def add_stats(df, variable1, variable2, ax, order):
stat = dunn_posthoc_test(df, y_variable, x_variable)
# label box pairs
box_pairs = [
(variable1, variable2),
(variable1, "control"),
(variable2, "control"),
]
# make empty list of p_values
p_values = []
# populate the list of p_values according to the box_pairs
for pair in box_pairs:
print(pair)
# select p value for each pair
p = stat.loc[pair[0], pair[1]]
p_values.append(p)
# add stats annotation to the plot
add_stat_annotation(
ax,
# plot=plot_type,
data=df,
x=x,
y=y,
order=order,
box_pairs=box_pairs,
text_format="star",
loc="outside",
verbose=2,
perform_stat_test=False,
pvalues=p_values,
test_short_name="Dunn",
)
def squat_plots(df_in, var_in, new_ylabel='Same'):
boxpairs = [
(('Avg', 'b'), ('Avg', 'p')),
(('Max', 'b'), ('Max', 'p')),
# (('Min', 'b'), ('Min', 'p')),
(('25%', 'b'), ('25%', 'p')),
(('75%', 'b'), ('75%', 'p')),
# (('Std', 'b'), ('Std', 'p')),
]
plot_out = sns.pointplot(x='Stat',
y=var_in,
hue='Mode',
join=False,
dodge=0.25,
data=df_in,
split=True,
palette="dark",
ci='sd')
add_stat_annotation(plot_out,
data=df_in,
x='Stat',
y=var_in,
hue='Mode',
box_pairs=boxpairs,
test='Mann-Whitney',
comparisons_correction=None,
text_format='star',
loc='inside',
verbose=1,
pvalue_thresholds=[[1e-4, "**"], [1e-3, "**"],
[1e-2, "**"], [0.05, "*"], [1, ""]])
plot_out.set_ylabel(new_ylabel)
plot_paper_params()
return plot_out
def foot_plots(df_in, var_in):
boxpairs = [(('Ball-Ball Dist.', 'b'), ('Ball-Ball Dist.', 'p')),
(('Heel-Heel Dist.', 'b'), ('Heel-Heel Dist.', 'p')),
(('Load Line L', 'b'), ('Load Line L', 'p')),
(('Load Line R', 'b'), ('Load Line R', 'p'))]
plot_out = sns.pointplot(x=' ',
y=var_in,
hue='mode',
join=False,
dodge=0.25,
data=df_in,
split=True,
palette="dark",
ci='sd')
add_stat_annotation(plot_out,
data=df_in,
x=' ',
y=var_in,
hue='mode',
box_pairs=boxpairs,
test='Mann-Whitney',
comparisons_correction=None,
text_format='star',
loc='inside',
verbose=1,
pvalue_thresholds=[[1e-4, "**"], [1e-3, "**"],
[1e-2, "**"], [0.05, "*"], [1, ""]])
plot_paper_params()
plt.xticks(rotation=90)
def draw_violin(input_file, output_file, watch):
data = pandas.read_csv(input_file)
seaborn.set(context="poster", style="whitegrid")
fig, ax = matplotlib.pyplot.subplots(figsize=(24, 24))
seaborn.violinplot(data=data,
x="Classification",
y=watch,
order=general.classes)
statannot.add_stat_annotation(ax,
data=data,
x="Classification",
y=watch,
box_pairs=[
(general.classes[i - 1],
general.classes[i])
for i in range(1, len(general.classes))
],
test="t-test_ind",
text_format="star",
verbose=0,
order=general.classes)
fig.savefig(general.check_exist(output_file))
matplotlib.pyplot.close(fig)
def en_trans_cli_plot(gene):
en_clinical_and_proteomics = en.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "transcriptomics",
metadata_cols = "Proteomics_Tumor_Normal",
omics_genes = gene)
en_clinical_and_proteomics.head()
## Show possible variations of Histologic_type
en_clinical_and_proteomics["Proteomics_Tumor_Normal"].unique()
sns.set(style ="white",
font_scale = 1.5)
ax = sns.boxplot(x = "Proteomics_Tumor_Normal",
y = gene + '_transcriptomics',
data = en_clinical_and_proteomics,
showfliers = False)
sns.stripplot(x = "Proteomics_Tumor_Normal",
y = gene + '_transcriptomics',
data = en_clinical_and_proteomics,
color = '.3')
add_stat_annotation(ax,
data = en_clinical_and_proteomics,
x = "Proteomics_Tumor_Normal",
y = gene + '_transcriptomics',
boxPairList = [("Tumor", "Adjacent_normal")],
test = 't-test_ind',
textFormat = 'star',
loc = 'inside',
verbose = 2)
plt.title('endometrial cancer')
def test_sig_boxplotAccuracy(self):
# print(self.acc.columns)
fig, ax3 = plt.subplots(figsize=(9, 7))
sns.set_style("whitegrid")
ax3 = sns.boxplot(data=self.acc, color="white")
ax3 = sns.swarmplot(data=self.acc, color=".25")
# plt.ylabel("single ELMs Accuracy")
# plt.ylabel("Boosting ELM Accuracy")
# plt.ylabel("Bagging ELM Accuracy")
# plt.ylabel("Majority ELM Accuracy")
plt.ylabel("Accuracy Comparison")
