def plot_tuning_mtype_grid(par_df, auc_df, use_clf, args, cdata):
par_count = len(use_clf.tune_priors)
fig, axarr = plt.subplots(figsize=(0.5 + 7 * par_count, 7 * par_count),
nrows=par_count,
ncols=par_count)
auc_vals = auc_df.quantile(q=0.25, axis=1)
auc_clrs = auc_vals.apply(auc_cmap)
size_vec = [
461 * sum(cdata.train_pheno(mtype)) /
(len(cdata.get_samples()) * par_count) for mtype in auc_vals.index
]
for i, (par_name, tune_distr) in enumerate(use_clf.tune_priors):
axarr[i, i].grid(False)
if detect_log_distr(tune_distr):
use_distr = [np.log10(par_val) for par_val in tune_distr]
par_lbl = par_name + '\n(log-scale)'
else:
use_distr = tune_distr
par_lbl = par_name
distr_diff = np.mean(
np.array(use_distr[1:]) - np.array(use_distr[:-1]))
plt_min = use_distr[0] - distr_diff / 2
plt_max = use_distr[-1] + distr_diff / 2
axarr[i, i].set_xlim(plt_min, plt_max)
axarr[i, i].set_ylim(plt_min, plt_max)
axarr[i, i].text((plt_min + plt_max) / 2, (plt_min + plt_max) / 2,
par_lbl,
ha='center',
fontsize=28,
weight='semibold')
for par_val in use_distr:
axarr[i, i].axhline(y=par_val,
color='#116611',
ls='--',
linewidth=4.1,
alpha=0.27)
axarr[i, i].axvline(x=par_val,
color='#116611',
ls='--',
linewidth=4.1,
alpha=0.27)
for (i, (par_name1, tn_distr1)), (j, (par_name2, tn_distr2)) in combn(
enumerate(use_clf.tune_priors), 2):
if detect_log_distr(tn_distr1):
use_distr1 = [np.log10(par_val) for par_val in tn_distr1]
par_meds1 = np.log10(par_df[par_name1]).median(axis=1)
par_means1 = np.log10(par_df[par_name1]).mean(axis=1)
distr_diff = np.mean(
np.log10(np.array(tn_distr1[1:])) -
np.log10(np.array(tn_distr1[:-1])))
plt_ymin = np.log10(tn_distr1[0]) - distr_diff / 2
plt_ymax = np.log10(tn_distr1[-1]) + distr_diff / 2
else:
use_distr1 = tn_distr1
par_meds1 = par_df[par_name1].median(axis=1)
par_means1 = par_df[par_name1].mean(axis=1)
distr_diff = np.mean(
np.array(tn_distr1[1:]) - np.array(tn_distr1[:-1]))
plt_ymin = tn_distr1[0] - distr_diff / 2
plt_ymax = tn_distr1[-1] + distr_diff / 2
if detect_log_distr(tn_distr2):
use_distr2 = [np.log10(par_val) for par_val in tn_distr2]
par_meds2 = np.log10(par_df[par_name2]).median(axis=1)
par_means2 = np.log10(par_df[par_name2]).mean(axis=1)
distr_diff = np.mean(
np.log10(np.array(tn_distr2[1:])) -
np.log10(np.array(tn_distr2[:-1])))
plt_xmin = np.log10(tn_distr2[0]) - distr_diff / 2
plt_xmax = np.log10(tn_distr2[-1]) + distr_diff / 2
else:
use_distr2 = tn_distr2
par_meds2 = par_df[par_name2].median(axis=1)
par_means2 = par_df[par_name2].mean(axis=1)
distr_diff = np.mean(
np.array(tn_distr2[1:]) - np.array(tn_distr2[:-1]))
plt_xmin = tn_distr2[0] - distr_diff / 2
plt_xmax = tn_distr2[-1] + distr_diff / 2
par_meds1 = par_meds1[auc_clrs.index]
par_meds2 = par_meds2[auc_clrs.index]
y_adj = (plt_ymax - plt_ymin) / len(tn_distr1)
x_adj = (plt_xmax - plt_xmin) / len(tn_distr2)
plt_adj = (plt_xmax - plt_xmin) / (plt_ymax - plt_ymin)
for med1, med2 in set(zip(par_meds1, par_meds2)):
use_indx = (par_meds1 == med1) & (par_meds2 == med2)
