@brainpilot/skills 0.0.6

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

@@ -0,0 +1,109 @@
+import os
+import numpy as np
+from matplotlib import pyplot as plt
+from matplotlib import gridspec as gridspec
+
+
+data_ablation = {
+    'methods': [
+        r'CellSpliceNet',
+        r'ViT',
+        r'SpliceFinder',
+        r'Pangolin',
+        r'SpliceTransformer',
+        r'SpliceAI',
+        r'ESM2',
+    ],
+    'colors': ['#0F4D92', "#F09F97", "#F1B3AC", "#EFBEB8", "#F0CDC8", "#F3D9D8", '#FCEEED'],
+    'metrics': [r'Spearman correlation', r'Pearson correlation', r'R$^2$ score'],
+    'result': {
+        # r'Spearman correlation': np.array([0.88, 0.81, 0.80, 0.79, 0.77, 0.71, 0.606]),
+        # r'Pearson correlation': np.array([0.88, 0.81, 0.80, 0.79, 0.77, 0.72, 0.613]),
+        # r'R$^2$ score': np.array([0.77, 0.66, 0.64, 0.62, 0.59, 0.52, 0.369]),
+        r'Spearman correlation': np.array([
+            [0.88, 0.88, 0.88],  # TODO: update!
+            [0.81, 0.81, 0.81],  # TODO: update!
+            [0.806, 0.793, 0.794],
+            [0.792, 0.785, 0.788],
+            [0.765, 0.765, 0.767],
+            [0.714, 0.716, 0.706],
+            [0.598, 0.625, 0.594],
+        ]),
+        r'Pearson correlation': np.array([
+            [0.88, 0.88, 0.88],  # TODO: update!
+            [0.81, 0.81, 0.81],  # TODO: update!
+            [0.812, 0.798, 0.801],
+            [0.792, 0.785, 0.788],
+            [0.765, 0.765, 0.766],
+            [0.722, 0.722, 0.714],
+            [0.604, 0.631, 0.605],
+        ]),
+        r'R$^2$ score': np.array([
+            [0.77, 0.77, 0.77],  # TODO: update!
+            [0.66, 0.66, 0.66],  # TODO: update!
+            [0.658, 0.632, 0.642],
+            [0.633, 0.615, 0.627],
+            [0.585, 0.585, 0.586],
+            [0.518, 0.520, 0.507],
+            [0.359, 0.384, 0.364],
+        ]),
+    }
+}
+
+def is_dark(color_in_hex, threshold=128):
+    color = color_in_hex.lstrip('#')
+    r = int(color[0:2], 16)
+    g = int(color[2:4], 16)
+    b = int(color[4:6], 16)
+
+    luminance = 0.299*r + 0.587*g + 0.114*b
+    return luminance < threshold
+
+
+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
+
+    fig = plt.figure(figsize=(45, 12))
+
+    gs = gridspec.GridSpec(1, 3)
+
+    for metric_idx, metric_name in enumerate(data_ablation['metrics']):
+        ax = fig.add_subplot(gs[metric_idx])
+
+        num_methods = len(data_ablation['methods'])
+        bars = ax.bar(
+            np.arange(num_methods),
+            data_ablation['result'][metric_name].mean(axis=1),
+            yerr=data_ablation['result'][metric_name].std(axis=1),
+            error_kw={
+                'elinewidth': 2,   # thickness of vertical error bar line
+                'capthick': 2,     # thickness of caps
+                'capsize': 15      # length of caps
+            },
+            color=data_ablation['colors'],
+            label=data_ablation['methods'],
+        )
+
+        for i, (bar, value) in enumerate(zip(bars, data_ablation['result'][metric_name].mean(axis=1))):
+            textcolor = 'white' if is_dark(data_ablation['colors'][i]) else 'black'
+            ax.text(bar.get_x() + bar.get_width()/2, bar.get_height() - 0.10,
+                f'{value:.2f}', ha='center', va='bottom', fontsize=32, color=textcolor)
+
+        ax.set_ylabel(metric_name, fontsize=54, labelpad=12)
+        ymax = np.max(data_ablation['result'][metric_name])
+        ax.set_ylim([0.0, ymax + 0.5])
+        ax.set_xticks([])
+        ax.set_yticks([0.00, 0.25, 0.50, 0.75, 1.00])
+        ax.tick_params(axis='y', labelsize=36, length=10, width=2)
+
+        ax.legend(bbox_to_anchor=(0.02, 1.08), loc='upper left', fontsize=38, frameon=False, ncols=2, columnspacing=0.6)
+
+    fig.tight_layout(pad=2)
+
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/comparison.png', dpi=300)
+    plt.close(fig)

