def run():
dic_primary_full, _ = return_metadata()
signature_2_treatment_signature_analyzer = {
'CISPLATIN': ['21_SBS31_0.953955_1', '14_1'],
'CARBOPLATIN': ['21_SBS31_0.953955_1', '25_1'],
'5-FU_CAPE': ['31_SBS17b_0.968799_1'],
'OXALIPLATIN': ['14_1', '37_1']
}
signature_2_treatment_sigprofiler = {
'CISPLATIN': ['1_SBS31_0.968153_0.98'],
'CARBOPLATIN': ['1_SBS31_0.968153_0.98'],
'OXALIPLATIN': ['20_0.92'],
'5-FU_CAPE': ['19_SBS17b_0.961548_0.99'],
}
outpath = 'figures/'
path_base = 'data/hartwig/signatures/extraction/results/{}/snvs/exposures/'
path_siganalyzer = path_base.format(
"SignatureAnalyzer") + "Pan_full/Pan_full.exposures.tsv"
path_sigprofiler = path_base.format(
"SigProfiler"
) + "PanNoSkinNoUnknown.snvs/PanNoSkinNoUnknown.snvs.exposures.tsv"
for drug in signature_2_treatment_signature_analyzer:
process_exposures_methods(path_siganalyzer, path_sigprofiler, drug,
signature_2_treatment_signature_analyzer,
signature_2_treatment_sigprofiler, outpath)
def __init__(self,
exposures_path,
discriminate_path,
matrix_treatments_path,
ttype='Pan'):
self.exposures = pd.read_csv(exposures_path, sep='\t', index_col=0)
self.exposures = self.exposures.transpose()
self.discriminate = pd.read_csv(discriminate_path, sep='\t')
self.discriminate = self.discriminate[self.discriminate['ttype'] ==
ttype]
self.discriminate = self.discriminate.astype({
'signature': str,
'effect_size': float,
'pvals': float,
'treatment': str,
'ttype': str,
'total_treated': float
})
self.treatments = pd.read_csv(matrix_treatments_path,
sep='\t',
index_col=0)
self.ttype = ttype
dic_ttypes, _ = return_metadata()
if ttype == 'Pan':
self.samples = dic_ttypes.keys()
else:
self.samples = [k for k, v in dic_ttypes.items() if v == ttype]
self.effect_thresh = EFFECT_SIZE # min effect size
self.pvalue_thresh = PVALUE # empirical_pvalue threshold
self.overlap_thresh = OVERLAP # min overlap to consider cotreatments
def process_exposures_methods(path_siganalyzer, path_sigprofiler, drug,
signature_2_treatment_signature_analyzer,
signature_2_treatment_sigprofiler, outpath):
df_sigpro = read_exposures(path_sigprofiler)
df_sigan = read_exposures(path_siganalyzer)
all_treated_samples = set()
treated_samples = pickle.load(gzip.open(Conf['treatment_specific_drug']))
dic_primary_full, _ = return_metadata()
dsiganalyzer = defaultdict(lambda: defaultdict(float))
dsiganalyzer_simple = defaultdict(float)
dsiganalyzer_simple_exposure = defaultdict(float)
sigs_affected_signature_analyzer = signature_2_treatment_signature_analyzer[
drug]
for sample, d in df_sigan.iterrows():
if sample in treated_samples[drug]['Pan']['YES']:
dsiganalyzer[dic_primary_full[sample]][sample] = d.loc[
sigs_affected_signature_analyzer].sum()
dsiganalyzer_simple[sample] = d.loc[
sigs_affected_signature_analyzer].sum()
dsiganalyzer_simple_exposure[sample] = d.loc[
sigs_affected_signature_analyzer].sum() / d.sum()
dsigprof = defaultdict(lambda: defaultdict(float))
dsigprof_simple = defaultdict(float)
dsigprof_simple_exposure = defaultdict(float)
sigs_affected_sigprofiler = signature_2_treatment_sigprofiler[drug]
for sample, d in df_sigpro.iterrows():
if sample in treated_samples[drug]['Pan']['YES']:
all_treated_samples.add(sample)
dsigprof[dic_primary_full[sample]][sample] = d.loc[
sigs_affected_sigprofiler].sum()
dsigprof_simple[sample] = d.loc[sigs_affected_sigprofiler].sum()
