def age_v_family_history(df, column, outfile=None):
''' Boxplot showing median age at diagnosis between
pateints with and without a family history.
'''
df = df.dropna(subset=[column, 'family_history'])
df = df[df['family_history'] != 'unknown']
df['family_history'] = convert2category(df['family_history'],
['yes', 'no'])
fig, ax = plt.subplots(figsize=(15, 12))
sns.set(font_scale=1.5)
sns.set_style("whitegrid")
sns.set_palette('Greys')
ax = sns.boxplot(data=df, x='family_history', y=column)
ax.set(title="", ylabel='Age at Diagnosis', xlabel='', ylim=(0, 100))
name_list = pm.rename_xtick(df, 'family_history')
ax.set_xticklabels(name_list)
family_history = df[df['family_history'] == 'yes'][column]
no_family_history = df[df['family_history'] == 'no'][column]
fam_p_val = str(
pm.RoundToSigFigs(
stats.ranksums(family_history, no_family_history)[1], 2))
pm.line_between_plots(ax,
x1=0,
x2=1,
height=90,
string='p = ' + fam_p_val,
fontsize=18)
if outfile:
ax.figure.savefig(outfile)
def fh_vs_genetic_diagnosis(df, outfile=None):
''' Countplot displaying counts for each variant classification
split by family history.
'''
df = df.dropna(subset=['New Category'])
df = df[df['family_history'] != 'unknown']
df['family_history'] = convert2category(df['family_history'],
['yes', 'no'])
sns.set_style("whitegrid")
sns.set_palette('Greys')
fig, ax = plt.subplots(figsize=(14, 12))
name_list = pm.rename_xtick(df, 'family_history', counts=False)
ax = sns.countplot(data=df, x='family_history', hue='New Category')
ax.set(title=" . ",
xlabel="",
ylabel="Samples",
xticklabels=name_list,
ylim=(0, 500))
ax.legend(bbox_to_anchor=(0., 1.0, 1., .102),
loc=8,
ncol=3,
mode="expand",
borderaxespad=0.)
# label percentages above bars
fh_total = len(df[df['family_history'] == 'yes'])
no_fh_total = len(df[df['family_history'] == 'no'])
for p, total in zip(
ax.patches,
[fh_total, no_fh_total, fh_total, no_fh_total, fh_total, no_fh_total]):
h = p.get_height()
ax.text(p.get_x() + 0.15,
h + 3,
str('{:.1f}'.format((h / total) * 100) + "%"),
ha='center')
# STATS: As I am an anlasying a large number of samples, the chi-square test is approriate
# and will yield a simliar result as the Fisher-Freeman-Halton test
table = df.groupby(['family_history',
'New Category']).size().unstack(1).fillna(0.0)
array = table.values.tolist()
chi2, pvalue, dof, expected = stats.chi2_contingency(array)
pm.line_between_plots(axs=ax,
x1=0,
x2=1,
height=array[1][2] + 20,
extend=10,
string="p = " + str(pm.RoundToSigFigs(pvalue, 3)),
fontsize=14)
if outfile:
ax.figure.savefig(outfile)
def gender_vs_genetic_diagnosis(df, outfile=None):
''' Countplot displaying counts for each variant classification
split by gender
'''
df = df.dropna(subset=['Gender', 'New Category'])
sns.set(font_scale=1.2)
sns.set_style("whitegrid")
sns.set_palette("Greys")
fig, ax = plt.subplots(figsize=(14, 12))
ax = sns.countplot(data=df, x='Gender', hue='New Category')
name_list = pm.rename_xtick(df, 'Gender', counts=False)
ax.set_xticklabels(name_list)
ax.set(title=' . ', xlabel="", ylabel="Samples", ylim=(0, 700))
ax.legend(bbox_to_anchor=(0., 1.0, 1., .102),
loc=8,
ncol=3,
mode="expand",
borderaxespad=0.)
# STATS
gen_total = len(df[df['Gender'] == 'Male'])
no_gen_total = len(df[df['Gender'] == 'Female'])
table = df.groupby(['Gender', 'New Category']).size().unstack(1)
array = table.values.tolist()
chi2, pvalue, dof, expected = stats.chi2_contingency(array)
pm.line_between_plots(axs=ax,
x1=0,
x2=1,
height=array[0][2] + 45,
extend=10,
string="p = " + str(round(pvalue, 3)),
fontsize=14)
# label percentages above bars
for p, total in zip(ax.patches, [
gen_total, no_gen_total, gen_total, no_gen_total, gen_total,
no_gen_total
]):
h = p.get_height()
ax.text(p.get_x() + 0.15,
h + 3,
str(round((h / total), 3) * 100)[:4] + "%",
ha='center')
if outfile:
fig = ax.get_figure()
fig.savefig(outfile, bbox_inches='tight')
return ax
def variant_class_violin(df, column, title='', outfile=None):
''' Produces a violin plot of the age at surgery vs the variant class.'''
