBH so that graphs can be made for the
# cluster centers.
Syn = pd.concat([WS, PS, KS, W4S], axis=0)
WBH = pd.concat([WH, PH, KH, W4H], axis=0)
com = pd.concat([Syn, WBH], axis=0)
order = ['Synaptosome', 'Homogenate']
hue_order = ['WT', 'Polg', 'Polg-PKO', 'Polg-W402A']

# Now we will plot out the Heteroplasmy Clusters vs Reference Position as well
# as vs gene. We will use the facetgrid from seabron so that the two (Syn and
# WBH) can be on the same figure.
x = sns.FacetGrid(data=com, col='Sample_type', col_wrap=1, despine=True,
hue='Strain',palette='colorblind', col_order=order, hue_order=hue_order, aspect=2)

# The graph title is adjusted to be on top of the graph, but not intruding on
# it, and it is emboldened.
x.fig.subplots_adjust(top=0.8)
x.fig.suptitle('Heteroplasmy Clusters', fontsize=28,weight='bold')
labels = hue_order
colors = sns.color_palette('colorblind').as_hex()[:len(labels)]

# Seaborn has an issue pulling the colors for the legend in the facetgrid.
# Therefore a patch from matplotlib.patches is called to correct for that.
handles = [patches.Patch(color=col, label=lab) for col, lab in zip(colors,
    labels)]

# Next, seaborn is called to add the legened as outlined in the patch
x.add_legend(legend_data={lab: hand for lab, hand in zip(labels, handles)})

# A histplot will be inserted into the facetgrid. X will be Reference
# Position, Y will be Heteroplasmy Cluster. The alpha command represents
# transperancy of the colors. The higher it is, the less transparent. 
x.map(sns.histplot, 'Mitochondrial Position', 'Heteroplasmy'
        ' Cluster',alpha=0.4)
#.set(yscale='log')

# Since facetgrids puts the title of each graph with '=', it is changes so that
# only the name without '=' is posted on the title of each graph.
x.set_titles(col_template='{col_name}',weight='bold')

# Set y limit from 0 to 100 so that 100% heteroplasmy is noticable.
x.set(ylim=(0,100))
plt.savefig(os.path.join(No_1,'Heteroplasmy_clusters.png'))


# Now we'll do it for the gene_name.
x = sns.FacetGrid(data=com, col='Sample_type', col_wrap=1, despine=True,
hue='Strain',palette='colorblind', col_order=order, hue_order=hue_order, aspect=4)

# The graph title is adjusted to be on top of the graph, but not intruding on
# it, and it is emboldened.
x.fig.subplots_adjust(top=0.8)
x.fig.suptitle('Heteroplasmy Clusters', fontsize=28,weight='bold')
labels = hue_order
colors = sns.color_palette('colorblind').as_hex()[:len(labels)]
# Seaborn has an issue pulling the colors for the legend in the facetgrid.
# Therefore a patch from matplotlib.patches is called to correct for that.
handles = [patches.Patch(color=col, label=lab) for col, lab in zip(colors,
    labels)]

# Next, seaborn is called to add the legened as outlined in the patch
x.add_legend(legend_data={lab: hand for lab, hand in zip(labels, handles)})
# A scatterplot will be inserted into the facetgrid. X will be Reference
# Position, Y will be Heteroplasmy Cluster. The alpha command represents
# transperancy of the colors. The higher it is, the less transparent. 
x.map(sns.histplot, 'Gene_name', 'Heteroplasmy'
        ' Cluster',alpha=0.4)
#.set(yscale='log')
# Since we want all gene names to be on the graph, the xticklables will be
# rotated 45 degress.
[plt.setp(ax.get_xticklabels(), rotation=45) for ax in x.axes.flat]

# Since facetgrids puts the title of each graph with '=', it is changes so that
# only the name without '=' is posted on the title of each graph.
x.set_titles(col_template='{col_name}',weight='bold')
# Set y limit from 0 to 100 so that 100% heteroplasmy is noticable.
x.set(ylim=(0,100))
plt.savefig(os.path.join(No_1,'Heteroplasmy_clusters_gene.png'), bbox_inches='tight')