# statistical notation
add_stat_annotation(ax3,
data=self.acc,
box_pairs=[("single ELMs", "majority voting ELMs"),
("single ELMs", "bagging-based ELMs"),
("single ELMs", "boosting-based ELMs"),
("majority voting ELMs",
"boosting-based ELMs"),
("bagging-based ELMs",
"boosting-based ELMs"),
("boosting-based ELMs", "SVMs")],
test='t-test_ind',
text_format='star',
loc='inside',
verbose=2)
# plt.show(ax3)
# fig.savefig('fig2SingleELMAccuracyStats.png')
# fig.savefig('fig3BaggingELMAccuracyStats.png')
# fig.savefig('fig4BoostingELMAccuracyStats.png')
# fig.savefig('fig5MajorityELMAccuracyStats.png')
fig.savefig('fig6comparisonELMSVMAccuracyStats.png')
def plot_clinical_status_vs_esm_params(res, output_dir):
plt.figure(figsize=(10, 13))
nrows = 2
ncols = 2
params = ["BETAS_est", "DELTAS_est", "BDR_log", "PUP_ROI_AB_Mean"]
face_pal = {"No": "cornflowerblue", "Yes": "indianred"}
titles = [
"Production Rate", "Clearance Rate", "Prod/Clear Ratio (Log)",
"Amyloid Beta"
]
for i, param in enumerate(params):
j = i + 1
plt.subplot(nrows, ncols, j)
g = sns.boxplot(x="Symptomatic",
y=param,
data=res[res.AB_Positive == "Yes"],
palette=face_pal)
add_stat_annotation(g,
data=res[res.AB_Positive == "Yes"],
x="Symptomatic",
y=param,
box_pairs=[("Yes", "No")],
test='t-test_ind',
text_format='star',
loc='inside',
verbose=2,
fontsize=18)
plt.ylabel("")
plt.title(titles[i], fontsize=22)
plt.xticks(fontsize=18)
plt.yticks(fontsize=18)
plt.xlabel("Symptomatic", fontsize=18)
plt.tight_layout()
plt.savefig(os.path.join(output_dir, "clinical_status_vs_esm_params.png"))
def plot_swap_acc(ax, path, parameter_configs):
global data
data = _prepare_data_swapacc(path, parameter_configs)
data.sort_values(by="Agent", inplace=True)
return
sns.barplot(
data=data,
x="Agent",
y="Swap acc.",
ax=ax,
edgecolor=".2",
capsize=0.01,
errwidth=1.5,
)
add_stat_annotation(
ax,
data=data,
x="Agent",
y="Swap acc.",
test="t-test_welch",
line_height=0.02,
line_offset_to_box=0.04,
box_pairs=[("DTI", "AE+MTM")],
)
def cat_plot(type, x, y, file_name, hue=None):
df_cat = pd.read_excel(
r'C:\Users\chaob\Documents\Biopsy Heterogeneity Data Sheet.xlsx',
sheet_name='Violin Plot')
fig, ax = plt.subplots(figsize=(5, 6))
if type == 'violin':
sns.violinplot(x=x, y=y, hue=hue, palette='pastel', data=df_cat, ax=ax)
fig.tight_layout()
fig.savefig(file_name, dpi=200)
if type == 'box':
sns.boxplot(x=x,
y=y,
hue=hue,
saturation=0.5,
showfliers=False,
palette='pastel',
data=df_cat,
ax=ax)
ax.xaxis.labelpad = 15
ax.yaxis.labelpad = 5
ax.set_xlabel(ax.get_xlabel(), fontsize=13)
ax.set_ylabel(ax.get_ylabel(), fontsize=13)
ax.tick_params(axis='both', which='major', labelsize=10.5)
xlabels = [l.get_text() for l in ax.get_xticklabels()]
if hue is None:
box_pairs = list(itertools.combinations(xlabels, 2))
sns.stripplot(x=x,
y=y,
hue=hue,
s=3,
data=df_cat,
alpha=0.6,
palette='tab10',
ax=ax)
else:
huelabels = df_cat[hue].unique().tolist()
hue_pairs = list(itertools.combinations(huelabels, 2))
box_pairs = []
for xlabel in xlabels:
for hue1, hue2 in hue_pairs:
pair = ((xlabel, hue1), (xlabel, hue2))
box_pairs.append(pair)
add_stat_annotation(ax,
data=df_cat,
x=x,
y=y,
hue=hue,
box_pairs=box_pairs,
perform_stat_test=True,
test='t-test_welch',
loc='inside',
verbose=0,
no_ns=True,
fontsize='large')
fig.tight_layout()
fig.savefig(file_name, dpi=200)
def main(dname, out_dir):
# prepare
out_dir.mkdir(parents=True, exist_ok=True)
df = pd.read_csv(dname / 'measures.csv')
# print statistics
print(df.groupby('method').count())
print(df.groupby(['method'])['roc_auc'].median())
print(df.groupby(['method'])['roc_auc'].std())
# aggregated plot
fig, ax = plt.subplots(figsize=(8, 6))
sns.boxplot(data=df, x='method', y='roc_auc', order=['dce', 'cor', 'pcor'])
for patch in ax.artists:
r, g, b, a = patch.get_facecolor()
patch.set_facecolor((r, g, b, 0.3))
sns.stripplot(data=df, x='method', y='roc_auc', order=['dce', 'cor', 'pcor'])
statannot.add_stat_annotation(
ax,
data=df,