dsigprof_simple_exposure[
sample] = d.loc[sigs_affected_sigprofiler].sum() / d.sum()
barplot_classes(dsigprof_simple_exposure, dsiganalyzer_simple_exposure,
drug, outpath, 'count')
plot_single_distribution(dsigprof_simple, dsiganalyzer_simple, drug,
outpath, 'count')
plot_single_correlation(dsigprof_simple, dsiganalyzer_simple, drug,
outpath, 'count')
plot_single_distribution(dsigprof_simple_exposure,
dsiganalyzer_simple_exposure, drug, outpath,
'exposure')
plot_single_correlation(dsigprof_simple_exposure,
dsiganalyzer_simple_exposure, drug, outpath,
'exposure')
bad_samples_drug = get_samples_similar_exposure(
dsigprof_simple_exposure, dsiganalyzer_simple_exposure, drug, outpath,
'exposure')
def merge_samples_timing(input_file, outpath):
os.makedirs(outpath, exist_ok=True)
dic_primary_full, _ = return_metadata()
all_hartwig = glob(input_file)
to_remove = [
'EXTENDED', 'MAJOR_CN', 'MINOR_CN', 'TOTAL_CN', 'NORMAL_CN',
'VAR_COUNTS', 'REF_COUNTS', 'GENDER', 'PURITY'
]
all_files = []
for f in tqdm(all_hartwig, total=len(all_hartwig)):
# remove those that do not belong to a good origin
sample_name = os.path.basename(f).split('_')[1].split('.')[0]
if sample_name in dic_primary_full:
df = pd.read_csv(f, sep='\t', low_memory=False)
df.drop(to_remove, axis=1, inplace=True)
all_files.append(df)
hart_df = pd.concat(all_files)
hart_df['PRIMARY'] = hart_df['SAMPLE'].map(dic_primary_full)
hart_df.sort_values(by=['CHROM', 'POS'], inplace=True)
outfile = '{}/Pan.snvs.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'SNV'].to_csv(outfile,
sep='\t',
header=True,
index=False,
compression='gzip')
outfile = '{}/Pan.indels.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'INDEL'].to_csv(outfile,
sep='\t',
header=True,
index=False,
compression='gzip')
outfile = '{}/Pan.dbs.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'DBS'].to_csv(outfile,
sep='\t',
header=True,
index=False,
compression='gzip')
# save ALL file
outfile = '{}/Pan.all.gz'.format(outpath)
hart_df.to_csv(outfile,
sep='\t',
header=True,
index=False,
compression='gzip')
def create_matrix_treatments_plot():
treated = pickle.load(gzip.open(Conf['treatment_FDA_drug']))
dic_primary, _ = return_metadata()
matrix_treated = defaultdict(lambda: defaultdict(int))
forbidden = ['CUP', 'Eye', 'Double-primary', 'Unknown']
for sample, t in dic_primary.items():
if t not in forbidden:
for k, d in treated.items():
if sample in d[t]['YES']:
matrix_treated[sample][k] = 1
elif sample in d[t]['NO']:
matrix_treated[sample][k] = 0
out = pd.DataFrame.from_dict(matrix_treated, orient='index')
out = out.dropna()
dic_t = defaultdict(list)
for i, row in out.iterrows():
dic_t[dic_primary[i]].append(i)
return out, dic_t
def get_items_to_plot(samples_set, expos, sig):
df = pd.read_csv(expos, sep='\t', index_col=0)
df = df.T
dic_primary_full, _ = return_metadata()
colors_dict = return_colors()
vals_to_plot = defaultdict(list)
colors = defaultdict(list)
count_exposed = defaultdict(int)
count_total = defaultdict(int)
for sample in samples_set:
ttype = dic_primary_full[sample]
if sample in df.index:
sample_exp = df.loc[sample][sig]
if sample_exp > 1:
count_exposed[ttype] += 1
vals_to_plot[ttype].append(sample_exp)
colors[ttype].append(colors_dict[ttype])
count_total[ttype] += 1
return vals_to_plot, colors, count_exposed, count_total