Syn.to_excel(os.path.join(syn,'het_cluster_syn.xlsx'),index=False)
WBH.to_excel(os.path.join(wbh,'het_cluster_wbh.xlsx'),index=False)

# Now we'll do it for each of the strains separately, Syn and WBH will be two
# separate figures .
x = sns.FacetGrid(data=Syn, col='Strain', col_wrap=1, despine=True,
hue='Strain',palette='colorblind', col_order=hue_order, hue_order=hue_order, aspect=4)

# The graph title is adjusted to be on top of the graph, but not intruding on
# it, and it is emboldened.
x.fig.subplots_adjust(top=0.8)
x.fig.suptitle('Synaptosome - Heteroplasmy Clusters', fontsize=28,weight='bold')

# A histplot will be inserted into the facetgrid. X will be Reference
# Position, Y will be Heteroplasmy Cluster. The alpha command represents
# transperancy of the colors. The higher it is, the less transparent. 
x.map(sns.histplot, 'Mitochondrial Position', 'Heteroplasmy'
        ' Cluster',alpha=1)

# Since facetgrids puts the title of each graph with '=', it is changes so that
# only the name without '=' is posted on the title of each graph.
x.set_titles(col_template='{col_name}',weight='bold')
# Set y limit from 0 to 100 so that 100% heteroplasmy is noticable.
x.set(ylim=(0,100))
plt.savefig(os.path.join(No_1,'Heteroplasmy_clusters_Syn_Strain.png'), bbox_inches='tight')

# Now we'll do it for gene name for Syn.
x = sns.FacetGrid(data=Syn, col='Strain', col_wrap=1, despine=True,
hue='Strain',palette='colorblind', col_order=hue_order, hue_order=hue_order, aspect=4)

# The graph title is adjusted to be on top of the graph, but not intruding on
# it, and it is emboldened.
x.fig.subplots_adjust(top=0.8)
x.fig.suptitle('Synaptosome - Heteroplasmy Clusters', fontsize=28,weight='bold')

# A histplot will be inserted into the facetgrid. X will be Reference
# Position, Y will be Heteroplasmy Cluster. The alpha command represents
# transperancy of the colors. The higher it is, the less transparent. 
x.map(sns.histplot, 'Gene_name', 'Heteroplasmy'
        ' Cluster',alpha=1)

# Since we want all gene names to be on the graph, the xticklables will be
# rotated 45 degress.
[plt.setp(ax.get_xticklabels(), rotation=45) for ax in x.axes.flat]

# Since facetgrids puts the title of each graph with '=', it is changes so that
# only the name without '=' is posted on the title of each graph.
x.set_titles(col_template='{col_name}',weight='bold')

# Set y limit from 0 to 100 so that 100% heteroplasmy is noticable.
x.set(ylim=(0,100))
plt.savefig(os.path.join(No_1,'Heteroplasmy_clusters_Syn_Strain_Gene.png'), bbox_inches='tight')

# Now ww'll do it for WBH.
x = sns.FacetGrid(data=WBH, col='Strain', col_wrap=1, despine=True,
hue='Strain',palette='colorblind', col_order=hue_order, hue_order=hue_order, aspect=4)

# The graph title is adjusted to be on top of the graph, but not intruding on
# it, and it is emboldened.
x.fig.subplots_adjust(top=0.8)
x.fig.suptitle('Homogenate - Heteroplasmy Clusters', fontsize=28,weight='bold')

# A histplot will be inserted into the facetgrid. X will be Reference
# Position, Y will be Heteroplasmy Cluster. The alpha command represents
# transperancy of the colors. The higher it is, the less transparent. 
x.map(sns.histplot, 'Mitochondrial Position', 'Heteroplasmy'
        ' Cluster',alpha=1)

# Since facetgrids puts the title of each graph with '=', it is changes so that
# only the name without '=' is posted on the title of each graph.
x.set_titles(col_template='{col_name}',weight='bold')

# Set y limit from 0 to 100 so that 100% heteroplasmy is noticable.
x.set(ylim=(0,100))
plt.savefig(os.path.join(No_1,'Heteroplasmy_clusters_WBH_Strain.png'), bbox_inches='tight')

# Now we'll do it for gene name for Syn.
x = sns.FacetGrid(data=WBH, col='Strain', col_wrap=1, despine=True,
hue='Strain',palette='colorblind', col_order=hue_order, hue_order=hue_order, aspect=4)