x='method',
y='roc_auc',
order=['dce', 'cor', 'pcor'],
box_pairs=[('dce', 'cor'), ('dce', 'pcor')],
test='Wilcoxon',
text_format='simple',
loc='outside',
verbose=2,
)
ax.set_xlabel('Method')
ax.set_ylabel('ROC-AUC')
fig.tight_layout()
fig.savefig(out_dir / 'method_comparison.pdf')
# stratified plot
g = sns.catplot(
data=df,
x='method',
y='roc_auc',
hue='perturbed_gene',
row='treatment',
kind='box',
hue_order=natsorted(df['perturbed_gene'].unique()),
aspect=2,
)
g.map(
lambda **kwargs: plt.axhline(0.5, ls='dashed', color='gray', alpha=1, zorder=-1)
)
g.set_axis_labels('Method', 'ROC-AUC')
g._legend.set_title('Perturbed gene(s)')
g.savefig(out_dir / 'method_comparison_stratified.pdf')
def plot_pubtator_clean():
sysid_primary = pd.read_csv(ROOT_DIR + "sysid/sysid_primary.csv",
usecols=["Entrez id"])["Entrez id"].to_list()
sysid_candidates = pd.read_csv(ROOT_DIR + "sysid/sysid_candidates.csv",
usecols=["Entrez id"
])["Entrez id"].to_list()
princeton_negative = pd.read_csv(
ROOT_DIR + "ASD_translated_to_ensembl.csv")["gene id"].to_list()
pubtator = pd.read_csv(
ROOT_DIR +
"pubtator_central/gene_scores/gene_scores_p_cutoff_0,0001_clean.csv")
pubtator["sys_primary"] = pubtator.gene_id.isin(sysid_primary).astype(int)
pubtator["sys_candidate"] = pubtator.gene_id.isin(sysid_candidates).astype(
int)
pubtator["sys"] = pubtator.gene_id.isin(sysid_primary +
sysid_candidates).astype(int)
pubtator["sys_category"] = "unknown"
pubtator.loc[pubtator.sys_candidate == 1, "sys_category"] = "candidate"
pubtator.loc[pubtator.sys_primary == 1, "sys_category"] = "known NDD"
pubtator.loc[pubtator.gene_id.isin(princeton_negative),
"sys_category"] = "negative control"
order = ["unknown", "negative control", "candidate", "known NDD"]
ax = plt.figure(figsize=(6, 6))
ax = sns.boxplot(x="sys_category",
y="gene_score",
data=pubtator,
showfliers=False,
order=order)
add_stat_annotation(ax,
data=pubtator,
x="sys_category",
y="gene_score",
order=order,
box_pairs=[("unknown", "candidate"),
("candidate", "negative control"),
("candidate", "known NDD")],
test='Mann-Whitney',
text_format='simple',
loc='outside',
line_offset_to_box=0.001,
line_height=0.05,
text_offset=2,
verbose=2)
ax.set(ylim=(0, 1800))
# ax.set(ylim=(0, 0.07))
ax.set_title(f"pubtator gene scores")
ax.set_xlabel("SysID category")
ax.set_ylabel(f"gene score")
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.get_figure().savefig(ROOT_DIR +
f"pubtator_central/plot_pubtator_clean.png")
plt.show()
def box_zoom(factors, sel, hue, b, color_df, palette=None, stat=False):
"""Display boxplot for a given TCA-component associated trial factor
Arguments:
factors {list} -- list of 3 arrays containing the TCA factors
sel {scalar} -- component selected
hue {string} -- entry in the color_df to color code trial factors
b {tuple} -- color list coded for learning, list of each block boundaries,
list of each first day trial
color_df {pandas dataframe} -- columns [Odor, Reward, Day, Behavior] with color coded
Keyword Arguments:
palette {list} -- color palette for plotting (default: {['red', 'black']})
stat {bool} -- add significance test stars (default: {False})
"""
box_df = make_box_df(factors, sel, b, color_df)
n_blocks = len(b[1])
if palette is None:
palette = ['red', 'black']
plt.rcParams['figure.figsize'] = 14, 6
#fig = plt.figure(figsize=(14, 6))
ax = sns.boxplot(x="Block",
y="Trial Factor",
hue=hue,
data=box_df,
palette=palette,
dodge=False,
linewidth=2,
fliersize=2,
width=.3)
xmin, xmax, ymin, ymax = ax.axis()
if stat:
c1, c2 = box_df[hue].unique()[:2]
for i in range(n_blocks):
if len(box_df[hue][box_df['Block'] == i].unique()) < 2:
continue
sta.add_stat_annotation(ax,
data=box_df,
x="Block",
y="Trial Factor",
hue=hue,
boxPairList=[((i, c1), (i, c2))],
test='t-test',
textFormat='star',
loc='inside',
fontsize='large',
lineYOffsetAxesCoord=0.05,
linewidth=0,
verbose=0)
ax.set_xlim(xmin, xmax)
ax.set_ylim(ymin, ymax * 1.15)
plt.show()
def violin_plot(data, colname, ymax, fname):