def plot_heatmap_treatment(file):
outf = os.path.basename(file).split('.')[0]
dic_primary_full, _ = return_metadata()
color_ttype = return_colors()
total_s = defaultdict(list)
total_count = defaultdict(int)
for sample, t in dic_primary_full.items():
total_s[t].append(sample)
total_count[t] += 1
sorted_ttyps = sorted(total_count, key=total_count.get, reverse=True)
treated = pickle.load(gzip.open(file))
forbidden = ['RADIATION', 'TOPOII']
matrix_treated = defaultdict(lambda: defaultdict(int))
for sample, t in dic_primary_full.items():
for k, d in treated.items():
if k not in forbidden:
if sample in d[t]['YES']:
matrix_treated[sample][k] = 1
elif sample in d[t]['NO']:
matrix_treated[sample][k] = 0
d_treatments = defaultdict(int)
for k, d in treated.items():
if k not in forbidden:
for ttype, l in d.items():
d_treatments[k] += len(l['YES'])
sorted_treatments = sorted(d_treatments,
key=d_treatments.get,
reverse=True)
out = pd.DataFrame.from_dict(matrix_treated, orient='index')
out['TTYPE'] = [dic_primary_full[t] for t in out.index.tolist()]
out['sum'] = out.sum(axis=1)
out = out[out['sum'] > 0].drop('sum', axis=1)
forbidden = ['Double-primary']
order_sample_plot = []
order = []
dic_len = defaultdict(int)
for ttype in tqdm(sorted_ttyps):
if ttype not in forbidden:
subs = out[out['TTYPE'] == ttype]
mat = subs.drop('TTYPE', axis=1).dropna()[sorted_treatments[:30]]
if len(mat) > 1:
n = classic_mutual_exclusivity_visualization(
mat, sorted_treatments[:30])
new_order = n.dendrogram_col.reordered_ind
sample_list = mat.reset_index().loc[new_order]['index'].tolist(
)
order_sample_plot.extend(sample_list)
order.append(ttype)
dic_len[ttype] = len(sample_list)
new_cmap = LinearSegmentedColormap.from_list(
"", ["lightgrey", "grey", "darkred"])
concat = out.loc[order_sample_plot].drop('TTYPE', axis=1)
concat = concat[sorted_treatments[:20]]
if 'specific' in outf:
new_cols = [s.lower() for s in concat.columns]
concat.columns = new_cols
config_params(2)
fig, ax = plt.subplots(1,
2,
figsize=(1, 3),
gridspec_kw={'width_ratios': [1, 27]})
ax2 = sns.heatmap(concat,
cmap=new_cmap,
yticklabels=False,
ax=ax[1],
cbar=False)
ax[1].xaxis.set_ticks_position('top')
bot = 0
for t in order[::-1]:
ax[0].bar(0, dic_len[t], bottom=bot, color=color_ttype[t])
bot += dic_len[t]
ax[0].set_ylim(0, bot)
ax[0].spines['top'].set_visible(False)
ax[0].spines['bottom'].set_visible(False)
ax[0].spines['left'].set_visible(False)
ax[0].spines['right'].set_visible(False)
ax[0].get_yaxis().set_visible(False)
ax[0].get_xaxis().set_visible(False)
plt.xticks(rotation=90)
plt.savefig('figures/EDF1_{}.png'.format(outf), dpi=600)
plt.close()
def create_matrix_treatments(drug, treated_path, exposures_path, tumor_type, keep_sigs=True):
# load treated samples
with gzip.open(treated_path, 'rb') as fd:
treated_samples = pickle.load(fd)
# return equivalence sample tumor type
dic_primary_full, _ = return_metadata()
# load mutational signature exposures at the Pan level
d = pd.read_csv(exposures_path, sep='\t', index_col=0)
# get all signatures
all_sigs = d.index.tolist()
# divide samples treated vs not treated
if tumor_type == 'Pan':
treated = list(treated_samples[drug]['Pan']['YES'])
notreated = list(treated_samples[drug]['Pan']['NO'])
else:
treated = list(treated_samples[drug][tumor_type]['YES'])
notreated = list(treated_samples[drug][tumor_type]['NO'])
