@brainpilot/skills 0.0.6

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_ImmunoStruct/plot_bars.py

@@ -0,0 +1,216 @@
+import os
+import numpy as np
+from matplotlib import pyplot as plt
+from raw_data import data_comparison_IEDB, data_ablation_IEDB, data_comparison_Cancer, data_ablation_Cancer
+
+
+def decode_ablation(data_dict):
+    binary_list = data_dict['ablations']
+    component_str = data_dict['components']
+    decoded_list = []
+    for binary_code in binary_list:
+        assert len(binary_code) == len(component_str)
+        decoded_str = []
+        for i, c in enumerate(binary_code):
+            if c == '1':
+                decoded_str.append(component_str[i])
+        decoded_list.append(' + '.join(decoded_str))
+    return decoded_list
+
+
+if __name__ == '__main__':
+    plt.rcParams['font.family'] = 'helvetica'
+    plt.rcParams['font.size'] = 24
+    plt.rcParams['axes.spines.right'] = False
+    plt.rcParams['axes.spines.top'] = False
+    plt.rcParams['axes.linewidth'] = 3
+    plt.rcParams['svg.fonttype'] = 'none'
+
+    fig = plt.figure(figsize=(28, 6))
+
+    ax = fig.add_subplot(1, 4, 1)
+    ax.bar(range(len(data_comparison_IEDB['mean'])),
+           data_comparison_IEDB['mean'][:, 0],
+           yerr=data_comparison_IEDB['std'][:, 0],
+           capsize=5,
+           color=data_comparison_IEDB['colors'],
+           label=data_comparison_IEDB['methods'])
+    handles, labels = ax.get_legend_handles_labels()
+    ax.set_xticks([])
+    ax.set_ylim([0.5, 0.9])
+    ax.set_ylabel(data_comparison_IEDB['metrics'][0], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 2)
+    ax.bar(range(len(data_comparison_IEDB['mean'])),
+           data_comparison_IEDB['mean'][:, 1],
+           yerr=data_comparison_IEDB['std'][:, 1],
+           capsize=5,
+           color=data_comparison_IEDB['colors'])
+    ax.set_xticks([])
+    ax.set_ylim([0.15, 0.75])
+    ax.set_ylabel(data_comparison_IEDB['metrics'][1], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 3)
+    ax.bar(range(len(data_comparison_IEDB['mean'])),
+           data_comparison_IEDB['mean'][:, 2],
+           yerr=data_comparison_IEDB['std'][:, 2],
+           capsize=5,
+           color=data_comparison_IEDB['colors'])
+    ax.set_xticks([])
+    ax.set_ylim([0.18, 0.55])
+    ax.set_ylabel(data_comparison_IEDB['metrics'][2], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 4)
+    ax.legend(handles, labels)
+    ax.set_axis_off()
+
+    fig.tight_layout(pad=2)
+
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/bars_comparison_IEDB.png', dpi=600)
+    plt.close(fig)
+
+
+    fig = plt.figure(figsize=(24, 8))
+
+    ax = fig.add_subplot(1, 3, 1)
+    ax.barh(range(len(data_ablation_IEDB['mean'][:, 0])),
+            data_ablation_IEDB['mean'][:, 0],
+            xerr=data_ablation_IEDB['std'][:, 0],
+            color=[(0.215686, 0.458824, 0.729412, alpha) for alpha in np.linspace(0.2, 1.0, 12)],
+            ecolor='k',
+            capsize=5,
+    )
+
+    ax.set_yticks(range(len(data_ablation_IEDB['ablations'])))
+    ax.set_yticklabels(decode_ablation(data_ablation_IEDB))
+    ax.set_xlim([0.75, 0.9])
+    ax.set_xticks([0.75, 0.8, 0.85, 0.9])
+    ax.set_xticklabels([0.75, 0.8, 0.85, 0.9])