""" Plot a violin plot with the length of each read by novelty category"""
sns.set_context("paper", font_scale=1.3)
ax = sns.stripplot(x='DE_type',
y=colname,
data=data,
color="black",
alpha=0.5,
size=1.5,
jitter=True)
ax = sns.boxplot(x='DE_type', y=colname, data=data, palette="Blues")
add_stat_annotation(ax,
data=data,
x='DE_type',
y=colname,
box_pairs=[("Higher in Illumina", "Higher in PacBio"),
("Higher in Illumina", "Not DE"),
("Higher in PacBio", "Not DE")],
test='Mann-Whitney',
text_format='star',
loc='outside',
verbose=2)
#ax = sns.violinplot(x='DE_type', y=colname, legend = False,
# data=data,
# #order=cat_order,
# linewidth = 1,
# inner = 'box', cut = 0)
# Calculate number of obs per group & position labels
nobs = list(data.groupby("DE_type").size())
nobs = [str(x) for x in nobs]
nobs = ["n=" + i for i in nobs]
# Add it to the plot
ypos = data.groupby(['DE_type'])[colname].max().dropna().values
pos = range(len(nobs))
for tick, label in zip(pos, ax.get_xticklabels()):
ax.text(pos[tick],
ypos[tick] + ypos[tick] * 0.1,
nobs[tick],
horizontalalignment='center',
size='x-small',
color='black',
weight='semibold')
ax.legend().set_visible(False)
plt.xlabel("")
plt.ylabel("GC percentage of gene")
#ymin = min(data.groupby(['transcript_novelty'])['read_length'].min().values)
plt.ylim(0, 100)
plt.tight_layout()
plt.savefig(fname, dpi=600, bbox_inches='tight')
plt.close()
def plot(self):
x, y, hue, order, hue_order, box_pairs = self.x, self.y, self.hue, self.order, self.hue_order, self.box_pairs
if not isinstance(self.x, str) and not isinstance(self.y, str):
vis_df = pd.DataFrame()
vis_df['x'] = x
vis_df['y'] = y
x = 'x'
y = 'y'
if self.hue is not None:
vis_df['colour'] = self.hue
hue = 'colour'
if order is None:
order = list(set(vis_df['x'].values))
order.sort()
else:
vis_df = self.df
# set the orders
if hue_order is None and hue is not None:
hue_order = list(set(vis_df[hue].values))
hue_order.sort()
if order is None:
order = list(set(vis_df[x].values))
order.sort()
ax = sns.violinplot(data=vis_df, x=x, y=y, hue=hue, hue_order=hue_order, order=order, palette=self.palette,
showfliers=self.showfliers)
if self.add_dots:
ax = sns.stripplot(data=vis_df, x=x, y=y, hue_order=hue_order, order=order, alpha=0.9, s=1, color='.2')
if self.add_stats:
# Add all pairs in the order if the box pairs is none
pairs = []
if box_pairs is None:
box_pairs = []
for i in order:
for j in order:
if i != j:
# Ensure we don't get duplicates
pair = f'{i}{j}' if i < j else f'{j}{i}'
if pair not in pairs:
box_pairs.append((i, j))
pairs.append(pair)
# Add stats annotation
add_stat_annotation(ax, data=vis_df, x=x, y=y, order=order,
box_pairs=box_pairs,
test=self.stat_method, text_format='star', loc='inside', verbose=2,
pvalue_thresholds=[[1e-4, "****"], [1e-3, "***"], [1e-2, "**"], [0.05, "*"]])
ax.set_xticklabels(ax.get_xticklabels(), rotation=45, horizontalalignment='right')
ax.tick_params(labelsize=self.label_font_size)
plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0., fontsize=self.label_font_size)
self.add_labels()
self.set_ax_params(ax)
plt.tight_layout()
return ax
def plot_perspective(ax, path_persp, path_no_persp, parameter_configs_p,
parameter_configs_nop):
if path_persp == "results/gridsweep":
print("Getting data from gridsweep")
data_perp = _prepare_data_perspective_grid()
data_no_perp = _prepare_data_noperspective_grid()
else:
print("Getting data NOT from gridsweep")
data_perp = _prepare_data_perspective(path_persp, parameter_configs_p)
data_no_perp = _prepare_data_perspective(path_no_persp,
parameter_configs_nop)
ax.set_ylim((0.5, 1))
data_no_perp["Perspective"] = "No"
data_no_perp.loc[data_no_perp["Agent"] == "AE+MTM", "Value"] += 0.002
data_perp["Perspective"] = "Yes"
data = pd.concat([data_perp, data_no_perp])
data.sort_values(by="Agent", inplace=True)
print(data)
sns.barplot(
data=data,
x="Perspective",
y="Value",
hue="Agent",
edgecolor=".2",
capsize=0.01,
errwidth=1.5,
ax=ax,
)
remove_legend_titles(ax)