# The graph title is adjusted to be on top of the graph, but not intruding on
# it, and it is emboldened.
x.fig.subplots_adjust(top=0.8)
x.fig.suptitle('Homogenate - Heteroplasmy Clusters', fontsize=28,weight='bold')

# A histplot will be inserted into the facetgrid. X will be Reference
# Position, Y will be Heteroplasmy Cluster. The alpha command represents
# transperancy of the colors. The higher it is, the less transparent. 
x.map(sns.histplot, 'Gene_name', 'Heteroplasmy'
        ' Cluster',alpha=1)
# Since we want all gene names to be on the graph, the xticklables will be
# rotated 45 degress.
[plt.setp(ax.get_xticklabels(), rotation=45) for ax in x.axes.flat]

# Since facetgrids puts the title of each graph with '=', it is changes so that
# only the name without '=' is posted on the title of each graph.
x.set_titles(col_template='{col_name}',weight='bold')

# Set y limit from 0 to 100 so that 100% heteroplasmy is noticable.
x.set(ylim=(0,100))
plt.savefig(os.path.join(No_1,'Heteroplasmy_clusters_WBH_Strain_Gene.png'), bbox_inches='tight')

# Now we will create separate dataframes with minimum heteroplasmies in order
# to calculate and graph if there are significant differences between the
# strains.
Syn10 = Syn[Syn['Heteroplasmy'] >= 10]
Syn20 = Syn[Syn['Heteroplasmy'] >= 20]
Syn30 = Syn[Syn['Heteroplasmy'] >= 30]
Syn50 = Syn[Syn['Heteroplasmy'] >= 50]
WBH10 = WBH[WBH['Heteroplasmy'] >= 10]
WBH20 = WBH[WBH['Heteroplasmy'] >= 20]
WBH30 = WBH[WBH['Heteroplasmy'] >= 30]
WBH50 = WBH[WBH['Heteroplasmy'] >= 50]


# Create a new column in both Syn and WBH named 'Heteroplasmic Burden'. This
# column is a multiple of the length of mutations per animal multiplied by the
# mean heteroplasmy.

# First we create a column 'Len of name' which give the length (i.e. the number
# of mutations) per animal. This will be done for each strain in a separate
# dataframe because otherwise the entire heteroplasmy gets taken into
# consideration.
WS['Len of name'] = WS.groupby(['Name'])['count'].transform('count')
PS['Len of name'] = PS.groupby(['Name'])['count'].transform('count')
KS['Len of name'] = KS.groupby(['Name'])['count'].transform('count')
W4S['Len of name'] = W4S.groupby(['Name'])['count'].transform('count')
WH['Len of name'] = WH.groupby(['Name'])['count'].transform('count')
PH['Len of name'] = PH.groupby(['Name'])['count'].transform('count')
KH['Len of name'] = KH.groupby(['Name'])['count'].transform('count')
W4H['Len of name'] = W4H.groupby(['Name'])['count'].transform('count')

# Next we create a column 'Heteroplasmic Burden' by multiplying 'Len of name by
# the mean heteroplasmy. * Let's figure out how to get mean of heteroplasmy per
# animal instead of mean of entire strain, probably won't be much different,
# but more precise.
WS['Heteroplasmic Burden'] = WS['Len of name'] * WS['Heteroplasmy'].mean()
PS['Heteroplasmic Burden'] = PS['Len of name'] * PS['Heteroplasmy'].mean()
KS['Heteroplasmic Burden'] = KS['Len of name'] * KS['Heteroplasmy'].mean() 
W4S['Heteroplasmic Burden'] = W4S['Len of name'] * W4S['Heteroplasmy'].mean() 
WH['Heteroplasmic Burden'] = WH['Len of name'] * WH['Heteroplasmy'].mean()
PH['Heteroplasmic Burden'] = PH['Len of name'] * PH['Heteroplasmy'].mean()
KH['Heteroplasmic Burden'] = KH['Len of name'] * KH['Heteroplasmy'].mean() 
W4H['Heteroplasmic Burden'] = W4H['Len of name'] * W4H['Heteroplasmy'].mean() 

