def rotational_bars(rot_high_files, rot_low_files):
nuceriod_plt.config_params(11)
df_high = increase.load_d(rot_high_files)
df_high['rot'] = 'high'
df_low = increase.load_d(rot_low_files)
df_low['rot'] = 'low'
df = pd.concat([df_high, df_low])
fig, axs = plt.subplots(nrows=len(df.groupby(by='name')),
ncols=1,
figsize=(1.75, 8))
order = ['low', 'high']
for ix, (sig, data) in enumerate(df.groupby(by='name')):
xvals = []
yvals = []
colors = []
count = 0
for i in order:
val = data[data['rot'] == i]['snr'].tolist()[0]
yvals.append(val)
xvals.append(count)
colors.append(COLORS[sig])
count += 1
axs[ix].bar(xvals, yvals, color=colors, label=['low', 'high'])
axs[ix].set_xticks([0, 1])
axs[ix].set_xticklabels(('low', 'high'), fontsize=11)
axs[ix].set_ylabel('SNR')
axs[ix].spines['right'].set_visible(False)
axs[ix].spines['top'].set_visible(False)
plt.tight_layout()
def compare(files_deconstructsigs, files_sigfit):
nuceriod_plt.config_params(11)
df_deconstructsigs = increase.load_d(files_deconstructsigs)
df_deconstructsigs['control'] = 'deconstructsigs'
df_sigfit = increase.load_d(files_sigfit)
df_sigfit['control'] = 'sigfit'
toplot = pd.concat([df_deconstructsigs, df_sigfit])
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(3, 2))
sig_list = []
decon_list = []
sig_list2 = []
decon_list2 = []
for i, data in toplot.groupby(by='name'):
if len(data) == 2:
row = data.iloc[0]
snr_decon = data[data['control'] == 'sigfit']['snr'].iloc[0]
snr_sigfit = data[data['control'] ==
'deconstructsigs']['snr'].iloc[0]
sig_list.append(snr_sigfit)
decon_list.append(snr_decon)
if row['cross_validation_max'] < 0:
ax.scatter(-np.log(snr_decon),
-np.log(snr_sigfit),
c=COLORS[i])
sig_list2.append(-np.log(snr_sigfit))
decon_list2.append(-np.log(snr_decon))
else:
ax.scatter(np.log(snr_decon), np.log(snr_sigfit), c=COLORS[i])
sig_list2.append(np.log(snr_sigfit))
decon_list2.append(np.log(snr_decon))
plt.xlabel('Period')
ylabels = [str(2**abs(i)) for i in range(2, 10, 2)]
yfinal = ylabels[::-1] + ylabels[1:]
plt.xticks(np.arange(-6, 8, 2), yfinal)
ylabels = [str(2**abs(i)) for i in range(2, 10, 2)]
yfinal = ylabels[::-1] + ylabels[1:]
plt.yticks(np.arange(-6, 8, 2), yfinal)
slope, intercept, r_value, p_value, std_err = stats.linregress(
sig_list2, decon_list2)
xvals = np.arange(-6, 8, 2)
yvals = [slope * y + intercept for y in xvals]
plt.plot(xvals, yvals)
R, pval = stats.pearsonr(sig_list, decon_list)
plt.text(-4, 3, 'R = {}\npval = {}'.format(round(R, 3), round(pval, 3)))
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.ylabel('SNR (Sigfit)')
plt.xlabel('SNR (Deconstructsig)')
plt.tight_layout()
def compare(cohorts_5mer, cohorts_3mer, cohorts_linker, tumors=None):
nuceriod_plt.config_params(14)
df_5mer = increase.load_d(cohorts_5mer)
df_5mer['control'] = 'mer5'
df_3mer = increase.load_d(cohorts_3mer)
df_3mer['control'] = 'mer3'
df_linker = increase.load_d(cohorts_linker)
df_linker['control'] = 'linker'
df = pd.concat([df_5mer, df_3mer, df_linker])
df['increase_in'] = df['observed_in'] - df['expected_in']
df['prop_increase_in'] = df['increase_in'] / df['expected_in']
df['ttype'] = df['name'].map(TTYPES)
if tumors is not None:
df = df[df['name'].isin(tumors)]
toplot = df.sort_values(by='prop_increase_in', ascending=True)