add_stat_annotation(
ax,
line_height=0.02,
line_offset_to_box=0.04,
data=data,
x="Perspective",
y="Value",
hue="Agent",
test="t-test_welch",
box_pairs=[
(("Yes", "AE"), ("Yes", "DTI")),
(("Yes", "AE"), ("Yes", "AE+MTM")),
(("No", "AE"), ("No", "DTI")),
(("No", "AE"), ("No", "AE+MTM")),
],
)
change_width_(ax, 0.22)
ax.set_ylabel(r"Accuracy (\%)")
def col_pho_cliplot(gene):
col_clinical_and_proteomics = col.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "phosphoproteomics",
metadata_cols = "Stage")
col_clinical_and_proteomics["Stage"] = col_clinical_and_proteomics["Stage"].fillna("Normal")
col_clinical_and_proteomics.head()
## Show possible variations of Histologic_type
col_clinical_and_proteomics["Stage"].unique()
PhosphoSite = list(col_clinical_and_proteomics.filter(like = gene).columns.values.tolist())
for i in PhosphoSite:
try:
print(i)
col_clinical_and_proteomics = col.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "phosphoproteomics",
metadata_cols = "Stage")
col_clinical_and_proteomics["Stage"] = col_clinical_and_proteomics["Stage"].fillna("Normal")
col_clinical_and_proteomics = col_clinical_and_proteomics.dropna(subset = [i])
plt.figure()
sns.set(style ="white",
font_scale = 1.5)
order = ["Normal",
"Stage I",
"Stage II",
"Stage III",
"Stage IV"]
ax = sns.boxplot(x = "Stage",
y = i,
data = col_clinical_and_proteomics,
showfliers = False,
order = order)
sns.stripplot(x = "Stage",
y = i,
data = col_clinical_and_proteomics,
color = '.3',
order = order)
add_stat_annotation(ax,
data = col_clinical_and_proteomics,
x = "Stage",
y = i,
order = order,
boxPairList = [("Normal", "Stage I"),
("Normal", "Stage II"),
("Normal", "Stage III"),
("Normal", "Stage IV")],
test = 't-test_ind',
textFormat = 'star',
loc = 'inside',
verbose = 2)
plt.title('colon cancer')
except: ValueError
pass
def annotate_anova(ax, data, y, anova_path, anova_sheet):
df = pd.read_excel(anova_path, sheet_name=anova_sheet, index_col=0)
df = df[y.split(' ')[0]]
pvalues = []
box_pairs = []
for x in df.index:
p = df[x]
if p < 0.05:
pvalues.append(p)
box_pairs.append(((x, 'RU'), (x, 'LL')))
add_stat_annotation(ax, data=data, x='EVLP ID', y=y, hue='Location',
box_pairs=box_pairs, pvalues=pvalues, perform_stat_test=False,
loc='outside', verbose=0)
def ovcliplot(gene):
ov_clinical_and_proteomics = ov.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "phosphoproteomics",
# metadata_cols = "Tumor_Stage_Ovary_FIGO",
omics_genes = gene)
ov_clinical_and_proteomics["Tumor_Stage_Ovary_FIGO"] = ov_clinical_and_proteomics["Tumor_Stage_Ovary_FIGO"].fillna("Normal")
ov_clinical_and_proteomics.head()
## Show possible variations of Histologic_type
ov_clinical_and_proteomics["Tumor_Stage_Ovary_FIGO"].unique()
PhosphoSite = list(ov_clinical_and_proteomics.filter(like = gene).columns.values.tolist())
for i in PhosphoSite:
print(i)
ov_clinical_and_proteomics = ov.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "phosphoproteomics",
# metadata_cols = "Tumor_Stage_Ovary_FIGO",
omics_genes = gene)
ov_clinical_and_proteomics["Tumor_Stage_Ovary_FIGO"] = ov_clinical_and_proteomics["Tumor_Stage_Ovary_FIGO"].fillna("Normal")
# ov_clinical_and_proteomics = ov_clinical_and_proteomics.dropna(subset = [i])
plt.figure()
sns.set_style("white")
order = ["Normal",
"IIIA",
"IIIB",
"IIIC",
"IV"]
ax = sns.boxplot(x = "Tumor_Stage_Ovary_FIGO",
y = i,
data = ov_clinical_and_proteomics,
showfliers = False,
order = order)
sns.stripplot(x = "Tumor_Stage_Ovary_FIGO",
y = i,
data = ov_clinical_and_proteomics,
color = '.3',
order = order)
add_stat_annotation(ax,
data = ov_clinical_and_proteomics,
x = "Tumor_Stage_Ovary_FIGO",
y = i,
order = order,
boxPairList = [("Normal", "IIIA"),
("Normal", "IIIB"),
("Normal", "IIIC"),
("Normal", "IV")],
test = 't-test_ind',
textFormat = 'star',
loc = 'inside',
verbose = 2)
plt.title('ovarian cancer')
def add_annotation(ax, results_df, all_pairs, metric, box_pairs):
"""Add annotation for pairwise statistical tests to box plots."""