+    ax.set_xlabel(data_ablation_IEDB['metrics'][0], fontsize=32)
+
+    ax = fig.add_subplot(1, 3, 2)
+    ax.barh(range(len(data_ablation_IEDB['mean'][:, 1])),
+            data_ablation_IEDB['mean'][:, 1],
+            xerr=data_ablation_IEDB['std'][:, 1],
+            color=[(0.215686, 0.458824, 0.729412, alpha) for alpha in np.linspace(0.2, 1.0, 12)],
+            ecolor='k',
+            capsize=5,
+    )
+
+    ax.set_yticks([])
+    ax.set_xlim([0.4, 0.72])
+    ax.set_xticks([0.4, 0.5, 0.6, 0.7])
+    ax.set_xticklabels([0.4, 0.5, 0.6, 0.7])
+    ax.set_xlabel(data_ablation_IEDB['metrics'][1], fontsize=32)
+
+    ax = fig.add_subplot(1, 3, 3)
+    ax.barh(range(len(data_ablation_IEDB['mean'][:, 2])),
+            data_ablation_IEDB['mean'][:, 2],
+            xerr=data_ablation_IEDB['std'][:, 2],
+            color=[(0.215686, 0.458824, 0.729412, alpha) for alpha in np.linspace(0.2, 1.0, 12)],
+            ecolor='k',
+            capsize=5,
+    )
+
+    ax.set_yticks([])
+    ax.set_xlim([0.4, 0.55])
+    ax.set_xticks([0.4, 0.45, 0.5, 0.55])
+    ax.set_xticklabels([0.4, 0.45, 0.5, 0.55])
+    ax.set_xlabel(data_ablation_IEDB['metrics'][2], fontsize=32)
+
+    fig.tight_layout(pad=2)
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/bars_ablation_IEDB.png', dpi=600)
+    plt.close(fig)
+
+
+    fig = plt.figure(figsize=(28, 6))
+
+    ax = fig.add_subplot(1, 4, 1)
+    ax.bar(range(len(data_comparison_Cancer['mean'])),
+           data_comparison_Cancer['mean'][:, 0],
+           yerr=data_comparison_Cancer['std'][:, 0],
+           capsize=5,
+           color=data_comparison_Cancer['colors'],
+           label=data_comparison_Cancer['methods'])
+    handles, labels = ax.get_legend_handles_labels()
+    ax.set_xticks([])
+    ax.set_ylim([0.5, 0.82])
+    ax.set_ylabel(data_comparison_Cancer['metrics'][0], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 2)
+    ax.bar(range(len(data_comparison_Cancer['mean'])),
+           data_comparison_Cancer['mean'][:, 1],
+           yerr=data_comparison_Cancer['std'][:, 1],
+           capsize=5,
+           color=data_comparison_Cancer['colors'])
+    ax.set_xticks([])
+    ax.set_ylim([0.16, 0.52])
+    ax.set_ylabel(data_comparison_Cancer['metrics'][1], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 3)
+    ax.bar(range(len(data_comparison_Cancer['mean'])),
+           data_comparison_Cancer['mean'][:, 2],
+           yerr=data_comparison_Cancer['std'][:, 2],
+           capsize=5,
+           color=data_comparison_Cancer['colors'])
+    ax.set_xticks([])
+    ax.set_ylim([0.14, 0.44])
+    ax.set_ylabel(data_comparison_Cancer['metrics'][2], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 4)
+    ax.legend(handles, labels)
+    ax.set_axis_off()
+
+    fig.tight_layout(pad=2)
+
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/bars_comparison_Cancer.png', dpi=600)
+    plt.close(fig)
+
+    fig = plt.figure(figsize=(28, 6))
+
+    items_shown = [6, 4, 0] # transfer + contrastive, transfer, none
+
+    ax = fig.add_subplot(1, 4, 1)
+    ax.bar(range(len(items_shown)),
+           data_ablation_Cancer['mean'][:, 0][items_shown],
+           yerr=data_ablation_Cancer['std'][:, 0][items_shown],
+           capsize=5,