NewSyn = pd.concat([WS, PS, KS, W4S], axis=0)
NewWBH = pd.concat([WH, PH, KH, W4H], axis=0)


NewSyn.to_excel(os.path.join(syn,'Heteroburdensyn.xlsx'),index=False)
NewWBH.to_excel(os.path.join(wbh,'Heteroburdenwbh.xlsx'),index=False)

# We will create StrainS and StrainW for unique strains in NewSyn and NewWBH
# respectively. Do stats on them and graph them.
StrainS = NewSyn.Strain.unique()
StrainW = NewWBH.Strain.unique()

df3 = []
for s in StrainS:
    df3.append(NewSyn[NewSyn['Strain'] == s]['Heteroplasmic Burden'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=NewSyn['Heteroplasmic Burden'], groups=NewSyn['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'Syn_heteroburden_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(NewSyn, val_col="Heteroplasmic Burden",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'Heteroplasmic Burden'
order = ['WT', 'Polg', 'Polg-PKO', 'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
sns.set_palette('colorblind')
fig, axes = plt.subplots(1, 2,figsize=(10, 5))
fig.suptitle('Average Heteroplasmic Burden', weight='bold')
ax = sns.violinplot(ax=axes[0],data=NewSyn,x=x, y=y,order=order,color='red')
#ax = sns.swarmplot(ax=axes[0], data=NewSyn,x=x, y=y,order=order,color='black',
 #       size=3)
ax.set_title('Synaptosome')
ax.set_ylim(top=max(NewSyn['Heteroplasmic Burden']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=NewSyn, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)

# Now for WBH.
df3 = []
for s in StrainW:
    df3.append(NewWBH[NewWBH['Strain'] == s]['Heteroplasmic Burden'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=NewWBH['Heteroplasmic Burden'], groups=NewWBH['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'WBH_heteroburden_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(NewWBH, val_col="Heteroplasmic Burden",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

ax = sns.violinplot(ax=axes[1],data=NewWBH,x=x, y=y,order=order,color='blue')
#ax = sns.swarmplot(ax=axes[1], data=NewWBH,x=x, y=y,order=order,color='black',
 #       size=3)
ax.set_title('Homogenate')
ax.set_ylim(top=max(NewWBH['Heteroplasmic Burden']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=NewWBH, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)
plt.savefig(os.path.join(No_1,'Het_burden.png'))

# Group the dataframes by Name and Strain and aggregate by count in order to be
# able to do stats on them. Reset the index so that the variable is available
# downstream
Syn = Syn.groupby(['Name', 'Strain'],observed=True).count().reset_index()
Syn10 = Syn10.groupby(['Name', 'Strain'],observed=True).count().reset_index()
Syn20 = Syn20.groupby(['Name', 'Strain'],observed=True).count().reset_index()
Syn30 = Syn30.groupby(['Name', 'Strain'],observed=True).count().reset_index()
Syn50 = Syn50.groupby(['Name', 'Strain'],observed=True).count().reset_index()
WBH = WBH.groupby(['Name', 'Strain'],observed=True).count().reset_index()
WBH10 = WBH10.groupby(['Name', 'Strain'],observed=True).count().reset_index()
WBH20 = WBH20.groupby(['Name', 'Strain'],observed=True).count().reset_index()
WBH30 = WBH30.groupby(['Name', 'Strain'],observed=True).count().reset_index()
WBH50 = WBH50.groupby(['Name', 'Strain'],observed=True).count().reset_index()

Syn10.to_excel(os.path.join(syn,'10%SYN.xlsx'),index=False)
Syn20.to_excel(os.path.join(syn,'20%SYN.xlsx'),index=False)
Syn30.to_excel(os.path.join(syn,'30%SYN.xlsx'),index=False)
Syn50.to_excel(os.path.join(syn,'50%SYN.xlsx'),index=False)

WBH10.to_excel(os.path.join(wbh,'10%WBH.xlsx'),index=False)
WBH20.to_excel(os.path.join(wbh,'20%WBH.xlsx'),index=False)
WBH30.to_excel(os.path.join(wbh,'30%WBH.xlsx'),index=False)
WBH50.to_excel(os.path.join(wbh,'50%WBH.xlsx'),index=False)