cnt_adj = use_indx.sum()**0.49
use_sizes = [s for s, ix in zip(size_vec, use_indx) if ix]
sort_indx = sorted(enumerate(use_sizes),
key=lambda x: x[1],
reverse=True)
from circlify import circlify
mpl.use('Agg')
for k, circ in enumerate(circlify([s for _, s in sort_indx])):
axarr[i, j].scatter(
med2 + (1 / 23) * cnt_adj * circ.y * plt_adj,
med1 + (1 / 23) * cnt_adj * circ.x * plt_adj**-1,
s=sort_indx[k][1],
c=auc_clrs[use_indx][sort_indx[k][0]],
alpha=0.36,
edgecolor='black')
par_means1 += np.random.normal(0, y_adj / 27, auc_df.shape[0])
par_means2 += np.random.normal(0, x_adj / 27, auc_df.shape[0])
axarr[j, i].scatter(par_means1[auc_clrs.index],
par_means2[auc_clrs.index],
s=size_vec,
c=auc_clrs,
alpha=0.36,
edgecolor='black')
axarr[i, j].set_xlim(plt_xmin, plt_xmax)
axarr[i, j].set_ylim(plt_ymin, plt_ymax)
axarr[j, i].set_ylim(plt_xmin, plt_xmax)
axarr[j, i].set_xlim(plt_ymin, plt_ymax)
annot_placed = place_annot(par_meds2,
par_meds1,
size_vec=size_vec,
annot_vec=auc_vals.index,
x_range=plt_xmax - plt_xmin,
y_range=plt_ymax - plt_ymin)
for annot_x, annot_y, annot, halign in annot_placed:
axarr[i, j].text(annot_x, annot_y, annot, size=11, ha=halign)
for par_val1 in use_distr1:
axarr[i, j].axhline(y=par_val1,
color='#116611',
ls=':',
linewidth=2.3,
alpha=0.19)
axarr[j, i].axvline(x=par_val1,
color='#116611',
ls=':',
linewidth=2.3,
alpha=0.19)
for par_val2 in use_distr2:
axarr[i, j].axvline(x=par_val2,
color='#116611',
ls=':',
linewidth=2.3,
alpha=0.19)
axarr[j, i].axhline(y=par_val2,
color='#116611',
ls=':',
linewidth=2.3,
alpha=0.19)
plt.tight_layout()
fig.savefig(os.path.join(
plot_dir, args.expr_source,
"{}__samps-{}".format(args.cohort, args.samp_cutoff),
args.model_name.split('__')[0],
"{}__tuning-mtype-grid.svg".format(args.model_name.split('__')[1])),
bbox_inches='tight',
format='svg')
plt.close()
def plot_tuning_mtype(par_df, auc_df, use_clf, args, cdata):
fig, axarr = plt.subplots(figsize=(1 + 9 * len(use_clf.tune_priors), 13),
nrows=3,
ncols=len(use_clf.tune_priors),
gridspec_kw={'height_ratios': [1, 0.3, 1]},
squeeze=False,
sharex=False,
sharey=True)
auc_vals = auc_df.quantile(q=0.25, axis=1)
size_vec = [
198 * len(mtype.get_samples(cdata.mtree)) / len(cdata.get_samples())
for mtype in auc_vals.index
]
for i, (par_name, tune_distr) in enumerate(use_clf.tune_priors):
axarr[1, i].set_axis_off()
axarr[2, i].tick_params(length=6)
if detect_log_distr(tune_distr):
med_vals = np.log10(par_df[par_name]).median(axis=1)
mean_vals = np.log10(par_df[par_name]).mean(axis=1)
use_distr = [np.log10(par_val) for par_val in tune_distr]
par_lbl = par_name + '\n(log-scale)'
else:
med_vals = par_df[par_name].median(axis=1)
mean_vals = par_df[par_name].mean(axis=1)
use_distr = tune_distr
par_lbl = par_name
med_vals = med_vals[auc_vals.index]
mean_vals = mean_vals[auc_vals.index]
distr_diff = np.mean(
np.array(use_distr[1:]) - np.array(use_distr[:-1]))
for j in range(3):
axarr[j, i].set_xlim(use_distr[0] - distr_diff / 2,