+           color=[(0.215686, 0.458824, 0.729412, alpha) for alpha in [1.0, 0.7, 0.4]],
+           label=['ImmunoStruct', 'No Contrastive Learning',
+                  'No Contrastive Learning &\nNo Transfer Learning'])
+    handles, labels = ax.get_legend_handles_labels()
+    ax.set_xticks([])
+    ax.set_ylim([0.68, 0.80])
+    ax.set_ylabel(data_ablation_Cancer['metrics'][0], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 2)
+    ax.bar(range(len(items_shown)),
+           data_ablation_Cancer['mean'][:, 1][items_shown],
+           yerr=data_ablation_Cancer['std'][:, 1][items_shown],
+           capsize=5,
+           color=[(0.215686, 0.458824, 0.729412, alpha) for alpha in [1.0, 0.7, 0.4]])
+    ax.set_xticks([])
+    ax.set_ylim([0.30, 0.52])
+    ax.set_ylabel(data_ablation_Cancer['metrics'][1], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 3)
+    ax.bar(range(len(items_shown)),
+           data_ablation_Cancer['mean'][:, 2][items_shown],
+           yerr=data_ablation_Cancer['std'][:, 2][items_shown],
+           capsize=5,
+           color=[(0.215686, 0.458824, 0.729412, alpha) for alpha in [1.0, 0.7, 0.4]])
+    ax.set_xticks([])
+    ax.set_ylim([0.29, 0.43])
+    ax.set_ylabel(data_ablation_Cancer['metrics'][2], fontsize=32)
+
+    ax = fig.add_subplot(1, 4, 4)
+    ax.legend(handles, labels)
+    ax.set_axis_off()
+
+    fig.tight_layout(pad=2)
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/bars_ablation_Cancer.png', dpi=600)
+    plt.close(fig)

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_ImmunoStruct/raw_data.py

@@ -0,0 +1,125 @@
+import numpy as np
+
+
+data_comparison_IEDB = {
+    'methods': [r'Prime-2.1', r'NetMHCpan', r'MHCnuggets', r'MHCflurry', r'DeepNeo',
+                r'BigMHC-EL', r'BigMHC-IM', r'BigMHC$_\text{retrained}$', r'ImmunoStruct (ours)'],
+    'colors': ['#CFCECE', '#F4EEAC', '#FBDFE2', '#D9B9D4', '#DAA87C', '#DDF3DE', '#AADCA9', '#8BCF8B', '#3775BA'],
+    'metrics': ['AUROC', 'AUPRC', 'Mean PPVn'],
+    'mean': np.array([
+        [0.538, 0.212, 0.207],
+        [0.537, 0.214, 0.220],
+        [0.546, 0.246, 0.233],
+        [0.577, 0.260, 0.273],
+        [0.767, 0.438, 0.411],
+        [0.588, 0.269, 0.290],
+        [0.684, 0.319, 0.333],
+        [0.793, 0.462, 0.445],
+        [0.882, 0.696, 0.514],
+    ]),
+    'std': np.array([
+        [0.012, 0.031, 0.028 / np.sqrt(5)],
+        [0.027, 0.020, 0.018 / np.sqrt(5)],
+        [0.023, 0.019, 0.031 / np.sqrt(5)],
+        [0.021, 0.026, 0.028 / np.sqrt(5)],
+        [0.032, 0.024, 0.017 / np.sqrt(5)],
+        [0.018, 0.016, 0.024 / np.sqrt(5)],
+        [0.028, 0.048, 0.034 / np.sqrt(5)],
+        [0.013, 0.027, 0.009 / np.sqrt(5)],
+        [0.005, 0.020, 0.021 / np.sqrt(5)],
+    ]),
+}
+
+data_ablation_IEDB = {
+    'ablations': ['10000', '01000', '11000', '01100', '11100', '11110',
+                  '10001', '01001', '11001', '01101', '11101', '11111'],
+    'components': ['Structure', 'Sequence', 'Bchems', 'MMA', 'Transfer Learning'],
+    'metrics': ['AUROC', 'AUPRC', 'Mean PPVn'],
+    'mean': np.array([
+       [0.775, 0.442, 0.433],
+       [0.842, 0.553, 0.430],
+       [0.840, 0.547, 0.444],
+       [0.842, 0.554, 0.419],
+       [0.840, 0.547, 0.440],