# select samples treated if they exist in the dataframe.
s_treat = [s for s in d.columns if s in treated]
s_nottreat = [s for s in d.columns if s in notreated]
# get samples and transpose them
affected = d[list(s_treat)].T
notaffected = d[list(s_nottreat)].T
affected_reg = affected.copy()
notaffected_reg = notaffected.copy()
# add variable response
affected_reg['resp'] = 1
notaffected_reg['resp'] = 0
# concat all samples
toregression = pd.concat([affected_reg, notaffected_reg])
# if keep_sigs flag is True, we will select for the regression all the signatures which
# have at least exposure in one fifth of the samples
if keep_sigs:
keep_exposed_signatures = []
for signature in affected_reg.columns:
exposed_to_sig = len(affected_reg[affected_reg[signature] > 10])
if exposed_to_sig > (len(affected_reg) / 5):
keep_exposed_signatures.append(signature)
keep_exposed_signatures = keep_exposed_signatures + ['resp']
toregression = toregression[keep_exposed_signatures]
# add intercept
toregression['_intercept'] = 1
# add tumor type in the dataframe
for ttype_l in set(dic_primary_full.values()):
# get whether the sample belongs to a specific tumor type
val = [1 if dic_primary_full.get(s, s) == ttype_l else 0 for s in toregression.index.tolist()]
toregression[ttype_l] = val
return toregression, all_sigs, list(set(list(dic_primary_full.values()))), len(treated)
def merge_individual_samples_hartwig(path_hartwig, outpath):
"""
This should merge all tumor types, remove those which belong to a primary,
and then prepare for the extraction.
:param path_hartwig:
:return:
"""
os.makedirs(outpath, exist_ok=True)
# metadata of biopsy
dic_primary_full, dic_secondary_fixed = return_metadata()
all_hartwig = glob(path_hartwig)
dic_muts = defaultdict(int)
all_files = []
for f in tqdm(all_hartwig, total=len(all_hartwig)):
# remove those that do not belong to a good origin
sample_name = os.path.basename(f).split('.')[0].split('_')[1]
if sample_name in dic_primary_full:
df = pd.read_csv(f, sep='\t')
all_files.append(df)
dic_muts[sample_name] = len(df)
hart_df = pd.concat(all_files)
print('sorting...')
hart_df.sort_values(by=['CHROM', 'POS'], inplace=True)
hart_df['PRIMARY'] = hart_df['SAMPLE'].map(dic_primary_full)
hart_df['PRIMARY_EXTENDED'] = hart_df['SAMPLE'].map(dic_secondary_fixed)
# we will split indels and SNVs, without considering hypermutants
for ttype, data in tqdm(hart_df.groupby(by='PRIMARY')):
outfile = '{}/{}.all.gz'.format(outpath, ttype)
data.to_csv(outfile, sep='\t', header=True, index=False, compression='gzip')
outfile = '{}/{}.snvs.gz'.format(outpath, ttype)
data[data['CLASS'] == 'SNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}.indels.gz'.format(outpath, ttype)
data[data['CLASS'] == 'INDEL'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}.dbs.gz'.format(outpath, ttype)
data[data['CLASS'] == 'DBS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}.mnvs.gz'.format(outpath, ttype)
data[data['CLASS'] == 'MNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}.complex_indels.gz'.format(outpath, ttype)
data[data['CLASS'] == 'COMPLEX_INDELS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
# 2- save ALL file
# Pan No Skin and no Unknown for SigProfiler extraction
data = hart_df[
(hart_df['PRIMARY'] != 'Skin') &
(hart_df['PRIMARY'] != 'nan') &
(hart_df['PRIMARY'] != 'Unknown') &
(hart_df['PRIMARY'] != 'CUP')
]
outfile = '{}/PanNoSkinNoUnknown.snvs.gz'.format(outpath)
data[data['CLASS'] == 'SNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/PanNoSkinNoUnknown.indels.gz'.format(outpath)
