Upload 9 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+thermomutdb.json filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,534 @@
+
+import gradio as gr
+import matplotlib.pyplot as plt
+import io
+import base64
+import shap
+import numpy as np
+import pandas as pd
+import torch
+import torch.nn as nn
+import requests
+import re
+import json
+import pickle # Added for loading .pkl and .json files
+import os # Added for path checking
+
+# --- Device Configuration ---
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# --- Model Architecture (MultiTaskMLP) ---
+class MultiTaskMLP(nn.Module):
+    def __init__(self, input_dim, num_classes, dropout_rate=0.3):
+        super(MultiTaskMLP, self).__init__()
+        self.shared_layers = nn.Sequential(
+            nn.Linear(input_dim, 256),
+            nn.BatchNorm1d(256),
+            nn.ReLU(),
+            nn.Dropout(dropout_rate),
+            nn.Linear(256, 128),
+            nn.BatchNorm1d(128),
+            nn.ReLU(),
+            nn.Dropout(dropout_rate)
+        )
+        self.regression_head = nn.Sequential(
+            nn.Linear(128, 64),
+            nn.BatchNorm1d(64),
+            nn.ReLU(),
+            nn.Dropout(dropout_rate),
+            nn.Linear(64, 1)
+        )
+        self.classification_head = nn.Sequential(
+            nn.Linear(128, 64),
+            nn.BatchNorm1d(64),
+            nn.ReLU(),
+            nn.Dropout(dropout_rate),
+            nn.Linear(64, num_classes)
+        )
+
+    def forward(self, x):
+        shared_features = self.shared_layers(x)
+        ddg_output = self.regression_head(shared_features)
+        effect_output = self.classification_head(shared_features)
+        return ddg_output, effect_output
+
+# --- SHAP Wrapper Classes ---
+class RegressionHeadWrapper(nn.Module):
+    def __init__(self, original_model):
+        super().__init__()
+        self.original_model = original_model
+
+    def forward(self, x):
+        ddg_output, _ = self.original_model(x)
+        return ddg_output
+
+class ClassificationHeadWrapper(nn.Module):
+    def __init__(self, original_model):
+        super().__init__()
+        self.original_model = original_model
+
+    def forward(self, x):
+        _, effect_output = self.original_model(x)
+        return effect_output
+
+# --- Helper Function: plot_to_base64 ---
+def plot_to_base64(fig):
+    buffer = io.BytesIO()
+    fig.savefig(buffer, format='png', bbox_inches='tight')
+    buffer.seek(0)
+    image_base64 = base64.b64encode(buffer.read()).decode('utf-8')
+    plt.close(fig)
+    return f"<img src='data:image/png;base64,{image_base64}'>"
+
+# --- Helper Function: AA_PROPERTIES and get_aa_property_comparison ---
+# AA_PROPERTIES is now loaded from global_vars.json in the main script
+def get_aa_property_comparison(original_aa, mutant_aa):
+    # global AA_PROPERTIES is assumed to be loaded after global_vars.json
+    if original_aa not in AA_PROPERTIES or mutant_aa not in AA_PROPERTIES:
+        return "Invalid amino acid code(s) for comparison."