+       [0.860, 0.615, 0.462],
+       [0.805, 0.497, 0.414],
+       [0.845, 0.568, 0.427],
+       [0.844, 0.569, 0.440],
+       [0.844, 0.561, 0.442],
+       [0.840, 0.554, 0.441],
+       [0.882, 0.696, 0.514],
+    ]),
+   'std': np.array([
+       [0.012, 0.022, 0.020 / np.sqrt(5)],
+       [0.006, 0.014, 0.009 / np.sqrt(5)],
+       [0.007, 0.024, 0.028 / np.sqrt(5)],
+       [0.006, 0.020, 0.017 / np.sqrt(5)],
+       [0.006, 0.018, 0.032 / np.sqrt(5)],
+       [0.013, 0.045, 0.015 / np.sqrt(5)],
+       [0.006, 0.022, 0.027 / np.sqrt(5)],
+       [0.007, 0.024, 0.017 / np.sqrt(5)],
+       [0.005, 0.014, 0.035 / np.sqrt(5)],
+       [0.006, 0.017, 0.024 / np.sqrt(5)],
+       [0.006, 0.028, 0.029 / np.sqrt(5)],
+       [0.005, 0.020, 0.021 / np.sqrt(5)],
+    ]),
+}
+
+
+data_comparison_Cancer = {
+    'methods': [r'Prime-2.1', r'NetMHCpan', r'MHCnuggets', r'MHCflurry', r'DeepNeo',
+                r'BigMHC-EL', r'BigMHC-IM', r'BigMHC$_\text{retrained}$', r'NeoaPred',
+                r'ImmunoStruct (ours)'],
+    'colors': ['#CFCECE', '#F4EEAC', '#FBDFE2', '#D9B9D4', '#DAA87C', '#DDF3DE', '#AADCA9', '#8BCF8B', "#92E3F9", '#3775BA'],
+    'metrics': ['AUROC', 'AUPRC', 'Mean PPVn'],
+    'mean': np.array([
+        [0.645, 0.259, 0.295],
+        [0.559, 0.210, 0.211],
+        [0.530, 0.210, 0.175],
+        [0.658, 0.304, 0.329],
+        [0.535, 0.261, 0.222],
+        [0.632, 0.248, 0.245],
+        [0.771, 0.373, 0.357],
+        [0.682, 0.310, 0.325],
+        [0.556, 0.267, 0.292],
+        [0.771, 0.433, 0.364],
+    ]),
+    'std': np.array([
+        [0.026, 0.023, 0.029 / np.sqrt(5)],
+        [0.049, 0.037, 0.045 / np.sqrt(5)],
+        [0.014, 0.017, 0.023 / np.sqrt(5)],
+        [0.015, 0.032, 0.030 / np.sqrt(5)],
+        [0.016, 0.053, 0.070 / np.sqrt(5)],
+        [0.034, 0.039, 0.042 / np.sqrt(5)],
+        [0.040, 0.062, 0.036 / np.sqrt(5)],
+        [0.012, 0.020, 0.030 / np.sqrt(5)],
+        [0.016, 0.022, 0.081 / np.sqrt(5)],
+        [0.024, 0.069, 0.127 / np.sqrt(5)],
+    ]),
+}
+
+data_ablation_Cancer = {
+    'coeffs': [[False, 0], [False, 0.001], [False, 0.01], [False, 0.1],
+               [True, 0], [True, 0.001], [True, 0.01], [True, 0.1]],
+    'metrics': ['AUROC', 'AUPRC', 'Mean PPVn'],
+    'mean': np.array([
+       [0.723, 0.391, 0.358],
+       [0.712, 0.370, 0.331],
+       [0.727, 0.405, 0.387],
+       [0.700, 0.413, 0.354],
+       [0.756, 0.426, 0.362],
+       [0.762, 0.418, 0.351],
+       [0.771, 0.433, 0.365],
+       [0.725, 0.406, 0.348],
+    ]),
+   'std': np.array([
+       [0.027, 0.068, 0.096 / np.sqrt(5)],
+       [0.040, 0.077, 0.114 / np.sqrt(5)],
+       [0.037, 0.080, 0.111 / np.sqrt(5)],
+       [0.036, 0.081, 0.109 / np.sqrt(5)],
+       [0.031, 0.068, 0.103 / np.sqrt(5)],
+       [0.024, 0.064, 0.151 / np.sqrt(5)],
+       [0.024, 0.069, 0.127 / np.sqrt(5)],
+       [0.035, 0.079, 0.129 / np.sqrt(5)],
+    ]),
+}

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_RNAGenScape/plot_comparison.py

@@ -0,0 +1,228 @@
+import os
+import numpy as np
+import matplotlib as mpl
+import matplotlib.pyplot as plt
+
+
+summary_label = r'\textit{Improvement}'
+options = ['VAE', 'DDPM', 'LDM', 'FM',