# Get all unique strains into a variable called 'Strain'. Then create an empty
# dataframe 'df3' and for each unique value in strain, append the df values
# from 'count'. Finally, perform f_oneway statistics on df3.
Strain10 = Syn10.Strain.unique()
Strain20 = Syn20.Strain.unique()
Strain30 = Syn30.Strain.unique()
Strain50 = Syn50.Strain.unique()

df3 = []
for s in Strain10:
    df3.append(Syn10[Syn10['Strain'] == s]['count'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=Syn10['count'], groups=Syn10['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'Syn10_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(Syn10, val_col="count",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO', 'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
sns.set_palette('colorblind')
fig, axes = plt.subplots(2, 2,figsize=(17, 10))
fig.suptitle('Synaptosomes', weight='bold')
ax = sns.barplot(ax=axes[0,0],data=Syn10,x=x, y=y,order=order,
        facecolor=('green'),edgecolor='.2')
ax.set_title('Minimum 10% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[0,0],data=Syn10,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(Syn10['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=Syn10, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)

df3 = []
for s in Strain20:
    df3.append(Syn20[Syn20['Strain'] == s]['count'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=Syn20['count'], groups=Syn20['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'Syn20_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(Syn20, val_col="count",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO',   'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
ax = sns.barplot(ax=axes[0,1],data=Syn20,x=x, y=y,order=order,
                        facecolor=('blue'),edgecolor='.2')
ax.set_title('Minimum 20% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[0,1],data=Syn20,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(Syn20['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=Syn20, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)
df3 = []
for s in Strain30:
    df3.append(Syn30[Syn30['Strain'] == s]['count'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=Syn30['count'], groups=Syn30['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'Syn30_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(Syn30, val_col="count",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO', 'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
ax = sns.barplot(ax=axes[1,0],data=Syn30,x=x, y=y,order=order,
                        facecolor=('purple'),edgecolor='.2')
ax.set_title('Minimum 30% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[1,0],data=Syn30,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(Syn30['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=Syn30, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO', 'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
ax = sns.barplot(ax=axes[1,1],data=Syn50,x=x, y=y,order=order,color='red')
           #             facecolor=('red'),edgecolor='.2')
ax.set_title('Minimum 50% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[1,1],data=Syn50,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(Syn50['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=Syn30, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)

plt.savefig(os.path.join(No_1,'Het_sep_segs_Syn.png'))

Strain10 = WBH10.Strain.unique()
Strain20 = WBH20.Strain.unique()
Strain30 = WBH30.Strain.unique()
Strain50 = WBH50.Strain.unique()
df3 = []
for s in Strain10:
    df3.append(WBH10[WBH10['Strain'] == s]['count'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=WBH10['count'], groups=WBH10['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'WBH10_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(WBH10, val_col="count",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO', 'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
sns.set_palette('colorblind')
fig, axes = plt.subplots(2, 2,figsize=(17, 10))
fig.suptitle('Homogenate', weight='bold')
ax = sns.barplot(ax=axes[0,0],data=WBH10,x=x, y=y,order=order,
                        facecolor=('green'),edgecolor='.2')
ax.set_title('Minimum 10% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[0,0],data=WBH10,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(WBH10['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=WBH10, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)

df3 = []
for s in Strain20:
    df3.append(WBH20[WBH20['Strain'] == s]['count'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=WBH20['count'], groups=WBH20['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'WBH20_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(WBH20, val_col="count",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO', 'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
ax = sns.barplot(ax=axes[0,1],data=WBH20,x=x, y=y,order=order,
                        facecolor=('blue'),edgecolor='.2')
ax.set_title('Minimum 20% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[0,1],data=WBH20,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(WBH20['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=WBH20, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)
df3 = []
for s in Strain30:
    df3.append(WBH30[WBH30['Strain'] == s]['count'])
F = f_oneway(*df3)
# Cast the f_oneway statistics into a string so that it can then be saved as a
# dataframe. Use the StringIO to implement a file-like class on the string
# ('F') so that it can be read in as a CSV and hence become a dataframe.
F = str(F)
F = StringIO(F)
F = pd.read_csv(F, sep=';')

# Perform groupwise comparisons using tukey HSD
tukey = pairwise_tukeyhsd(endog=WBH30['count'], groups=WBH30['Strain'],
        alpha=0.05)