+    results = {}
+    for prop, _ in AA_PROPERTIES['A'].items():
+        orig_val = AA_PROPERTIES[original_aa].get(prop, 0)
+        mut_val = AA_PROPERTIES[mutant_aa].get(prop, 0)
+        change = mut_val - orig_val
+        results[prop] = {
+            'original': orig_val,
+            'mutant': mut_val,
+            'change': f'{change:.2f}'
+        }
+    return results
+
+# --- Helper Function: get_top_features_by_shap ---
+def get_top_features_by_shap(explainer_obj, background_data_tensor, feature_names, top_n=10, task_type='regression'):
+    if task_type == 'regression':
+        shap_values_background = explainer_obj.shap_values(background_data_tensor)
+        if isinstance(shap_values_background, list):
+            shap_values_background = np.array(shap_values_background).squeeze()
+        elif shap_values_background.ndim == 3:
+            shap_values_background = shap_values_background.squeeze(axis=-1)
+    elif task_type == 'classification':
+        shap_values_background_raw = explainer_obj.shap_values(background_data_tensor)
+        if isinstance(shap_values_background_raw, list):
+            abs_shap_values = np.mean(np.abs(np.array(shap_values_background_raw)), axis=0)
+        elif shap_values_background_raw.ndim == 3:
+            abs_shap_values = np.mean(np.abs(shap_values_background_raw), axis=-1)
+        else:
+            abs_shap_values = np.abs(shap_values_background_raw)
+        shap_values_background = abs_shap_values
+    else:
+        raise ValueError("task_type must be 'regression' or 'classification'")
+    mean_abs_shap = np.mean(np.abs(shap_values_background), axis=0)
+    feature_importance = pd.DataFrame({
+        'Feature': feature_names,
+        'Mean_Abs_SHAP': mean_abs_shap
+    })
+    top_features = feature_importance.sort_values(by='Mean_Abs_SHAP', ascending=False).head(top_n)
+    return top_features.to_dict(orient='records')
+
+# --- Helper Function: plot_local_feature_importance ---
+def plot_local_feature_importance(shap_values, feature_names, title="Local Feature Importance", top_n=10):
+    if shap_values.ndim > 1:
+        shap_values = shap_values.flatten()
+    feature_importance = pd.DataFrame({
+        'Feature': feature_names,
+        'SHAP_Value': shap_values
+    })
+    feature_importance['Abs_SHAP'] = np.abs(feature_importance['SHAP_Value'])
+    feature_importance = feature_importance.sort_values(by='Abs_SHAP', ascending=False).head(top_n)
+    feature_importance = feature_importance.sort_values(by='SHAP_Value', ascending=True)
+    fig, ax = plt.subplots(figsize=(10, 6))
+    colors = ['red' if x < 0 else 'blue' for x in feature_importance['SHAP_Value']]
+    ax.barh(feature_importance['Feature'], feature_importance['SHAP_Value'], color=colors)
+    ax.set_xlabel('SHAP Value (Impact on Model Output)')
+    ax.set_ylabel('Feature')
+    ax.set_title(title)
+    fig.tight_layout()
+    return plot_to_base64(fig)
+
+# --- Helper Function: get_mutation_context_summary ---
+def get_mutation_context_summary(input_data_df):
+    summary_html = "<h3>Mutation Structural Context:</h3>"
+    sst = input_data_df['sst'].iloc[0] if 'sst' in input_data_df.columns else 'N/A'
+    rsa_val = input_data_df['rsa'].iloc[0] if 'rsa' in input_data_df.columns else 'N/A'
+    res_depth_val = input_data_df['res_depth'].iloc[0] if 'res_depth' in input_data_df.columns else 'N/A'
+
+    summary_html += f"<p><b>Secondary Structure (sst):</b> {sst}</p>"
+    if isinstance(rsa_val, (int, float)):
+        summary_html += f"<p><b>Relative Solvent Accessibility (rsa):</b> {rsa_val:.2f} (0=Buried, 1=Exposed)</p>"
+    else:
+        summary_html += f"<p><b>Relative Solvent Accessibility (rsa):</b> {rsa_val}</p>"
+    if isinstance(res_depth_val, (int, float)):
+        summary_html += f"<p><b>Residue Depth:</b> {res_depth_val:.2f} (Higher value = More buried)</p>"
+    else:
+        summary_html += f"<p><b>Residue Depth:</b> {res_depth_val}</p>"
+
+    if sst == 'Strand':