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(10, 5))
labels = []
colors = []
red_patch = mpatches.Patch(color='red', label='3-mers')
green_patch = mpatches.Patch(color='green',
label='no nucleosomes in context')
orange_patch = mpatches.Patch(color='orange', label='5-mers')
for ix, (ttype, data) in enumerate(
toplot.sort_values(by='snr', ascending=True).groupby(by='ttype',
sort=False)):
snr1 = data[data['control'] == 'mer3']['snr'].tolist()[0]
snr2 = data[data['control'] == 'linker']['snr'].tolist()[0]
snr3 = data[data['control'] == 'mer5']['snr'].tolist()[0]
colors.append('red')
colors.append('green')
colors.append('orange')
ax.scatter(ix, math.log2(snr1), color='red', s=15, alpha=0.8)
ax.scatter(ix, math.log2(snr2), color='green', s=15, alpha=0.8)
ax.scatter(ix, math.log2(snr3), color='orange', s=15, alpha=0.8)
labels.append(ttype)
plt.xticks([i for i in range(ix + 1)], labels, rotation=90)
tick = [2, 4, 6, 8]
plt.yticks(tick, [str(2**t) for t in tick])
plt.ylabel('log2(SNR)')
plt.legend(handles=[red_patch, green_patch, orange_patch])
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.tight_layout()
def rotational(cohorts_high, cohorts_low, tumors=None):
nuceriod_plt.config_params(14)
df_high = increase.load_d(cohorts_high)
df_high['control'] = 'high'
df_low = increase.load_d(cohorts_low)
df_low['control'] = 'low'
df = pd.concat([df_high, df_low])
df['increase_in'] = df['observed_in'] - df['expected_in']
df['prop_increase_in'] = df['increase_in'] / df['expected_in']
df['ttype'] = df['name'].map(TTYPES)
if tumors is not None:
df = df[df['name'].isin(tumors)]
toplot = df.sort_values(by='prop_increase_in', ascending=True)
order = ['low', 'high']
count = 0
xvals = []
yvals = []
colors = []
dic_t = collections.defaultdict(dict)
labels = []
for sig, data in toplot.sort_values(by='snr').groupby(by='ttype',
sort=False):
if sig in COLORS:
labels.append('{}'.format(sig))
for i in order:
val = data[data['control'] == i]['snr'].tolist()[0]
dic_t['Sign {}'.format(sig)][i] = val
yvals.append(math.log2(val))
xvals.append(count)
colors.append(COLORS[sig])
count += 1
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(10, 1.5))
ax.bar(xvals, yvals, color=colors, label=labels)
ax.set_ylabel('SNR')
plt.xticks(np.arange(0.5, 54, 2), labels, rotation=90, fontsize=13)
tick = [2, 4, 6, 8]
plt.yticks(tick, [str(2**t) for t in tick])
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
def _load(files):
df = increase.load_d(files)
df['increase_in'] = df['observed_in'] - df['expected_in']
df['prop_increase_in'] = df['increase_in'] / df['expected_in']
df['outname'] = df['name'].str.replace('Signature_', 'Sign ')
df['empirical_pvalue_snr'].replace(0, 0.001, inplace=True)
df['empirical_pvalue_snr'] = df['empirical_pvalue_snr'].fillna(1)
qvals = multipletests(df['empirical_pvalue_snr'].tolist(), method='fdr_bh')
df['qvals_snr'] = qvals[1]
return df
def find(files, min_muts=None):
df = increase.load_d(files)
df['ttype'] = df['name'].map(TTYPES)
if min_muts is not None:
df = df[df['nmuts_whole_nucleosome'] > min_muts]
df.sort_values(by='nmuts_whole_nucleosome', ascending=False, inplace=True)
ttypes = set()
names = []
for i, row in df.iterrows():
if row['ttype'] not in ttypes:
names.append(row['name'])
ttypes.add(row['ttype'])
return names
def _load(cohorts, tumors):