use_distr[-1] + distr_diff / 2)
axarr[1, i].text((use_distr[0] + use_distr[-1]) / 2,
0.5,
par_lbl,
ha='center',
va='center',
fontsize=25,
weight='semibold')
med_vals += np.random.normal(0, (use_distr[-1] - use_distr[0]) /
(len(tune_distr) * 17), auc_df.shape[0])
mean_vals += np.random.normal(0, (use_distr[-1] - use_distr[0]) /
(len(tune_distr) * 23), auc_df.shape[0])
axarr[0, i].scatter(med_vals,
auc_vals,
s=size_vec,
c='black',
alpha=0.23)
axarr[2, i].scatter(mean_vals,
auc_vals,
s=size_vec,
c='black',
alpha=0.23)
axarr[0, i].set_ylim(0, 1)
axarr[2, i].set_ylim(0, 1)
axarr[0, i].set_ylabel("1st Quartile AUC", size=19, weight='semibold')
axarr[2, i].set_ylabel("1st Quartile AUC", size=19, weight='semibold')
axarr[0, i].axhline(y=0.5,
color='#550000',
linewidth=2.3,
linestyle='--',
alpha=0.32)
axarr[2, i].axhline(y=0.5,
color='#550000',
linewidth=2.3,
linestyle='--',
alpha=0.32)
for par_val in use_distr:
axarr[1, i].axvline(x=par_val,
color='#116611',
ls='--',
linewidth=3.4,
alpha=0.27)
axarr[0, i].axvline(x=par_val,
color='#116611',
ls=':',
linewidth=1.3,
alpha=0.16)
axarr[2, i].axvline(x=par_val,
color='#116611',
ls=':',
linewidth=1.3,
alpha=0.16)
annot_placed = place_annot(med_vals,
auc_vals.values.tolist(),
size_vec=size_vec,
annot_vec=auc_vals.index,
x_range=use_distr[-1] - use_distr[0] +
2 * distr_diff,
y_range=1)
for annot_x, annot_y, annot, halign in annot_placed:
axarr[0, i].text(annot_x, annot_y, annot, size=8, ha=halign)
plt.tight_layout(h_pad=0)
fig.savefig(os.path.join(
plot_dir, args.expr_source,
"{}__samps-{}".format(args.cohort, args.samp_cutoff),
args.model_name.split('__')[0],
"{}__tuning-mtype.svg".format(args.model_name.split('__')[1])),
bbox_inches='tight',
format='svg')
plt.close()
def plot_tuned_auc(out_dict, phn_dict, auc_dict, args):
tune_priors = tuple(out_dict.values())[0]['Clf'].tune_priors
fig, axarr = plt.subplots(figsize=(17, 1 + 7 * len(tune_priors)),
nrows=len(tune_priors),
ncols=1,
squeeze=False)
use_srcs = sorted(set(src for src, _, _ in out_dict.keys()))
src_mrks = dict(zip(use_srcs, use_marks[:len(use_srcs)]))
use_cohs = sorted(set(coh for _, coh, _ in out_dict.keys()))
coh_clrs = dict(
zip(use_cohs, sns.color_palette("muted", n_colors=len(use_cohs))))
for ax, (par_name, tune_distr) in zip(axarr.flatten(), tune_priors):
if detect_log_distr(tune_distr):
par_fnc = np.log10
plt_xmin = 2 * np.log10(tune_distr[0]) - np.log10(tune_distr[1])
plt_xmax = 2 * np.log10(tune_distr[-1]) - np.log10(tune_distr[-2])
else:
par_fnc = lambda x: x
plt_xmin = 2 * tune_distr[0] - tune_distr[1]
plt_xmax = 2 * tune_distr[-1] - tune_distr[-2]
for (src, coh, lvls), ols in out_dict.items():
tune_df = ols['Pars'].loc[:, (slice(None), par_name)]
phn_list = phn_dict[src, coh, lvls]
auc_df = auc_dict[src, coh, lvls]
for mtype, vals in tune_df.iterrows():
for (cis_lbl, _), val in vals.iteritems():
ax.scatter(par_fnc(val),
auc_df.loc[mtype, cis_lbl],
marker=src_mrks[src],
s=371 * np.mean(phn_list[mtype]),