+           'DiffAb', 'IgLM', 'NOS-C', 'NOS-D',
+           'OAE + gradient ascent',
+           'OAE + MCMC',
+           'OAE + hill climbing',
+           'OAE + stochastic hill climbing',
+           r'$\texttt{RNAGenScape}$ \textbf{(ours)}',
+           summary_label]
+colors = ["#cdcdcd", "#767676", "#4d4d4d", "#272727",
+          "#c4ece7", "#ecc4c4", "#ecc4e7", "#ea84dd",
+          "#d5e29b", "#bdd35c", "#9fbc1d", "#8ead03",
+          "#0f4d92"]
+xlabel = np.arange(len(options))
+xticks = []
+results_inference = [0.13, 0.91, 0.74, 5.82, 41.04, 157.57, 0.99, 0.96,
+                     0.50, 10.93, 81.52, 99.66, 0.57]
+results_openvaccine_delta_pos = [-0.23, -0.33, -0.07, -0.34, 0.06, 0.24, 0.09, 0.18,
+                                 -0.01, -0.11, 0.40, -0.12, 0.54]
+results_openvaccine_pct_pos = [42.1, 33.8, 47.5, 32.5, 55.0, 63.1, 54.6, 58.3,
+                               51.0, 45.1, 69.2, 43.5, 77.5]
+results_openvaccine_delta_neg = [-0.23, -0.33, -0.07, -0.34, 0.11, 0.01, -2.25, -0.29,
+                                 -0.19, -0.19, -0.29, -0.15, -2.81]
+results_openvaccine_pct_neg = [57.9, 66.2, 52.5, 67.5, 44.0, 49.6, 90.6, 65.4,
+                               61.3, 57.7, 64.2, 60.0, 97.9]
+results_zebrafish_delta_pos = [-0.01, 0.29, -0.95, 0.32, -0.21, -0.57, 0.03, 0.31,
+                               -1.21, -0.20, 0.76, 0.32, 0.77]
+results_zebrafish_pct_pos = [49.4, 58.2, 22.5, 59.8, 36.3, 32., 51.1, 57.9,
+                             18.0, 44.3, 74.4, 60.3, 75.0]
+results_zebrafish_delta_neg = [-0.01, 0.29, -0.95, 0.32, -0.21, -0.83, -1.07, 0.38,
+                               -0.33, -0.37, -0.87, -0.80, -1.29]
+results_zebrafish_pct_neg = [50.6, 41.8, 77.5, 40.2, 60.8, 74.3, 80.8, 40.2,
+                             66.5, 60.1, 75.2, 73.7, 85.1]
+results_ribosome_delta_pos = [-0.24, -0.11, -0.24, -0.10, -0.05, 0.63, 0.08, -0.09,
+                              0.43, -0.19, 0.53, 0.19, 0.63]
+results_ribosome_pct_pos = [41.7, 45.9, 41.8, 46.4, 45.0, 80.5, 53.6, 46.5,
+                            73.6, 43.0, 83.0, 60.4, 81.4]
+results_ribosome_delta_neg = [-0.24, -0.11, -0.24, -0.10, -0.04, -0.51, 0.10, -0.10,
+                              -0.05, -0.10, -0.06, -0.05, -0.58]
+results_ribosome_pct_neg = [58.3, 54.1, 58.2, 53.6, 54.2, 65.5, 46.0, 53.6,
+                            55.7, 54.2, 56.1, 55.2, 67.8]
+
+
+if __name__ == '__main__':
+    PLOT_DE_NOVO_SPEED = False
+
+    plt.rcParams['text.usetex'] = True
+    plt.rcParams['font.family'] = 'helvetica'
+    plt.rcParams['font.size'] = 16
+    plt.rcParams['axes.spines.right'] = False
+    plt.rcParams['axes.spines.top'] = False
+    plt.rcParams['axes.linewidth'] = 2
+
+    fig = plt.figure(figsize=(9, 5))
+    ax = fig.add_subplot(1, 1, 1)
+
+    if PLOT_DE_NOVO_SPEED:
+        ax.bar(np.arange(len(options)), 1 / np.array(results_inference), color=colors, label=options)
+        ax.hlines(y=1 / results_inference[-1], xmin=-1, xmax=len(options), color=colors[-1], linestyle='--')
+        ax.set_xlim([-1, len(options)])
+        ax.legend(loc='upper right', frameon=False, fontsize=12)
+        ax.set_xticks([])
+        ax.set_ylabel('Inference Throughput ' + r'$\uparrow$' +'\n(samples / ms)', fontsize=18)
+
+        # Add braces.