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_Cflows/diffusion_swiss_roll.py

@@ -0,0 +1,97 @@
+import numpy as np
+import matplotlib.pyplot as plt
+from scipy.spatial.distance import pdist, squareform
+
+# Generate Swiss Roll data
+def generate_swiss_roll_2d(n_samples=80, noise=0.1):
+    t = 1.5 * np.pi * (1 + 2 * np.random.rand(n_samples))
+    x = t * np.cos(t)
+    z = t * np.sin(t)
+
+    # Add some noise
+    x += noise * np.random.randn(n_samples)
+    z += noise * np.random.randn(n_samples)
+
+    return x, z, t
+
+# Compute diffusion matrix (transition probabilities)
+def compute_diffusion_matrix(x, z, t, sigma=1.0):
+    # Order points by manifold parameter t for better matrix visualization
+    sorted_indices = np.argsort(t)
+    x_sorted = x[sorted_indices]
+    z_sorted = z[sorted_indices]
+    t_sorted = t[sorted_indices]
+
+    # Compute pairwise distances in ambient space
+    points = np.column_stack([x_sorted, z_sorted])
+    distances = squareform(pdist(points))
+
+    # Compute manifold distances (along parameter t)
+    t_distances = np.abs(t_sorted[:, None] - t_sorted[None, :])
+
+    # Combine spatial and manifold distances (emphasize manifold structure)
+    combined_distances = distances + 0.5 * t_distances
+
+    # Gaussian kernel for transition probabilities
+    P = np.exp(-combined_distances**2 / (2 * sigma**2))
+
+    # Make matrix sparser by thresholding small values
+    P[P < 0.01] = 0
+
+    # Normalize rows to make it a proper transition matrix
+    row_sums = P.sum(axis=1)
+    row_sums[row_sums == 0] = 1  # Avoid division by zero
+    P = P / row_sums[:, None]
+
+    return P, sorted_indices
+
+if __name__ == '__main__':
+    # Create two subplots
+    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(16, 8))
+
+    # Generate swiss roll point cloud (smaller for visualization clarity)
+    x, z, t = generate_swiss_roll_2d(n_samples=500, noise=0.5)
+
+    # Compute diffusion matrix
+    P, sorted_indices = compute_diffusion_matrix(x, z, t, sigma=2)
+
+    # Left plot: Diffusion Matrix
+    im = ax1.imshow(P, cmap='Reds', aspect='equal', origin='upper')
+    ax1.axis('off')
+    ax1.set_facecolor('white')
+
+    # Right plot: Swiss Roll with probability-weighted connections
+    # Use original (unsorted) coordinates for the swiss roll plot
+    x_orig, z_orig, t_orig = x, z, t
+
+    # Draw line segments between points with opacity = transition probability
+    threshold = 0.02  # Only draw lines above this probability threshold
+
+    for i in range(len(x_orig)):
+        for j in range(i+1, len(x_orig)):
+            # Find corresponding indices in sorted matrix
+            orig_i_in_sorted = np.where(sorted_indices == i)[0][0]
+            orig_j_in_sorted = np.where(sorted_indices == j)[0][0]
+
+            # Get transition probability from matrix
+            prob = max(P[orig_i_in_sorted, orig_j_in_sorted], P[orig_j_in_sorted, orig_i_in_sorted])
+
+            if prob > threshold:
+                ax2.plot([x_orig[i], x_orig[j]], [z_orig[i], z_orig[j]],
+                        color='black', linewidth=2, alpha=prob*2, zorder=1)
+
+    # Plot the swiss roll points on top
+    scatter = ax2.scatter(x_orig, z_orig, c=t_orig, cmap='viridis', s=100,
+                          alpha=0.5, edgecolors='white', linewidth=1, zorder=2)
+
+    # Styling for swiss roll plot
+    ax2.set_aspect('equal')
+    ax2.axis('off')
+    ax2.set_facecolor('white')
+
+    # Clean overall styling
+    fig.patch.set_facecolor('white')
+    plt.subplots_adjust(left=0, right=1, top=1, bottom=0, wspace=0.02)
+
+    plt.savefig('figures/diffusion_swiss_roll.png', dpi=300, bbox_inches='tight',
+                facecolor='white', edgecolor='none', pad_inches=1)

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_Cflows/plot_comparison_Ablation.py