import itertools as it
from statannot import add_stat_annotation
# do rank-based tests for all pairs, with Bonferroni correction
pairwise_tests_df = _pairwise_compare(results_df, all_pairs, metric)
# specify statistical tests to plot
box_pvals = (pairwise_tests_df.set_index(
['data_type_1', 'data_type_2']).loc[box_pairs, :]).corr_pval.values
# only display nearby pairs
_ = add_stat_annotation(ax,
data=results_df.sort_values(by='gene'),
x='training_data',
y='delta_mean',
order=all_pairs,
box_pairs=box_pairs,
perform_stat_test=False,
pvalues=box_pvals,
pvalue_thresholds=[(1e-3, '***'), (1e-2, '**'),
(0.05, '*'), (1, 'ns')],
text_format='star',
loc='inside',
verbose=0,
fontsize=16)
return pairwise_tests_df
def sns_violinplot(dd,
my_pal,
figname,
plot_xlabels,
x="Gene",
y='value',
hue=None,
no_legend=True,
rotation=0,
annot=False):
fig = plt.figure(figsize=(10, 10))
ax = sns.violinplot(x=x, y=y, data=dd, hue=hue, palette=my_pal)
gene_list = constants.analysis_config['MRNA_GENES']
if annot:
box_pairs = []
for i in range(1, len(gene_list) + 1):
if i % 2 == 0:
box_pairs.append(((gene_list[i - 2], gene_list[i - 1])))
add_stat_annotation(ax,
data=dd,
x=x,
y=y,
hue=hue,
box_pairs=box_pairs,
test='t-test_ind',
text_format='star',
loc='inside',
verbose=2)
ax.set_xlabel("")
ax.set_ylabel("")
ax.yaxis.grid(which="major", color='black', linestyle='-', linewidth=0.25)
ax.tick_params(right=False,
top=False,
direction='out',
length=8,
width=3,
colors='black')
ax.spines['left'].set_linewidth(3)
ax.set_xticklabels(plot_xlabels, rotation=rotation)
plt.yticks(fontsize=30)
plt.xticks(fontsize=20)
plt.gcf().subplots_adjust(bottom=0.2, left=0.2)
if no_legend:
ax.legend_.remove()
fig.savefig(figname, format='png', dpi=600)
plt.close()
def plot_param_diff_acc_status(esm_res, output_dir):
sns.set_style("whitegrid", {'axes.grid': False})
yaxis_labels = [
"Deltas (Clearance Parameter)", "Betas (Production Parameter)",
"Beta Delta Ratio (Log)"
]
plt.figure(figsize=(19, 7))
nrows = 1
ncols = 3
titles = ["Clearance", "Production", "Production/Clearance"]
for i, y in enumerate(["DELTAS_est", "BETAS_est", "BDR_log"]):
j = i + 1
plt.subplot(nrows, ncols, j)
yaxis_label = yaxis_labels[i]
pal = {"No": "mediumblue", "Yes": "red"}
face_pal = {"No": "cornflowerblue", "Yes": "indianred"}
y = y
x = "Accumulator"
data = esm_res[esm_res.Mutation == 1]
g = sns.boxplot(data=data, x=x, y=y, palette=face_pal, fliersize=0)
sns.stripplot(x=x,
y=y,
data=data,
jitter=True,
dodge=True,
linewidth=0.5,
palette=pal)
g.set_xticklabels(["Non-accumulator", "Accumulator"], fontsize=24)
add_stat_annotation(g,
data=data,
x=x,
y=y,
box_pairs=[("No", "Yes")],
test='t-test_ind',
text_format='star',
loc='inside',
verbose=2,
fontsize=18)
plt.xlabel("", fontsize=24)
plt.ylabel("", fontsize=18)
plt.title(titles[i], fontsize=24)
plt.rc('xtick', labelsize=24)
plt.rc('ytick', labelsize=24)
plt.tight_layout()
plt.savefig(os.path.join(output_dir, "param_diff_acc_status.png"))
plt.close()
def ov_pho_cli_plot(gene):
ov_clinical_and_proteomics = ov.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "phosphoproteomics",
metadata_cols = "Sample_Tumor_Normal",
omics_genes = gene)
ov_clinical_and_proteomics.head()
ov_clinical_and_proteomics = ov_clinical_and_proteomics.loc[:, ~ov_clinical_and_proteomics.columns.duplicated()]
## Show possible variations of Histologic_type
ov_clinical_and_proteomics["Sample_Tumor_Normal"].unique()
Genes = list(ov_clinical_and_proteomics.filter(like = gene).columns.values.tolist())
for i in Genes:
print(i)
ov_clinical_and_proteomics = ov.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "phosphoproteomics",
metadata_cols = "Sample_Tumor_Normal",
omics_genes = gene)
ov_clinical_and_proteomics = ov_clinical_and_proteomics.loc[:, ~ov_clinical_and_proteomics.columns.duplicated()]
ov_clinical_and_proteomics = ov_clinical_and_proteomics.dropna(subset = [i])
plt.figure()
sns.set_style("white")
order = ["Normal", "Tumor"]
ax = sns.boxplot(x = "Sample_Tumor_Normal",
y = i,
data = ov_clinical_and_proteomics,
showfliers = False,
order = order)
sns.stripplot(x = "Sample_Tumor_Normal",
y = i,
data = ov_clinical_and_proteomics,
color = '.3',
order = order)
add_stat_annotation(ax,
data = ov_clinical_and_proteomics,
x = "Sample_Tumor_Normal",
y = i,
order = order,
boxPairList = [("Normal", "Tumor")],