+        ymin, ymax = ax.get_ylim()
+        ax.set_ylim(ymin - 1, ymax)
+        len_de_novo = 4
+        mid_de_novo = 0 + (len_de_novo - 0) / 2
+        start_opt = len_de_novo + 1
+        end_opt = len(options) - 1
+        mid_opt = (start_opt + end_opt) / 2
+        ax.text(mid_de_novo, ymin - 0.3,
+            r'$\underbrace{\rule{5cm}{0pt}}_{\textrm{\textit{de\ novo}\ generative\ models}}$',
+            ha='center', va='top', fontsize=14)
+        ax.text(mid_opt, ymin - 0.3,
+            r'$\underbrace{\rule{9.2cm}{0pt}}_{\textrm{property\ optimization\ methods}}$',
+            ha='center', va='top', fontsize=14)
+        ax.spines['bottom'].set_position(('data', 0))
+        ax.xaxis.set_ticks_position('bottom')
+        ax.spines['bottom'].set_linewidth(2)
+        ax.spines['left'].set_bounds(0, ymax)
+
+    else:
+        # speed_values = 1 / np.array(results_inference[4:])
+        # options_speed = options[4:-1]
+        # ax.bar(np.arange(len(options_speed)), speed_values, color=colors[4:], label=options_speed)
+        speed_values = 1 / np.array(results_inference[4:8] + [results_inference[-1]])
+        options_speed = options[4:8] + [options[-2]]
+        ax.bar(np.arange(len(options_speed)), speed_values, color=colors[4:8] + [colors[-1]], label=options_speed)
+        ax.set_xlim([-1, len(options_speed)])
+        ax.legend(loc='upper left', frameon=False, fontsize=16, ncol=1)
+        ax.set_xticks([])
+        ax.set_ylabel('Inference Throughput ' + r'$\uparrow$' +'\n(samples / ms)', fontsize=18)
+        ax.set_yscale('log')
+        ymin, ymax = ax.get_ylim()
+        ax.set_ylim(ymin, ymax * 20)
+
+        # Add numbers on top of bars
+        for i, val in enumerate(speed_values):
+            ax.text(i, val * 1.1, f'{val:.3f}', ha='center', va='bottom', fontsize=16)
+
+    fig.tight_layout(pad=1)
+    os.makedirs('./figures', exist_ok=True)
+    fig.savefig('./figures/results_comparison_speed.png', dpi=300)
+    plt.close(fig)
+
+
+    fig = plt.figure(figsize=(20, 9))
+    ax = fig.add_subplot(1, 2, 1)
+    improvements = np.stack([results_openvaccine_delta_pos, results_openvaccine_delta_neg,
+                             results_zebrafish_delta_pos, results_zebrafish_delta_neg,
+                             results_ribosome_delta_pos, results_ribosome_delta_neg], axis=1)
+    base_improvements = improvements[:-1]
+    best_improvements = []
+    for j in range(base_improvements.shape[1]):
+        col = base_improvements[:, j]
+        best_improvements.append(col.max() if j % 2 == 0 else col.min())
+    best_improvements = np.array(best_improvements)
+    denom = np.where(best_improvements == 0, np.nan, best_improvements)
+    improvement_over_best = 100 * (improvements[-1] - best_improvements) / denom
+    improvements = np.vstack([improvements, improvement_over_best])
+    improvements_display = improvements.copy()
+    improvements_display[-1, :] = np.nan
+    n_rows, n_cols = improvements.shape
+    vmin, vmax = improvements[:-1].min(0), improvements[:-1].max(0)
+    cmap_red, cmap_blue = plt.cm.Reds, plt.cm.Blues_r
+    for j in range(n_cols):
+        cmap = cmap_red if j % 2 == 0 else cmap_blue
+        cmap = cmap.copy()
+        cmap.set_bad(color="white")
+        norm = mpl.colors.Normalize(vmin=vmin[j] if j % 2 == 1 else 0, vmax=vmax[j] if j % 2 == 0 else 0)
+        ax.imshow(improvements_display[:, j:j+1], cmap=cmap, norm=norm,
+                aspect="auto", extent=[j - 0.5, j + 0.5, 0, n_rows], origin="lower")
+    for (i, j), val in np.ndenumerate(improvements):