c=[coh_clrs[coh]],
alpha=0.17,
edgecolor='none')
ax.set_xlim(plt_xmin, plt_xmax)
ax.set_ylim(0.48, 1.02)
ax.tick_params(labelsize=19)
ax.set_xlabel('Tuned {} Value'.format(par_name),
fontsize=27,
weight='semibold')
ax.axhline(y=1.0, color='black', linewidth=2.1, alpha=0.37)
ax.axhline(y=0.5,
color='#550000',
linewidth=2.7,
linestyle='--',
alpha=0.29)
fig.text(-0.01,
0.5,
'Aggregate AUC',
ha='center',
va='center',
fontsize=27,
weight='semibold',
rotation='vertical')
plt.tight_layout(h_pad=1.7)
fig.savefig(os.path.join(plot_dir,
"{}__tuned-auc.svg".format(args.classif)),
bbox_inches='tight',
format='svg')
plt.close()
def plot_tuning_profile(out_dict, auc_dict, args):
tune_priors = tuple(out_dict.values())[0]['Clf'].tune_priors
fig, axarr = plt.subplots(figsize=(17, 1 + 7 * len(tune_priors)),
nrows=len(tune_priors),
ncols=1,
squeeze=False)
use_aucs = pd.concat(auc_dict.values()).round(4)
auc_bins = pd.qcut(
use_aucs.values.flatten(),
q=[0., 0.5, 0.75, 0.8, 0.85, 0.9, 0.92, 0.94, 0.96, 0.98, 0.99, 1.],
precision=5).categories
use_srcs = sorted(set(src for src, _, _ in out_dict.keys()))
src_mrks = dict(zip(use_srcs, use_marks[:len(use_srcs)]))
use_cohs = sorted(set(coh for _, coh, _ in out_dict.keys()))
coh_clrs = dict(
zip(use_cohs, sns.color_palette("muted", n_colors=len(use_cohs))))
for ax, (par_name, tune_distr) in zip(axarr.flatten(), tune_priors):
if detect_log_distr(tune_distr):
par_fnc = np.log10
plt_xmin = 2 * np.log10(tune_distr[0]) - np.log10(tune_distr[1])
plt_xmax = 2 * np.log10(tune_distr[-1]) - np.log10(tune_distr[-2])
else:
par_fnc = lambda x: x
plt_xmin = 2 * tune_distr[0] - tune_distr[1]
plt_xmax = 2 * tune_distr[-1] - tune_distr[-2]
plot_df = pd.DataFrame([])
for (src, coh, lvls), ols in out_dict.items():
for cis_lbl in cis_lbls:
tune_vals = pd.DataFrame.from_records(ols['Acc'].loc[:,
(cis_lbl,
'avg')])
tune_vals -= pd.DataFrame.from_records(ols['Acc'].loc[:,
(cis_lbl,
'std')])
par_vals = pd.DataFrame.from_records(
ols['Acc'].loc[:, (cis_lbl, 'par')]).applymap(
itemgetter(par_name)).applymap(par_fnc)
tune_vals.index = ols['Acc'].index
par_vals.index = ols['Acc'].index
par_df = pd.concat(
[par_vals.stack(), tune_vals.stack()],
axis=1,
keys=['par', 'auc'])
par_df['auc_bin'] = [
auc_bins.get_loc(
round(auc_dict[src, coh, lvls].loc[mtype, cis_lbl], 4))
for mtype, _ in par_df.index
]
plot_df = pd.concat([plot_df, par_df])
for auc_bin, bin_vals in plot_df.groupby('auc_bin'):
plot_vals = bin_vals.groupby('par').mean()
ax.plot(plot_vals.index, plot_vals.auc)
ax.set_xlim(plt_xmin, plt_xmax)
ax.set_ylim(0.45, 1.01)
ax.tick_params(labelsize=19)
ax.set_xlabel('Tested {} Value'.format(par_name),
fontsize=27,
weight='semibold')
ax.axhline(y=1.0, color='black', linewidth=2.1, alpha=0.37)
ax.axhline(y=0.5,
color='#550000',
linewidth=2.7,
linestyle='--',
alpha=0.29)
fig.text(-0.01,
0.5,
'Aggregate AUC',
ha='center',
va='center',
fontsize=27,
weight='semibold',
rotation='vertical')
plt.tight_layout(h_pad=1.7)
fig.savefig(os.path.join(plot_dir,