@@ -0,0 +1,64 @@
+import os
+import numpy as np
+from matplotlib import pyplot as plt
+
+
+data_comparison_Ablation = {
+    'methods': [r'OT-CFM', r'SB-CFM', r'SF2M', r'Cflows (w/o growth + energy)', r'Cflows (w/o growth)', r'Cflows'],
+    'colors': ['#AADCA9', '#8BCF8B', '#E9A6A1', '#B8C9E5', '#7097CA', '#3775BA'],
+    'metrics': [r'RMSE$\downarrow$', r'MAE$\downarrow$', r'PCC$\uparrow$', r'SCC$\uparrow$'],
+    'mean': {
+        r'RMSE$\downarrow$': np.array([0.94, 1.05, 0.99, 0.89, 0.75, 0.62]),
+        r'MAE$\downarrow$': np.array([0.75, 0.85, 0.78, 0.70, 0.59, 0.48]),
+        r'PCC$\uparrow$': np.array([0.53, 0.49, 0.55, 0.58, 0.65, 0.72]),
+        r'SCC$\uparrow$': np.array([0.50, 0.47, 0.52, 0.55, 0.68, 0.70]),
+    },
+    'std': {
+        r'RMSE$\downarrow$': np.array([0.08, 0.09, 0.09, 0.07, 0.06, 0.05]),
+        r'MAE$\downarrow$': np.array([0.07, 0.08, 0.07, 0.06, 0.05, 0.04]),
+        r'PCC$\uparrow$': np.array([0.04, 0.02, 0.01, 0.03, 0.02, 0.02]),
+        r'SCC$\uparrow$': np.array([0.03, 0.02, 0.03, 0.03, 0.03, 0.01]),
+    },
+}
+
+
+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
+
+    fig = plt.figure(figsize=(35, 7))
+
+    num_methods = len(data_comparison_Ablation['methods'])
+    for metric_idx, metric_name in enumerate(data_comparison_Ablation['metrics']):
+        ax = fig.add_subplot(1, 5, metric_idx + 1)
+
+        ax.bar(
+            np.arange(num_methods),
+            data_comparison_Ablation['mean'][metric_name],
+            yerr=data_comparison_Ablation['std'][metric_name],
+            capsize=8,
+            error_kw={'capthick': 2},
+            color=data_comparison_Ablation['colors'],
+            label=data_comparison_Ablation['methods'],
+        )
+
+        if metric_idx == 0:
+            handles, labels = ax.get_legend_handles_labels()
+
+        ax.set_xticks([])
+        ax.set_ylabel(data_comparison_Ablation['metrics'][metric_idx], fontsize=36, labelpad=12)
+        ax.ticklabel_format(axis='y', style='sci', scilimits=(0, 0))
+
+    ax = fig.add_subplot(1, 5, 5)
+    ax.legend(handles, labels, fontsize=30, loc='lower left', frameon=False)
+    ax.set_axis_off()
+
+    fig.tight_layout(pad=2)
+
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/figX_comparison_Ablation.png', dpi=300)
+    fig.savefig('./figures/figX_comparison_Ablation.pdf', dpi=300)
+    plt.close(fig)

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_Cflows/plot_comparison_GeneRegulatory.py