test = 't-test_ind',
textFormat = 'star',
loc = 'inside',
verbose = 2)
plt.title('ovarian cancer')
def analyze_whole_chr_anp_frac_samplewise_all(wgd_plus_fracs, wgd_minus_fracs):
wgd_status = ["WGD+"] * len(wgd_plus_fracs.index) + ["WGD-"] * len(
wgd_minus_fracs.index)
print(wgd_plus_fracs.head())
print(wgd_minus_fracs.head())
df_all = pd.concat([wgd_plus_fracs, wgd_minus_fracs],
axis=0,
ignore_index=True)
# df_all = pd.concat([df_all, pd.Series(wgd_status)], axis = 1)
df_all["wgd_status"] = wgd_status
print(df_all.head())
df_all.to_csv(os.path.join(src_folder, "Type_wise_df_wca_frac.tsv"),
sep="\t",
index=0)
plt.figure(figsize=(30, 10))
ax = sns.boxplot(
x="wgd_status",
y="wca_frac",
data=df_all,
hue="wgd_status",
palette="Set1",
showfliers=False,
)
ax.set_ylim((0, 1.1))
ax.set_xticklabels(ax.get_xticklabels(), rotation=90)
pairs = [("WGD+", "WGD-")]
add_stat_annotation(
ax,
x="wgd_status",
y="wca_frac",
data=df_all,
box_pairs=pairs,
test="t-test_ind",
text_format="star",
loc="inside",
verbose=0,
)
plt.show()
def add_annot(data, gene_list, ax, test):
dd = pd.DataFrame(dict([
(k, pd.Series(v)) for k, v in data.items()
])).melt().dropna().rename(columns={"variable": "gene"})
box_pairs = []
for i in range(1, len(gene_list) + 1):
if i % 2 == 0:
box_pairs.append(tuple((gene_list[i - 2], gene_list[i - 1])))
# test value should be one of the following:
add_stat_annotation(ax,
data=dd,
x='gene',
y='value',
hue=None,
box_pairs=box_pairs,
test=test,
text_format='star',
loc='inside',
verbose=2)
def barplots(alphas, betas, thetas):
sns.set(style='white', font_scale=2)
p_values = []
list = [thetas, alphas, betas]
for l, frequency in enumerate(['theta', 'alpha', 'beta']):
p_values.append(
stats.ttest_ind(list[l][0], list[l][1], equal_var=False)[1])
p_values = np.array(p_values) * len(p_values)
dataframe = pd.DataFrame()
dataframe['States'] = ['slow'] * len(alphas[0]) + ['fast'] * len(alphas[1])
dataframe['Alpha'] = alphas[0] + alphas[1]
dataframe['Theta'] = thetas[0] + thetas[1]
dataframe['Beta'] = betas[0] + betas[1]
fig, ax = plt.subplots(1, 3, figsize=(17, 20), squeeze=False)
place = {0: (0, 0), 1: (0, 1), 2: (0, 2)}
for l, frequency in enumerate(['Theta', 'Alpha', 'Beta']):
sns.barplot(ax=ax[place[l]],
y=frequency,
x='States',
data=dataframe,
capsize=0.1).set(xlabel='', ylabel=frequency)
add_stat_annotation(ax[place[l]],
y=frequency,
x='States',
data=dataframe,
box_pairs=[('slow', 'fast')],
perform_stat_test=False,
pvalues=[p_values[l]],
text_format='star',
loc='outside',
verbose=2,
comparisons_correction=None,
line_offset=0.02,
text_offset=0.01)
fig.subplots_adjust(wspace=0.38, hspace=0.62)
fig.show()
return p_values
def en_phos_cli_plot(gene):
en_clinical_and_proteomics = en.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "phosphoproteomics_gene",
metadata_cols = "tumor_Stage-Pathological",
omics_genes = gene)
en_clinical_and_proteomics["tumor_Stage-Pathological"] = en_clinical_and_proteomics["tumor_Stage-Pathological"].fillna("Normal")
en_clinical_and_proteomics.head()
## Show possible variations of Histologic_type
en_clinical_and_proteomics["tumor_Stage-Pathological"].unique()
PhosphoSite = list(en_clinical_and_proteomics.filter(like = gene).columns.values.tolist())
for i in PhosphoSite:
print(i)
en_clinical_and_proteomics = en_clinical_and_proteomics.dropna(subset = [i])
plt.figure()
sns.set(style ="white",
font_scale = 1.5)
order = ["Normal", "Stage I", "Stage II", "Stage III", "Stage IV"]
ax = sns.boxplot(x = "tumor_Stage-Pathological",
y = i,
data = en_clinical_and_proteomics,
showfliers = False,
order = order)
sns.stripplot(x = "tumor_Stage-Pathological",
y = i,
data = en_clinical_and_proteomics,
color = '.3',
order = order)
add_stat_annotation(ax,
data = en_clinical_and_proteomics,
x = "tumor_Stage-Pathological",
y = i,
order = order,
boxPairList = [("Normal", "Stage I"),
("Normal", "Stage II"),
("Normal", "Stage III"),
("Normal", "Stage IV")],
test = 't-test_ind',
textFormat = 'star',
loc = 'inside',
verbose = 2)
plt.title('endometrial cancer')
def plot_allele_frequency(df: pd.DataFrame,
fname: str,
af_col: str = "gnomADg_AF"):
"""
Plots allele frequencies for each class
:param pd.DataFrame df: Input df
:param str fname: Output basemame
:param str af_col: Column name that accounts
for allele frequencies. Default: `gnomAD_genomes`.