data[data['CLASS'] == 'INDEL'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/PanNoSkinNoUnknown.dbs.gz'.format(outpath)
data[data['CLASS'] == 'DBS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/PanNoSkinNoUnknown.mnvs.gz'.format(outpath)
data[data['CLASS'] == 'MNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/PanNoSkinNoUnknown.complex_indels.gz'.format(outpath)
data[data['CLASS'] == 'COMPLEX_INDELS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
# define hypermutators based on SNV count. An hypermutated sample is a sample
# with the total number of SNVs higher than 2.5IQR from the median of the entire distribution
sample_counts = hart_df[hart_df['CLASS'] == 'SNV']['SAMPLE'].value_counts().to_dict()
IQR = iqr(list(sample_counts.values()))
median = np.median(list(sample_counts.values()))
cutoff = median + 2.5 * IQR
hypersamples = [s for s, v in sample_counts.items() if v > cutoff]
nothypersamples = [s for s, v in sample_counts.items() if v <= cutoff]
hypers = [hypersamples, nothypersamples]
labels = ['Hypermutated', 'NotHypermutated']
ttype = 'Pan'
# Split into hypermutants and no hypermutants for Signature Analyzer extraction
for h, l in zip(hypers, labels):
subsdata = hart_df[hart_df['SAMPLE'].isin(h)]
outfile = '{}/{}{}.all.gz'.format(outpath, ttype, l)
subsdata.to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}{}.snvs.gz'.format(outpath, ttype, l)
subsdata[subsdata['CLASS'] == 'SNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}{}.indels.gz'.format(outpath, ttype, l)
subsdata[subsdata['CLASS'] == 'INDEL'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}{}.dbs.gz'.format(outpath, ttype, l)
subsdata[subsdata['CLASS'] == 'DBS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}{}.mnvs.gz'.format(outpath, ttype, l)
subsdata[subsdata['CLASS'] == 'MNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/{}{}.complex_indels.gz'.format(outpath, ttype, l)
subsdata[subsdata['CLASS'] == 'COMPLEX_INDELS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/Pan.all.gz'.format(outpath)
hart_df.to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/Pan.snvs.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'SNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/Pan.indels.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'INDEL'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/Pan.dbs.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'DBS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/Pan.mnvs.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'MNV'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
outfile = '{}/Pan.complex_indels.gz'.format(outpath)
hart_df[hart_df['CLASS'] == 'COMPLEX_INDELS'].to_csv(
outfile, sep='\t', header=True, index=False, compression='gzip'
)
def plot_piecharts_signatures(exposures_path, type_mut, type_extraction,
figsize, min_val):
colors_full = {
"SBS": {
25: '#003d7c',
10: '#224ba5',
5: '#4459ce',
2.5: '#6767f7',
1: '#9968c8',
0.5: '#cc6999',
0.25: '#ff6b6b',
0.1: '#ff8962',
0.05: '#ffa759',
0: '#ffc651'
},
"ID": {
1: '#003d7c',
0.5: '#224ba5',
0.1: '#4459ce',
0.05: '#6767f7',
0.04: '#9968c8',
0.03: '#cc6999',
0.02: '#ff6b6b',
0.01: '#ff8962',
0.001: '#ffa759',
0: '#ffc651'
},
"DBS": {
0.5: '#003d7c',
0.1: '#224ba5',
0.05: '#4459ce',
0.03: '#6767f7',
0.02: '#9968c8',
0.01: '#cc6999',
0.008: '#ff6b6b',
0.005: '#ff8962',
0.001: '#ffa759',
0: '#ffc651'
}
}
colors = colors_full[type_mut]
result = read_exposures(exposures_path)
config_params()
dic_primary_full, _ = return_metadata()
# result = pd.concat([df, df_mela])
result = result.fillna(0)