"{}__tuning-profile.svg".format(args.classif)),
bbox_inches='tight',
format='svg')
plt.close()
def plot_tuning_auc(out_list, args):
tune_priors = out_list[0][0]['Clf'].tune_priors
fig, axarr = plt.subplots(figsize=(17, 1 + 7 * len(tune_priors)),
nrows=len(tune_priors),
ncols=1,
squeeze=False)
use_cohs = sorted(set(coh for _, _, (coh, _) in out_list))
coh_mrks = dict(zip(use_cohs, use_marks[:len(use_cohs)]))
use_genes = sorted(set(gene for _, _, (_, gene) in out_list))
gene_clrs = dict(
zip(use_genes, sns.color_palette("muted", n_colors=len(use_genes))))
for ax, (par_name, tune_distr) in zip(axarr.flatten(), tune_priors):
if detect_log_distr(tune_distr):
par_fnc = np.log10
plt_xmin = 2 * np.log10(tune_distr[0]) - np.log10(tune_distr[1])
plt_xmax = 2 * np.log10(tune_distr[-1]) - np.log10(tune_distr[-2])
else:
par_fnc = lambda x: x
plt_xmin = 2 * tune_distr[0] - tune_distr[1]
plt_xmax = 2 * tune_distr[-1] - tune_distr[-2]
for out_dict, (pheno_dict, auc_list), (coh, gene) in out_list:
tune_df = out_dict['Tune'].loc[:, (slice(None), par_name)]
for mtype, (all_val, iso_val) in tune_df.iterrows():
ax.scatter(par_fnc(all_val),
auc_list.loc[mtype, 'All'],
marker=coh_mrks[coh],
s=551 * np.mean(pheno_dict[mtype]),
c=gene_clrs[gene],
alpha=0.23)
ax.scatter(par_fnc(iso_val),
auc_list.loc[mtype, 'Iso'],
marker=coh_mrks[coh],
s=551 * np.mean(pheno_dict[mtype]),
c=gene_clrs[gene],
alpha=0.23)
ax.set_xlim(plt_xmin, plt_xmax)
ax.set_ylim(0.48, 1.02)
ax.tick_params(labelsize=19)
ax.set_xlabel('Tuned {} Value'.format(par_name),
fontsize=27,
weight='semibold')
ax.axhline(y=1.0, color='black', linewidth=2.1, alpha=0.37)
ax.axhline(y=0.5,
color='#550000',
linewidth=2.7,
linestyle='--',
alpha=0.29)
fig.text(-0.01,
0.5,
'Aggregate AUC',
ha='center',
va='center',
fontsize=27,
weight='semibold',
rotation='vertical')
plt.tight_layout(h_pad=1.7)
fig.savefig(os.path.join(plot_dir,
"{}__tuning-auc.svg".format(args.classif)),
bbox_inches='tight',
format='svg')
plt.close()
def plot_tuning_grid(out_list, args):
par_count = len(use_clf.tune_priors)
fig, axarr = plt.subplots(figsize=(0.5 + 7 * par_count, 7 * par_count),
nrows=par_count,
ncols=par_count)
coh_vec = reduce(add, [[out_path[0]] * infer_df['Iso'].shape[0] * 2
for out_path, infer_df, _, in out_list])
use_cohs = sorted(set(coh_vec))
mark_vec = [use_marks[use_cohs.index(coh)] for coh in coh_vec]
gene_vec = reduce(add, [[out_path[1]] * infer_df['Iso'].shape[0] * 2
for out_path, infer_df, _, in out_list])
use_genes = sorted(set(gene_vec))
gene_clrs = sns.color_palette("muted", n_colors=len(use_genes))
clr_vec = [gene_clrs[use_genes.index(gn)] for gn in gene_vec]
size_vec = np.concatenate([
np.repeat([np.sum(stat_dict[mcomb]) for mcomb in tune_df.index], 2)
for (_, _, tune_df), (stat_dict, _) in zip(out_list, score_list)
])
size_vec = 341 * size_vec / np.max(size_vec)
par_vals = {
par_name: np.concatenate([
tune_df.loc[:, (slice(None), par_name)].values.flatten()
for (_, _, tune_df) in out_list