df = increase.load_d(cohorts)
df['increase_in'] = df['observed_in'] - df['expected_in']
df['prop_increase_in'] = df['increase_in'] / df['expected_in']
df['ttype'] = df['name'].map(TTYPES)
df = df[df['name'].isin(tumors)].sort_values(by='prop_increase_in',
ascending=True)
# add Q-value
df['empirical_pvalue_snr'] = df['empirical_pvalue_snr'].fillna(1)
df['empirical_pvalue_snr'].replace(0, 0.001, inplace=True)
qvals = multipletests(df['empirical_pvalue_snr'].tolist(), method='fdr_bh')
df['qvals_snr'] = qvals[1]
return df
def _load(files):
df = increase.load_d(files)
df['increase_in'] = df['observed_in'] - df['expected_in']
df['prop_increase_in'] = df['increase_in'] / df['expected_in']
df = df.sort_values(by='prop_increase_in', ascending=True)
df = df[df['nmuts_whole_nucleosome'] > 1000]
# add Q-value
df['empirical_pvalue_snr'] = df['empirical_pvalue_snr'].fillna(1)
df['empirical_pvalue_snr'].replace(0, 0.001, inplace=True)
qvals = multipletests(df['empirical_pvalue_snr'].tolist(), method='fdr_bh')
df['qvals_snr'] = qvals[1]
df['corrected_snr'] = df.apply(_sign_SNR, axis=1)
return df
def cohorts(files, colors):
nuceriod_plt.config_params_full(font_size=7)
df = increase.load_d(files)
df['increase_in'] = df['observed_in'] - df['expected_in']
df['prop_increase_in'] = df['increase_in'] / df['expected_in']
df['signature_name'] = df['name'].str.split('__').str.get(1)
df['ttype'] = df['name'].str.split('__').str.get(0)
df['empirical_pvalue_snr'].replace(0, 0.001, inplace=True)
df['empirical_pvalue_snr'] = df['empirical_pvalue_snr'].fillna(1)
qvals = multipletests(df['empirical_pvalue_snr'].tolist(), method='fdr_bh')
df['qvals_snr'] = qvals[1]
fig, axs = plt.subplots(nrows=2, ncols=1, sharex=True, figsize=(7, 2))
signature_groups = df.groupby('signature_name')
for i in range(1, 31):
ix = i - 1
try:
data = signature_groups.get_group('Signature_{}'.format(i))
except KeyError:
continue
else:
for _, row in data.iterrows():
if row['cross_validation_max'] > 0:
if row['qvals_snr'] < 0.05 and row['snr'] > 8 and 10 < row[
'peak'] < 10.4:
axs[0].scatter(jitter(ix),
np.log2(row['snr']),
c=colors[row['ttype']],
s=25,
edgecolor='black',
linewidth=0.5)
else:
axs[0].scatter(jitter(ix),
np.log2(row['snr']),
c='grey',
s=6)
elif row['cross_validation_max'] < 0:
if row['qvals_snr'] < 0.05 and row['snr'] > 8 and 10 < row[
'peak'] < 10.4:
axs[1].scatter(jitter(ix),
-np.log2(row['snr']),
c=colors[row['ttype']],
s=25,
edgecolor='black',
linewidth=0.5)
else:
axs[1].scatter(jitter(ix),
-np.log2(row['snr']),
c='grey',
s=8)
yvals = [i for i in range(2, 10, 2)]
axs[0].set_yticks(yvals)
yvals = [i for i in range(-8, 0, 2)]
axs[1].set_yticks(yvals)
ylabels = [str(2**abs(i)) for i in range(2, 10, 2)]
axs[0].set_yticklabels(ylabels)
ylabels = ['{}'.format(str(2**abs(i))) for i in range(-8, 0, 2)]
axs[1].set_yticklabels(ylabels)
axs[0].spines['right'].set_visible(False)
axs[0].spines['top'].set_visible(False)
axs[0].set_xlim(-1, 30)
xlabels = [i for i in range(1, 31)]
xpos = [i for i in range(0, 30)]
axs[1].set_xticklabels(xlabels)
plt.xticks(xpos, xlabels)
axs[1].xaxis.set_ticks_position('top')
axs[1].spines['bottom'].set_visible(False)
axs[1].spines['right'].set_visible(False)
axs[0].set_ylim(1.5, 10)
axs[1].set_ylim(-10, -1.5)
plt.tight_layout()