If column does not exist, analysis will be skipped.
"""
if af_col not in df.columns:
return
df['grouper'] = df['outcome'].astype(
str) + '\nN = ' + df['count_class'].astype(str)
order = sorted(list(df['grouper'].unique()))
ax = sns.boxplot(data=df, x="grouper", order=order, y=af_col)
try:
add_stat_annotation(ax,
data=df,
x="grouper",
y=af_col,
order=order,
box_pairs=[tuple(order)],
test='Mann-Whitney',
text_format='star',
loc='inside',
verbose=0,
pvalue_format_string='{:.4f}')
plt.xlabel("")
plt.ylabel("Allele frequency")
plt.tight_layout()
out = fname + '.pdf'
plt.savefig(out)
plt.close()
except ValueError:
plt.close()
pass
def boxes(self):
plt.clf()
tests_combinations = list(combinations(self.experiments, 2))
full_data_agreg = pd.read_csv(f'{self.dir}{self.anal}_full_data.csv')
full_data_agreg = full_data_agreg[full_data_agreg['experiment'].isin(
self.experiments)]
full_data_agreg = full_data_agreg[full_data_agreg['episode'] ==
full_data_agreg["episode"].max()]
print(full_data_agreg)
for idx_measure, measure in enumerate(self.measures):
sb.set()
sb.set_style("whitegrid")
plot = sb.boxplot(x='experiment',
y=measure,
data=full_data_agreg,
palette=self.clrs) # hue='Style',
#remove bonferroni correction?
if len(tests_combinations) > 0:
add_stat_annotation(
plot,
data=full_data_agreg,
x='experiment',
y=measure, # order=order,
box_pairs=tests_combinations,
test='Wilcoxon',
text_format='star',
loc='inside',
verbose=2)
plt.title(self.anal)
plot.get_figure().savefig(
f'{self.dir}{self.anal}_{measure}_box.png')
plt.clf()
def col_tra_cli_plot(gene):
col_clinical_and_proteomics = col.join_metadata_to_omics(
metadata_df_name = "clinical",
omics_df_name = "transcriptomics",
metadata_cols = "Stage",
omics_genes = gene)
col_clinical_and_proteomics["Stage"] = col_clinical_and_proteomics["Stage"].fillna("Normal")
col_clinical_and_proteomics.head()
## Show possible variations of Histologic_type
col_clinical_and_proteomics["Stage"].unique()
sns.set(style ="white",
font_scale = 1.5)
order = ["Normal",
"Stage I",
"Stage II",
"Stage III",
"Stage IV"]
ax = sns.boxplot(x = "Stage",
y = gene + '_transcriptomics',
data = col_clinical_and_proteomics,
showfliers = False,
order = order)
ax = sns.stripplot(x = "Stage",
y = gene + '_transcriptomics',
data = col_clinical_and_proteomics,
color = '.3',
order = order)
add_stat_annotation(ax,
data = col_clinical_and_proteomics,
x = "Stage",
y = gene + '_transcriptomics',
order = order,
boxPairList = [("Normal", "Stage I"),
("Normal", "Stage II"),
("Normal", "Stage III"),
("Normal", "Stage IV")],
test = 't-test_ind',
textFormat = 'star',
loc = 'inside',
verbose = 2)
plt.title('colon cancer')
def plot_agreement(ax, path):
data = _prepare_data_agreement(path)
data.sort_values(by="Agent", inplace=True)
sns.barplot(
data=data,
ax=ax,
x="Agent",
y="Agreement",
edgecolor=".2",
capsize=0.01,
errwidth=1.5,
)
add_stat_annotation(
ax,
data=data,
x="Agent",
y="Agreement",
test="t-test_welch",
line_height=0.02,
line_offset_to_box=0.04,
box_pairs=[("DTI", "AE+MTM")],
)