])
for par_name, _ in use_clf.tune_priors
}
auc_vals = np.concatenate([
auc_df.loc[tune_df.index].values.flatten()
for (_, _, tune_df), (_, auc_df) in zip(out_list, score_list)
])
auc_clrs = [auc_cmap(auc_val) for auc_val in auc_vals]
for i, (par_name, tune_distr) in enumerate(use_clf.tune_priors):
axarr[i, i].grid(False)
if detect_log_distr(tune_distr):
use_distr = [np.log10(par_val) for par_val in tune_distr]
par_lbl = par_name + '\n(log-scale)'
else:
use_distr = tune_distr
par_lbl = par_name
distr_diff = np.array(use_distr[-1]) - np.array(use_distr[0])
plt_min = use_distr[0] - distr_diff / 9
plt_max = use_distr[-1] + distr_diff / 9
axarr[i, i].set_xlim(plt_min, plt_max)
axarr[i, i].set_ylim(plt_min, plt_max)
axarr[i, i].text((plt_min + plt_max) / 2, (plt_min + plt_max) / 2,
par_lbl,
ha='center',
fontsize=31,
weight='semibold')
for par_val in use_distr:
axarr[i, i].axhline(y=par_val,
color='#116611',
ls='--',
linewidth=1.9,
alpha=0.23)
axarr[i, i].axvline(x=par_val,
color='#116611',
ls='--',
linewidth=1.9,
alpha=0.23)
for (i, (par_name1, tn_distr1)), (j, (par_name2, tn_distr2)) in combn(
enumerate(use_clf.tune_priors), 2):
if detect_log_distr(tn_distr1):
use_vals1 = np.log10(par_vals[par_name1])
distr_diff = np.log10(np.array(tn_distr1[-1]))
distr_diff -= np.log10(np.array(tn_distr1[0]))
plt_ymin = np.log10(tn_distr1[0]) - distr_diff / 9
plt_ymax = np.log10(tn_distr1[-1]) + distr_diff / 9
else:
use_vals1 = par_vals[par_name1]
distr_diff = tn_distr1[-1] - tn_distr1[0]
plt_ymin = tn_distr1[0] - distr_diff / 9
plt_ymax = tn_distr1[-1] + distr_diff / 9
if detect_log_distr(tn_distr2):
use_vals2 = np.log10(par_vals[par_name2])
distr_diff = np.log10(np.array(tn_distr2[-1]))
distr_diff -= np.log10(np.array(tn_distr2[0]))
plt_xmin = np.log10(tn_distr2[0]) - distr_diff / 9
plt_xmax = np.log10(tn_distr2[-1]) + distr_diff / 9
else:
use_vals2 = par_vals[par_name2]
distr_diff = tn_distr2[-1] - tn_distr2[0]
plt_xmin = tn_distr2[0] - distr_diff / 9
plt_xmax = tn_distr2[-1] + distr_diff / 9
use_vals1 += np.random.normal(0, (plt_ymax - plt_ymin) /
(len(tn_distr1) * 11), auc_vals.shape[0])
use_vals2 += np.random.normal(0, (plt_xmax - plt_xmin) /
(len(tn_distr2) * 11), auc_vals.shape[0])
for use_val2, use_val1, mark_val, size_val, auc_val in zip(
use_vals2, use_vals1, mark_vec, size_vec, auc_clrs):
axarr[i, j].scatter(use_val2,
use_val1,
marker=mark_val,
s=size_val,
c=auc_val,
alpha=0.35,
edgecolor='black')
for use_val1, use_val2, mark_val, size_val, gene_val in zip(
use_vals1, use_vals2, mark_vec, size_vec, clr_vec):
axarr[j, i].scatter(use_val1,
use_val2,
marker=mark_val,
s=size_val,
c=gene_val,
alpha=0.35,
edgecolor='black')
axarr[i, j].set_xlim(plt_xmin, plt_xmax)
axarr[i, j].set_ylim(plt_ymin, plt_ymax)
axarr[j, i].set_ylim(plt_xmin, plt_xmax)
axarr[j, i].set_xlim(plt_ymin, plt_ymax)
plt.tight_layout()
fig.savefig(os.path.join(plot_dir,
"{}__tuning-grid.svg".format(args.classif)),
bbox_inches='tight',
format='svg')
plt.close()