signatures = result.columns.tolist()
# get list of similar found signatures in the extraction
similar = [s for s in signatures if type_mut in s]
notsimilar = [s for s in signatures if type_mut not in s]
result['TTYPE'] = [dic_primary_full[t] for t in result.index.tolist()]
dic_sig = defaultdict(lambda: defaultdict(float))
dic_proportion = defaultdict(lambda: defaultdict(float))
for ttype, data in result.groupby(by='TTYPE'):
data2 = data.copy()
data2.drop('TTYPE', axis=1, inplace=True)
for col in data2:
# we normalize it by the number of MB in the human genome (3234)
dic_sig[ttype][col] = data2[
data2[col] > min_val][col].median() / 3234
if type_mut not in col:
dic_proportion[ttype][col.split('_')[0]] = len(
data2[data2[col] > min_val]) / len(data2)
else:
dic_proportion[ttype][col] = len(
data2[data2[col] > min_val]) / len(data2)
medians = pd.DataFrame.from_dict(dic_sig)
# sorting the already known signatures
keep_order_similar = defaultdict(list)
for s in similar:
number = s.split('_')[1].split(type_mut)[1]
try:
keep_n = int(number)
except Exception:
keep_n = int(number[:-1])
keep_order_similar[keep_n].append(str(s))
order_prev_labels = []
order_prev = []
for i in sorted(keep_order_similar, reverse=True):
all_s = []
d_equiv = defaultdict(str)
for sig in keep_order_similar[i]:
ID_sig = '{}_{}_{}'.format(
sig.split('_')[1],
sig.split('_')[2],
sig.split('_')[0])
d_equiv[ID_sig] = sig
sorted_final_k = sorted(d_equiv)
sorted_sigs_list = [d_equiv[ss] for ss in sorted_final_k]
for sim, sig in reversed(list(enumerate(sorted_sigs_list, start=1))):
if len(keep_order_similar[i]) == 1:
order_prev_labels.append('E-{}{} ({}-like, {})'.format(
type_mut,
sig.split('_')[0],
sig.split('_')[1], round(float(sig.split('_')[2]), 3)))
order_prev.append(sig)
else:
order_prev_labels.append('E-{}{} ({}-like {}, {})'.format(
type_mut,
sig.split('_')[0],
sig.split('_')[1], sim, round(float(sig.split('_')[2]),
3)))
order_prev.append(sig)
no_similar_signatures = medians.loc[notsimilar]
new_index = [
int(l.split('_')[0]) for l in no_similar_signatures.index.tolist()
]
no_similar_signatures.index = new_index
no_similar_signatures.sort_index(inplace=True, ascending=True)
names_notsimilar = [
'E-{} {}'.format(type_mut, c)
for c in no_similar_signatures.index.tolist()[::-1]
]
# merge new and old
merged = pd.concat([
no_similar_signatures.sort_index(ascending=False),
medians.loc[order_prev],
])
# merged = merged.loc[order_prev+small_newset.index.tolist()]
merged_labels = names_notsimilar + order_prev_labels
config_params(5)
fig, ax = plt.subplots(1, 1, figsize=figsize)
# plt.grid(b=True, which='major',)
for yval, (i, row) in enumerate(merged.iterrows()):
for xval, t in enumerate(merged.columns.tolist()):
val = row[t]
if val > 0:
color = None
for number in sorted(colors.keys(), reverse=True):
if (val > number) & (color is None):
color_scatter = colors[number]
break
if type_mut in str(i):
plt.scatter(xval,
yval,
c=color_scatter,
s=dic_proportion[t][i] * 20)
else:
plt.scatter(xval,
yval,
c=color_scatter,
s=dic_proportion[t][str(i)] * 20)
ax.set_xticks(np.arange(len(merged.T)))
ax.set_xticklabels(merged.columns.tolist(), rotation=90)
ax.set_yticks(np.arange(len(merged)))
ax.set_yticklabels(merged_labels)
ax.xaxis.set_ticks_position('top')
ax.set_axisbelow(True)
ax.yaxis.grid(color='gray', linestyle='dashed', alpha=0.3)
ax.xaxis.grid(color='gray', linestyle='dashed', alpha=0.3)
ax.spines['top'].set_visible(False)
ax.spines['bottom'].set_visible(False)
ax.spines['left'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.xaxis.set_ticks_position('none')