df = df.dropna(subset=[column, 'New Category'])
y_label = column.replace("age at diagnosis", "Age at Diagnosis")
name_list = pm.rename_xtick(df,
'New Category',
NA=True,
validated_path=True)
fig, ax = plt.subplots(figsize=(15, 12))
sns.set(font_scale=1.5)
sns.set_style("whitegrid")
sns.set_palette('Greys')
ax = sns.violinplot(data=df, x='New Category', y=column, cut=0)
ax.set(title=title,
ylabel=y_label,
xlabel='',
xticklabels=name_list,
ylim=(0, 105))
# Seperate and clean each variant class Series
pathogenic = df[df['New Category'] == "Pathogenic/Likely Pathogenic"]
pathogenic = sf.validated_only(pathogenic)[column].dropna()
damaging = df[df['New Category'] == "VUS"][column].dropna()
benign = df[df['New Category'] ==
"Likely Benign / No Variant"][column].dropna()
# Get unpaired ranksum wilcoxon p-values between each new category and place them upon the plot
path_ben_p_val = str(
pm.RoundToSigFigs(stats.ranksums(pathogenic, benign)[1], 3))
pm.line_between_plots(ax,
x1=0,
x2=2,
height=benign.max() + 15,
string="p = " + path_ben_p_val,
fontsize=15)
path_dam_p_val = str(
pm.RoundToSigFigs(stats.ranksums(pathogenic, damaging)[1], 3))
pm.line_between_plots(ax,
x1=0,
x2=1,
height=benign.max() + 10,
string="p = " + path_dam_p_val,
fontsize=15)
dam_ben_p_val = str(
pm.RoundToSigFigs(stats.ranksums(damaging, benign)[1], 3))
pm.line_between_plots(ax,
x1=1,
x2=2,
height=benign.max() + 5,
string="p = " + dam_ben_p_val,
fontsize=15)
if outfile:
ax.figure.savefig(outfile)
return ax
def age_group_v_pathogenic_piechart(df, outfile=None):
''' Two side-by-side piecharts displaying patients under and over 50
and the percentage of each variant class is present in each group.
'''
# remove non validated PLP variants and NaN values in Age Group
df = df[~((df['validation'] == 0) &
(df['New Category'] == 'Pathogenic/Likely Pathogenic'))]
df = df.dropna(subset=['Age Group', 'New Category'])
# create percentage lists for piechart
table = df.groupby(['Age Group', 'New Category']).size()
percentage_table = table.groupby(
level=0).apply(lambda x: x / float(x.sum())).unstack().T
young_percent = percentage_table['Under 50'].tolist(
)[::-1] # [PLP, VUS, Benign]
old_percent = percentage_table['Over 50'].tolist()[::-1]
# create values for xtick labels
total_counts = df['Age Group'].value_counts().to_dict()
young_xlabel = "Under 50\nn = {}".format(total_counts.get('Under 50'))
old_xlabel = "Over 50\nn = {}".format(total_counts.get('Over 50'))
# sharing an axis allows the use of pm.line_between_plots
fig, ax = plt.subplots(figsize=(15, 12))
mpl.rcParams['font.size'] = 12 # cannot find an ax method for this
ax.pie(young_percent,
labels=[
'Likely Benign / No Variant', 'VUS',
'Pathogenic/Likely Pathogenic'
],
colors=['grey', 'silver', 'white'],
shadow=False,
autopct='%1.1f%%',
center=(0, 0),
startangle=90,
labeldistance=1.2,
pctdistance=0.7)
ax.pie(old_percent,
labels=[
'Likely Benign / No Variant', 'VUS',
'Pathogenic/Likely Pathogenic'
],
colors=['grey', 'silver', 'white'],
shadow=False,
autopct='%1.1f%%',
center=(2.5, 0),
startangle=90,
labeldistance=1.2,
pctdistance=0.7)
# axis modifications
ax.axis('equal')
ax.set_xticks([0, 2.5])
ax.set_xticklabels([young_xlabel, old_xlabel])
ax.tick_params(labelsize=15)
# STATS
# As I am an analysing a large number of samples, the chi-square test is approriate
# and will yield a simliar result as the Fisher-Freeman-Halton test
contingency_table = table.unstack().values.tolist()
chi2, pvalue, dof, expected = stats.chi2_contingency(contingency_table)
pm.line_between_plots(axs=ax,
x1=0,
x2=2.5,
height=1.5,
fontsize=15,
extend=0.2,
string="p = {}".format(pm.RoundToSigFigs(pvalue, 2)))
if outfile:
ax.figure.savefig(outfile)
return ax