+        if i == n_rows - 1:
+            color = "forestgreen" if val >= 0 else "darkred"
+            label = f"{val:+.1f} \\%"
+            fontsize = 16
+        else:
+            cmap = cmap_red if j % 2 == 0 else cmap_blue
+            norm = mpl.colors.Normalize(vmin=vmin[j] if j % 2 == 1 else 0, vmax=vmax[j] if j % 2 == 0 else 0)
+            r, g, b, _ = cmap(norm(val))
+            lum = 0.299 * r + 0.587 * g + 0.114 * b
+            color = "white" if lum < 0.5 else "black"
+            label = f"{val:.2f}"
+            fontsize = 14
+        ax.text(j, i + 0.5, label, ha="center", va="center", fontsize=fontsize, color=color)
+    ax.set_title('Median change in property', fontsize=32, pad=24)
+    ax.set_xlim(-0.5, n_cols - 0.5)
+    ax.set_xticks(np.arange(n_cols))
+    ax.set_xticklabels([r'$\texttt{OpenVaccine}~(+)$', r'$\texttt{OpenVaccine}~(-)$',
+                        r'$\texttt{Zebrafish}~(+)$', r'$\texttt{Zebrafish}~(-)$',
+                        r'$\texttt{Ribosome}~(+)$', r'$\texttt{Ribosome}~(-)$'],
+                    fontsize=20, rotation=30)
+    ax.tick_params(axis='x', which='both', bottom=False, top=False, length=0)
+    ax.set_yticks(np.arange(n_rows) + 0.5)
+    ax.set_yticklabels(options, rotation=0, fontsize=20, ha='right')
+    for tick in ax.get_yticklabels():
+        if r'$\texttt{RNAGenScape}$' in tick.get_text():
+            tick.set_fontsize(24)
+    ax.set_frame_on(False)
+    ax.invert_yaxis()
+    # n_rows, n_cols = improvements.shape
+    # rect = patches.Rectangle((-0.5, n_rows - 1), n_cols, 1, fill=False, edgecolor='black', linewidth=3)
+    # rect.set_clip_on(False)
+    # ax.add_patch(rect)
+
+    ax = fig.add_subplot(1, 2, 2)
+    percentages = np.stack([results_openvaccine_pct_pos, results_openvaccine_pct_neg,
+                            results_zebrafish_pct_pos, results_zebrafish_pct_neg,
+                            results_ribosome_pct_pos, results_ribosome_pct_neg], axis=1)
+    base_percentages = percentages[:-1]
+    best_percentages = base_percentages.max(0)
+    pct_improvement_over_best = 100 * (percentages[-1] - best_percentages) / best_percentages
+    percentages = np.vstack([percentages, pct_improvement_over_best])
+    percentages_display = percentages.copy()
+    percentages_display[-1, :] = np.nan
+    n_rows, n_cols = percentages.shape
+    vmin, vmax = percentages[:-1].min(0), percentages[:-1].max(0)
+    cmap_red = plt.cm.Reds.copy()
+    cmap_red.set_bad(color="white")
+    for j in range(n_cols):
+        norm = mpl.colors.Normalize(vmin=max(50, vmin[j]), vmax=vmax[j])
+        ax.imshow(percentages_display[:, j:j+1], cmap=cmap_red, norm=norm,
+                aspect="auto", extent=[j - 0.5, j + 0.5, 0, n_rows], origin="lower")
+    for (i, j), val in np.ndenumerate(percentages):
+        if i == n_rows - 1:
+            color = "forestgreen" if val >= 0 else "darkred"
+            label = f"{val:+.1f} \\%"
+            fontsize = 16
+        else:
+            norm = mpl.colors.Normalize(vmin=max(50, vmin[j]), vmax=vmax[j])
+            r, g, b, _ = cmap_red(norm(val))
+            lum = 0.299 * r + 0.587 * g + 0.114 * b
+            color = "white" if lum < 0.5 else "black"
+            label = f"{val:.1f} \\%"
+            fontsize = 14
+        ax.text(j, i + 0.5, label, ha="center", va="center",
+                fontsize=fontsize, color=color)
+    ax.set_title('Success rate', fontsize=32, pad=24)
+    ax.set_xlim(-0.5, n_cols - 0.5)
+    ax.set_xticks(np.arange(n_cols))
+    ax.set_xticklabels([r'$\texttt{OpenVaccine}~(+)$', r'$\texttt{OpenVaccine}~(-)$',