ax.yaxis.set_ticks_position('none')
plt.ylim(-1, len(merged))
plt.tight_layout()
plt.savefig('figures/{}/supp1_{}.svg'.format(type_extraction, type_mut))
plt.savefig('figures/{}/supp1_{}.png'.format(type_extraction, type_mut),
dpi=600)
plt.close()
def do_plot(samples_lowbraca, samples_braca, not_breaked, only_treated, exp):
dic_primary_full, _ = return_metadata()
color_ttype = return_colors()
fig, ax = plt.subplots(1, 1, figsize=(2, 1.5))
config_params(6.5)
sigs = ['12_ID6_0.962941_1']
exposures_not_breaked_1 = exp[sigs].sum(axis=1).loc[[
i for i in not_breaked if i in samples_lowbraca
]].dropna()
exposures_not_breaked_2 = exp[sigs].sum(axis=1).loc[[
i for i in not_breaked if i in samples_braca
]].dropna()
exposures_breaked_1 = exp[sigs].sum(axis=1).loc[[
i for i in only_treated if i in samples_lowbraca
]].dropna()
exposures_breaked_2 = exp[sigs].sum(axis=1).loc[[
i for i in only_treated if i in samples_braca
]].dropna()
sns.boxplot(data=[
exposures_not_breaked_1, exposures_breaked_1, exposures_not_breaked_2,
exposures_breaked_2
],
linewidth=0.6,
showfliers=False,
color='#cbcacbff')
plt.ylabel('Indels DSB repair by\nnon-homologous end-joining')
plt.xticks([0, 1, 2, 3], [
'Not radiated no BRCAness ({})'.format(len(exposures_not_breaked_1)),
'Radiated no BRCAness ({})'.format(len(exposures_breaked_1)),
'Not radiated BRCAness ({})'.format(len(exposures_not_breaked_2)),
'Radiated BRCAness ({})'.format(len(exposures_breaked_2))
],
rotation=90)
plotdot = []
colors = []
for sample in [i for i in not_breaked if i in samples_lowbraca]:
plotdot.append(exp[sigs].sum(axis=1).loc[sample])
colors.append(color_ttype[dic_primary_full[sample]])
ax.scatter(
[0 + np.random.uniform(-0.2, 0.2, 1)[0] for i in range(len(plotdot))],
plotdot,
color=colors,
s=1,
alpha=0.2)
plotdot = []
colors = []
for sample in [i for i in only_treated if i in samples_lowbraca]:
plotdot.append(exp[sigs].sum(axis=1).loc[sample])
colors.append(color_ttype[dic_primary_full[sample]])
ax.scatter(
[1 + np.random.uniform(-0.2, 0.2, 1)[0] for i in range(len(plotdot))],
plotdot,
color=colors,
s=1,
alpha=0.2)
plotdot = []
colors = []
for sample in [i for i in not_breaked if i in samples_braca]:
plotdot.append(exp[sigs].sum(axis=1).loc[sample])
colors.append(color_ttype[dic_primary_full[sample]])
ax.scatter(
[2 + np.random.uniform(-0.2, 0.2, 1)[0] for i in range(len(plotdot))],
plotdot,
color=colors,
s=1,
alpha=0.2)
plotdot = []
colors = []
for sample in [i for i in only_treated if i in samples_braca]:
plotdot.append(exp[sigs].sum(axis=1).loc[sample])
colors.append(color_ttype[dic_primary_full[sample]])
ax.scatter(
[3 + np.random.uniform(-0.2, 0.2, 1)[0] for i in range(len(plotdot))],
plotdot,
color=colors,
s=1,
alpha=0.2)
plt.ylim(0, 700)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.savefig('figures/radiation.svg')
plt.show()
##################
fig, ax = plt.subplots(
1,
1,
)
stat, pval1 = mannwhitneyu(exposures_not_breaked_1, exposures_breaked_1)
print("Not radiated no BRCAnes vs Radiated no BRCAness", pval1)
stat, pval2 = mannwhitneyu(exposures_not_breaked_2, exposures_breaked_2)
print("Not radiated BRCAnes vs Radiated BRCAness", pval2)
stat, pval3 = mannwhitneyu(exposures_breaked_1, exposures_breaked_2)
print("radiated no BRCAnes vs Radiated BRCAness", pval3)
ax.text(1, 1, "$\it{P}$" + " = {}".format(sci_notation(pval1)), fontsize=7)
ax.text(1, 4, "$\it{P}$" + " = {}".format(sci_notation(pval2)), fontsize=7)
ax.text(1, 8, "$\it{P}$" + " = {}".format(sci_notation(pval3)), fontsize=7)
plt.xlim(0, 5)
plt.ylim(0, 10)
plt.savefig('figures/radiation_pvals.svg')
sys.exit()