@@ -0,0 +1,74 @@
+import os
+import numpy as np
+from matplotlib import pyplot as plt
+
+
+data_comparison_GeneRegulatory = {
+    'methods': [r'OCE', r'PC', r'mTE', r'mMI', r'NRI', r'DCRNN', r'GTS', r'NIR', r'GC (ours)'],
+    'colors': ['#D0A3A3', '#EFE7B1', '#F4C2C2', '#D7C4E2', '#E5C09F', '#A8C6C2', '#B7D3B0', '#F5B5A0', '#3775BA'],
+    'metric': 'Graph Edit Distance',
+    'datasets': [
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (100, 137)',
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (150, 329)',
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (200, 507)',
+    ],
+    'mean': {
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (100, 137)':
+            np.array([138.6, 140.4, 126.4, 51.2, 72.1, 158.14, 215.4, 62.7, 51.2]),
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (150, 329)':
+            np.array([293.4, 317.2, 261.0, 99.8, 106.6, 303.79, 347.2, 86.3, 109.0]),
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (200, 507)':
+            np.array([449.8, 495.6, 397.4, 162.8, 219.8, 508.25, 481.8, 159.2, 158.8]),
+    },
+    'std': {
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (100, 137)':
+            np.array([3.5, 3.9, 2.4, 3.3, 6.2, 8.6, 13.8, 3.2, 3.3]),
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (150, 329)':
+            np.array([2.9, 3.7, 2.2, 4.0, 5.4, 12.4, 19.3, 2.8, 6.4]),
+        r'($|\mathcal{V}|$, $|\mathcal{E}|$) = (200, 507)':
+            np.array([1.1, 6.5, 8.8, 6.2, 13.4, 23.6, 7.0, 11.6, 12.6]),
+    },
+}
+
+
+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
+
+    fig = plt.figure(figsize=(36, 6))
+
+    num_methods = len(data_comparison_GeneRegulatory['methods'])
+    for dataset_idx, dataset_name in enumerate(data_comparison_GeneRegulatory['datasets']):
+        ax = fig.add_subplot(1, 4, dataset_idx + 1)
+
+        ax.bar(
+            np.arange(num_methods),
+            data_comparison_GeneRegulatory['mean'][dataset_name],
+            yerr=data_comparison_GeneRegulatory['std'][dataset_name],
+            capsize=8,
+            error_kw={'capthick': 2},
+            color=data_comparison_GeneRegulatory['colors'],
+            label=data_comparison_GeneRegulatory['methods'],
+        )
+
+        if dataset_idx == 0:
+            handles, labels = ax.get_legend_handles_labels()
+
+        ax.set_xticks([])
+        ax.set_ylabel(data_comparison_GeneRegulatory['metric'], fontsize=36, labelpad=12)
+        ax.set_xlabel(dataset_name, fontsize=36, labelpad=12)
+        ax.ticklabel_format(axis='y', style='sci', scilimits=(0, 0))
+
+    ax = fig.add_subplot(1, 4, 4)
+    ax.legend(handles, labels, fontsize=30, loc='lower left', ncols=2, frameon=False)
+    ax.set_axis_off()
+
+    fig.tight_layout(pad=2)
+
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/fig2_comparison_GeneRegulatory.png', dpi=300)
+    fig.savefig('./figures/fig2_comparison_GeneRegulatory.pdf', dpi=300)
+    plt.close(fig)

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_Cflows/plot_comparison_Trajectory.py

@@ -0,0 +1,74 @@
+import os
+import numpy as np
+from matplotlib import pyplot as plt
+
+
+data_comparison_Trajectory = {
+    'methods': [r'TrajectoryNet', r'OT-CFM', r'SB-CFM', r'BEMIOflow (ours)'],
+    'colors': ['#DDF3DE', '#AADCA9', '#8BCF8B', '#3775BA'],
+    'metrics': ['PHATE Space RMSE', 'Gene Space RMSE', 'Interpolation EMD'],
+    'datasets': ['Bifurcation', 'Cycle', 'Unidirectional'],
+    'mean': {
+        'PHATE Space RMSE': {
+            'Bifurcation': np.array([6.16, 2.98, 2.83, 2.60]) * 1e-3,
+            'Cycle': np.array([2.49, 3.79, 1.58, 0.777]) * 1e-3,
+            'Unidirectional': np.array([8.08, 5.33, 5.67, 3.40]) * 1e-3,
+        },
+        'Gene Space RMSE': {
+            'Bifurcation': np.array([0.150, 0.0842, 0.0835, 0.0773]),
+            'Cycle': np.array([0.151, 0.209, 0.141, 0.118]),
+            'Unidirectional': np.array([0.118, 0.0960, 0.0978, 0.0853]),
+        },
+        'Interpolation EMD': {
+            'Bifurcation': np.array([1.07, 0.495, 0.516, 0.465]) * 1e-2,
+            'Cycle': np.array([0.710, 0.818, 0.526, 0.465]) * 1e-2,
+            'Unidirectional': np.array([1.50, 1.32, 1.28, 0.935]) * 1e-2,
+        },
+    },
+}
+
+
+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
+
+    fig = plt.figure(figsize=(36, 6))
+
+    num_methods = len(data_comparison_Trajectory['methods'])
+    for metric_idx, metric_name in enumerate(data_comparison_Trajectory['metrics']):
+        ax = fig.add_subplot(1, 4, metric_idx + 1)
+        xtick_list = []
+
+        for dataset_idx, dataset_name in enumerate(data_comparison_Trajectory['datasets']):
+
+            ax.bar(
+                np.arange(num_methods) + dataset_idx * (num_methods + 1),
+                data_comparison_Trajectory['mean'][metric_name][dataset_name],
+                color=data_comparison_Trajectory['colors'],
+                label=data_comparison_Trajectory['methods'],
+            )
+
+            xtick_list.append(np.mean(np.arange(num_methods)) + dataset_idx * (num_methods + 1))
+
+            if dataset_idx == 0:
+                handles, labels = ax.get_legend_handles_labels()
+
+        ax.set_xticks(xtick_list)
+        ax.set_xticklabels(data_comparison_Trajectory['datasets'])
+        ax.set_ylabel(data_comparison_Trajectory['metrics'][metric_idx], fontsize=36, labelpad=12)
+        ax.set_xlabel('Dataset', fontsize=36, labelpad=12)
+        ax.ticklabel_format(axis='y', style='sci', scilimits=(0, 0))
+
+    ax = fig.add_subplot(1, 4, 4)
+    ax.legend(handles, labels, fontsize=30, loc='lower left', frameon=False)
+    ax.set_axis_off()
+
+    fig.tight_layout(pad=2)
+
+    os.makedirs('./figures/', exist_ok=True)
+    fig.savefig('./figures/fig2_comparison_Trajectory.png', dpi=300)
+    fig.savefig('./figures/fig2_comparison_Trajectory.pdf', dpi=300)
+    plt.close(fig)