+                        r'$\texttt{Zebrafish}~(+)$', r'$\texttt{Zebrafish}~(-)$',
+                        r'$\texttt{Ribosome}~(+)$', r'$\texttt{Ribosome}~(-)$'],
+                    fontsize=20, rotation=30)
+    ax.tick_params(axis='x', which='both', bottom=False, top=False, length=0)
+    ax.set_yticks([])
+    # ax.set_yticks(np.arange(n_rows) + 0.5)
+    # ax.set_yticklabels(options, fontsize=16, ha='right')
+    ax.set_frame_on(False)
+    ax.invert_yaxis()
+    # rect = patches.Rectangle((-0.5, n_rows - 1), n_cols, 1,
+    #                         fill=False, edgecolor='black', linewidth=3)
+    # rect.set_clip_on(False)
+    # ax.add_patch(rect)
+
+    fig.tight_layout(pad=2)
+    os.makedirs('./figures', exist_ok=True)
+    fig.savefig('./figures/results_comparison_optimization.png', dpi=300)
+    plt.close(fig)

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_RNAGenScape/plot_hole_manifold.py

@@ -0,0 +1,82 @@
+import os
+import numpy as np
+import matplotlib.pyplot as plt
+from matplotlib.colors import LinearSegmentedColormap
+
+def function(x, y):
+    z = 0.6 * np.exp(-((x - 1)**2 + (y + 1)**2))
+    z += 0.5 * np.exp(-((x - 1)**2 + (y - 4)**2))
+    z += 0.3 * np.exp(-((x - 2)**2 + (y - 2)**2))
+    z += 0.2 * np.exp(-((x + 3)**2 + (y + 1)**2))
+    z += 0.3 * np.exp(-((x + 1)**2 + (y + 1)**2))
+    z -= 0.1 * np.exp(-((x + 1)**2 + (y - 2)**2))
+    z += 0.3 * np.exp(-((x + 2)**2 + (y - 2)**2))
+    z += 0.3 * np.exp(-((x + 2)**2 + (y - 1)**2))
+    return z
+
+x = np.linspace(-3, 3, 200)
+y = np.linspace(-3, 3, 200)
+X, Y = np.meshgrid(x, y)
+Z = function(X, Y)
+
+peak_i, peak_j = np.unravel_index(np.argmax(Z), Z.shape)
+x_peak, y_peak = X[peak_i, peak_j], Y[peak_i, peak_j]
+
+rng = np.random.default_rng(42)
+num_patches = 30
+r = 0.3
+r_forbid = 0.9
+pad = r + 0.1
+
+centers = []
+attempts = 0
+while len(centers) < num_patches and attempts < 5000:
+    cx = rng.uniform(x.min()+pad, x.max()-pad)
+    cy = rng.uniform(y.min()+pad, y.max()-pad)
+    if (cx - x_peak)**2 + (cy - y_peak)**2 < r_forbid**2:
+        attempts += 1
+        continue
+    ok = True
+    for px, py in centers:
+        if (cx - px)**2 + (cy - py)**2 < (1.6*r)**2:
+            ok = False
+            break
+    if ok:
+        centers.append((cx, cy))
+    attempts += 1
+
+mask = np.zeros_like(Z, dtype=bool)
+for cx, cy in centers:
+    mask |= (X - cx)**2 + (Y - cy)**2 <= r**2
+
+cmap = LinearSegmentedColormap.from_list(
+    "softgreen", ["#e9f5ec", "#d9f0e1", "#c9e5d3", "#a9cbb8", "#7f9e8a", "#4f5c4f"], N=256
+)
+
+norm = plt.Normalize(Z.min(), Z.max())
+facecolors = cmap(norm(Z))
+facecolors_with_gray = facecolors.copy()
+facecolors_with_gray[mask] = [0.7, 0.7, 0.7, 0.5]
+
+fig = plt.figure(figsize=(14, 6))
+for i, (fc, title) in enumerate([(facecolors, "Smooth Manifold"),
+                                 (facecolors_with_gray, "Manifold with Gray Patches")], 1):
+    ax = fig.add_subplot(1, 2, i, projection="3d")
+    ax.plot_surface(
+        X, Y, Z,
+        facecolors=fc,
+        rstride=4, cstride=4,
+        linewidth=0.05, edgecolor="k",
+        antialiased=True, shade=False, alpha=0.95
+    )
+    ax.set_title(title, fontsize=14)
+    ax.set_xticks([]); ax.set_yticks([]); ax.set_zticks([])
+    for a in (ax.xaxis, ax.yaxis, ax.zaxis):
+        a.pane.set_visible(False)
+        a.line.set_color((0,0,0,0))
+    ax.set_box_aspect([1, 1, 0.5])
+    ax.view_init(elev=20, azim=50)
+
+fig.tight_layout(pad=2)
+os.makedirs("./figures", exist_ok=True)
+fig.savefig("./figures/manifold_holes.png", dpi=300)