#!/usr/bin/env python
# Copyright 2016-2023 Biomedical Imaging Group Rotterdam, Departments of
# Medical Informatics and Radiology, Erasmus MC, Rotterdam, The Netherlands
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
import numpy as np
import tikzplotlib
[docs]def manhattan_importance(values, labels, feature_labels,
output_png=None, output_tex=None,
mapping=None, threshold_annotated=0.05):
# Assume labels are numeric and sorted
f = plt.figure(figsize=(20, 10))
# Generate X-positions
positions = np.arange(len(values))
# Initialize several objects
values = np.asarray(values)
unique_labels = set(list(labels))
n_labels = len(unique_labels)
colormap = ['#7dcfe2', '#4b78b5', 'darkgrey', 'dimgray'] * n_labels
for lnum, i in enumerate(unique_labels):
# Shift positions for each class
for pnum in range(len(positions)):
if labels[pnum] == i:
positions[pnum] += lnum
# NOTE: use lnum to leave space between labels for vlines
# Only take first set of points corresponding to first label
plot_positions = [p for p, l in zip(positions, labels) if l == i]
plot_values = [v for v, l in zip(values, labels) if l == i]
plt.scatter(plot_positions, plot_values, c=colormap[lnum])
# Set a line after each label ends
label_previous = labels[0]
pos_previous = positions[0]
color_end = list()
vlines = list()
# NOTE: leave space between groups to plot vline
for i, p in zip(labels, positions):
if i != label_previous:
# New color starts here
color_end.append((pos_previous + p) / 2.0)
label_previous = i
pos_previous = p
# Plot vlines just between classes
vlines.append(p - 1)
# Add the last color and line
color_end.append((pos_previous + p) / 2.0)
# Decide y-limits
ymax = np.max(values)
for i in range(0, 100):
if 10**(-i) < ymax:
ymaxlim = i - 1
break
ymin = np.min(values) # might be zero
yposmin = max(np.min(values[values > 0]), np.finfo(values.dtype).eps)
for i in range(0, 100):
if 10**(-i) < (ymin if ymin > 0 else yposmin):
yminlim = i
break
# Set several figure lay-out options
plt.gca().invert_yaxis()
if ymin > 0:
plt.yscale('log')
plt.ylim((10**-ymaxlim, 10**-yminlim))
else:
plt.yscale('symlog', linthresh=10**-yminlim)
plt.ylim((10**-ymaxlim, 0.0))
plt.xlim((0, max(positions)))
plt.yticks([10**-i for i in range(ymaxlim, yminlim + 1)],
[f'10-{i}' for i in range(ymaxlim, yminlim + 1)])
if mapping is None:
# Use raw labels (=numbers) as ticks
plt.xticks(color_end, np.arange(len(color_end)) + 1, size=16)
else:
xticks = [mapping[i] for i in unique_labels]
plt.xticks(color_end, xticks, size=8)
plt.vlines(vlines, 10**-ymaxlim, 10**-yminlim, linestyles='dotted', linewidth=0.3)
if threshold_annotated > 10**-yminlim:
y_value_annotated = threshold_annotated
plt.hlines(threshold_annotated, 0, max(positions), linestyles='dashed', linewidth=1, color='magenta')
else:
y_value_annotated = 10**-yminlim
plt.hlines(10**-yminlim, 0, max(positions), linestyles='dashed', linewidth=1, color='magenta')
plt.annotate(f'p={round(threshold_annotated, 5)}',
(1, y_value_annotated),
xytext=(1, y_value_annotated*0.95), size=8, color='magenta')
if 0.05 > 10**-yminlim:
plt.hlines(0.05, 0, max(positions), linestyles='dashed', linewidth=1, color='magenta')
plt.annotate('p=0.05',
(1, 0.05),
xytext=(1, 0.05*0.95), size=8, color='magenta')
plt.xlabel("Feature groups", size=12)
plt.ylabel("P-value Mann-Whitney U", size=12)
# Annotate points above the threshold
offset = len(values) / 200
offset = np.clip(offset, 0.1, 100)
annotated_values = [v for v in values if v < threshold_annotated]
annotated_pos = [p for p, v in zip(positions, values) if v < threshold_annotated]
annotated_labels = [p for p, v in zip(feature_labels, values) if v < threshold_annotated]
y_offset = -0.1
for x, y, text in zip(annotated_pos, annotated_values, annotated_labels):
plt.annotate(text,
(x, y),
xytext=(x + offset, y * (1 - y_offset)), size=6)
y_offset = y_offset * -1
plt.gca().spines['right'].set_color('none')
plt.gca().spines['top'].set_color('none')
if output_png is not None:
plt.savefig(output_png, bbox_inches='tight', pad_inches=0)
print(f"Plot saved as {output_png}!")
if output_tex is not None:
tikzplotlib.save(output_tex)
print(f"Plot saved as {output_tex}!")
return f