package/skills/13_Visualization/nature-figure/assets/figures4papers/figure_Dispersion/plot_idea.py

@@ -0,0 +1,76 @@
+import os
+import numpy as np
+import matplotlib.pyplot as plt
+
+
+EPSILON = 1e-6
+
+def sample_points_in_ball(num_points, theta_range=2*np.pi):
+    r = np.sqrt(np.random.uniform(0, 0.95, num_points))
+    theta = np.random.uniform(np.pi/2 - theta_range/2, np.pi/2 + theta_range/2, num_points)
+    x = r * np.cos(theta)
+    y = r * np.sin(theta)
+    return np.stack([x, y], axis=1)
+
+def plot_ball_with_points(ax, pts, facecolor):
+    num_points_grid = 512
+
+    xs = np.linspace(-1, 1, num_points_grid)
+    ys = np.linspace(-1, 1, num_points_grid)
+    x, y = np.meshgrid(xs, ys)
+    r2 = x**2 + y**2
+    mask = r2 <= 1.0
+    z = np.zeros_like(x)
+    z[mask] = np.sqrt(1.0 - r2[mask])
+
+    nx, ny, nz = x.copy(), y.copy(), z.copy()
+    norm = np.sqrt(nx**2 + ny**2 + nz**2) + EPSILON
+    nx, ny, nz = nx / norm, ny / norm, nz / norm
+
+    # light from top left.
+    light_dir = np.array([-0.5, 0.5, 0.8])
+    light_dir /= np.linalg.norm(light_dir)
+    intensity = np.maximum(0.0, nx*light_dir[0] + ny*light_dir[1] + nz*light_dir[2])
+    shade = np.clip(-0.5 + 2.0*intensity, 0, 1)
+
+    img = np.ones((num_points_grid, num_points_grid))
+    img[mask] = shade[mask]
+
+    ax.imshow(img, cmap='gray', origin='lower', extent=[-1, 1, -1, 1], vmin=0, vmax=1, alpha=0.3)
+    ax.set_xlim([-2, 2])
+    ax.set_ylim([-2, 2])
+    ax.set_axis_off()
+
+    # add points on sphere.
+    ax.scatter(pts[:, 0], pts[:, 1], s=80, facecolor=facecolor, edgecolor='black', alpha=0.8)
+    for i in range(pts.shape[0]):
+        ax.plot([pts[i, 0], 0], [pts[i, 1], 0],
+                linestyle="--", color="black", linewidth=1, alpha=0.8)
+
+    return ax
+
+
+if __name__ == "__main__":
+    save_path = './figures/idea.png'
+    num_points = 16
+
+    np.random.seed(1)
+    plt.rcParams['text.usetex'] = True
+    plt.rcParams['font.family'] = 'sans-serif'
+    fig = plt.figure(figsize=(18, 6))
+
+    ax = fig.add_subplot(1, 3, 1)
+    pts = sample_points_in_ball(num_points=num_points)
+    plot_ball_with_points(ax, pts, facecolor='#cde5f8')
+
+    ax = fig.add_subplot(1, 3, 2)
+    pts = sample_points_in_ball(num_points=num_points, theta_range=np.pi/4)
+    plot_ball_with_points(ax, pts, facecolor='#6a98cb')
+
+    ax = fig.add_subplot(1, 3, 3)
+    pts = sample_points_in_ball(num_points=num_points)
+    plot_ball_with_points(ax, pts, facecolor='#6a98cb')
+
+    os.makedirs(os.path.dirname(save_path), exist_ok=True)
+    fig.tight_layout(pad=2)
+    fig.savefig(save_path, dpi=300)