# Save the tukey data as a dataframe and append the F stats from F dataframe
# into it. Finally, export the dataframe to csv as 'Syn_stats.csv'.
tukey = pd.DataFrame(data=tukey._results_table.data[1:],
                columns=tukey._results_table.data[0])
tukey = pd.concat([tukey, F])
tukey.to_csv(os.path.join(No_1,'WBH30_stats.csv'),index=False)

# In order to annotate the resulting graphs with stars for statistical
# significance, the tukey results need to be put into a simple matrix (just
# showing the p values, not other parameters such as meandiff, lower and upper
# etc).
tukey_df = posthoc_tukey(WBH30, val_col="count",
                group_col="Strain")

# The matrix needs to be converted to a non-redundant list of comparisons with
# the p-value. This is done by removing the lower half and diagonal of the
# matrix and turning the matrix format into a long dataframe using melt(). The
# code and resulting dataframe are shown below.
remove = np.tril(np.ones(tukey_df.shape), k=0).astype("bool")
tukey_df[remove] = np.nan
molten_df = tukey_df.melt(ignore_index=False).reset_index().dropna()

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO', 'Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
ax = sns.barplot(ax=axes[1,0],data=WBH30,x=x, y=y,order=order,
                        facecolor=('purple'),edgecolor='.2')
ax.set_title('Minimum 30% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[1,0],data=WBH30,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(WBH30['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=WBH30, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)

# x, y, and order are defined so that they can be used in the graphs below.
x = "Strain"
y = 'count'
order = ['Polg', 'Polg-PKO','Polg-W402A']
sns.set_style("whitegrid", {'axes.grid' : False})
ax = sns.barplot(ax=axes[1,1],data=WBH50,x=x, y=y,order=order,
                        facecolor=('red'),edgecolor='.2')
ax.set_title('Minimum 50% Heteroplasmy')

# Add a swarmplot to visualize the individual datapoints on the barplot. Color
# it black so that the points are easy to spot.
ax = sns.swarmplot(ax=axes[1,1],data=WBH50,x=x, y=y,color='.2', order=order)
ax.set_ylabel('Number of Mutations')
ax.set_ylim(top=max(WBH50['count']) + 2)

# In order to only annotate the graph where there are significant differences,
# the dataframe 'molten_df', which contains the p values, will be filtered so
# that only significant p values (<= 0.05) are in there. Note, if all notations
# are desireable, skip the filtering step.
molten_df = molten_df.loc[molten_df['value'] <= 0.05]

# The pairs for multiple comparisons is defined as all strains in the p value
# table.
pairs = [(i[1]["index"], i[1]["variable"]) for i in molten_df.iterrows()]

# A list of p values is generated from the molten_df dataframe. The annotator
# is then defnied and configured to annotate the graph with stars using the p
# values from the list.
p_values = [i[1]["value"] for i  in molten_df.iterrows()]
annotator = Annotator(ax, pairs, data=Syn30, x=x, y=y, order=order)
annotator.configure(text_format="star", loc="inside")
annotator.set_pvalues_and_annotate(p_values)

plt.savefig(os.path.join(No_1,'Het_sep_segs_WBH.png'))

# Define a function that will assign the variable x to the column of interest,
# in this case 'Heteroplasmy'. It will then turn the column into an np array
# and reshape it to (-1,1) in order to run kmeans on  it. 
#def Hetero(x):
 #   x = np.array(x)
  #  x = x.reshape(-1,1)
   # print(x)
   # return x
#PS['Heteroplasmy'] = PS['Heteroplasmy'].apply(Hetero)
#print(type(PS['Heteroplasmy']))
#km = km    
#def kmeans(x):
 #   K = range(1,2)
  #  distortions = []
   # for k in K:
    #    i = km(n_clusters=k, random_state=42)
     #   i.fit(x)
      #  distortions.append(i.inertia_)
       # plt.figure(figsize=(16,8))
        #plt.plot(k, distortions, 'bx-')
        #plt.ylabel('Distortions')
        #plt.title('The Elbow Method')
        #plt.savefig('x.png')
#PS['Heteroplasmy'] = PS['Heteroplasmy'].apply(kmeans)