+        summary_html += "<p><i>Insight:</i> Mutations in strand regions might significantly disrupt beta-sheet structures, affecting protein folding and stability.</p>"
+    elif sst == 'AlphaHelix':
+        summary_html += "<p><i>Insight:</i> Alpha-helical mutations can alter helix packing, flexibility, or interactions, potentially impacting stability or function.</p>"
+    elif sst == 'None':
+        summary_html += "<p><i>Insight:</i> Mutations in coil/loop regions might introduce flexibility or alter surface interactions.</p>"
+
+    if isinstance(rsa_val, (int, float)):
+        if rsa_val < 0.2:
+            summary_html += "<p><i>Insight:</i> This residue is highly **buried**, suggesting it plays a critical role in the protein's core stability or internal packing. Changes here are often destabilizing.</p>"
+        elif rsa_val > 0.8:
+            summary_html += "<p><i>Insight:</i> This residue is highly **exposed**, indicating it might be involved in surface interactions, ligand binding, or protein-protein interfaces. Mutations here could affect function or solvent interactions.</p>"
+        else:
+            summary_html += "<p><i>Insight:</i> This residue has intermediate solvent accessibility, potentially involved in both structural integrity and surface interactions.</p>"
+
+    if isinstance(res_depth_val, (int, float)):
+        if res_depth_val > 6.0:
+            summary_html += "<p><i>Insight:</i> The residue is very deep within the protein, reinforcing its role in core structural stability.</p>"
+        elif res_depth_val < 3.0:
+            summary_html += "<p><i>Insight:</i> The residue is near the surface, potentially affecting surface interactions or flexibility.</p>"
+
+    return summary_html
+
+# --- Artifact Loading Logic ---
+def load_object_from_path(filepath, alternative_filepath=None):
+    if os.path.exists(filepath):
+        with open(filepath, 'rb') as f:
+            return pickle.load(f)
+    elif alternative_filepath and os.path.exists(alternative_filepath):
+        with open(alternative_filepath, 'rb') as f:
+            return pickle.load(f)
+    else:
+        raise FileNotFoundError(f"Object file not found at {filepath} or {alternative_filepath}")
+
+def load_json_from_path(filepath, alternative_filepath=None):
+    if os.path.exists(filepath):
+        with open(filepath, 'r') as f:
+            return json.load(f)
+    elif alternative_filepath and os.path.exists(alternative_filepath):
+        with open(alternative_filepath, 'r') as f:
+            return json.load(f)
+    else:
+        raise FileNotFoundError(f"JSON file not found at {filepath} or {alternative_filepath}")
+
+# Define file paths
+preprocessor_file = 'preprocessor.pkl'
+global_vars_file = 'global_vars.json'
+feature_names_file = 'feature_names.json'
+shap_background_data_file = 'shap_background_data.pt'
+model_save_path = 'final_multi_task_protein_stability_model_non_plm.pth'
+
+# Load preprocessor
+try:
+    preprocessor = load_object_from_path(preprocessor_file, f'src/{preprocessor_file}')
+except FileNotFoundError as e:
+    raise SystemExit(f"Error loading preprocessor: {e}. Please ensure 'preprocessor.pkl' is available.")
+
+# Load global variables
+try:
+    global_vars = load_json_from_path(global_vars_file, f'src/{global_vars_file}')
+except FileNotFoundError as e:
+    raise SystemExit(f"Error loading global variables: {e}. Please ensure 'global_vars.json' is available.")
+
+# Unpack global variables
+label_encoder = LabelEncoder()
+label_encoder.classes_ = np.array(global_vars['label_encoder_classes'])
+num_classes = global_vars['num_classes']
+best_hyperparams = global_vars['best_hyperparams']
+expected_value_reg = global_vars['expected_value_reg']
+expected_value_cls = np.array(global_vars['expected_value_cls']) # Already converted to list, convert back to np array
+AA_PROPERTIES = global_vars['AA_PROPERTIES']
+
+# Load feature names
+try:
+    feature_names = load_json_from_path(feature_names_file, f'src/{feature_names_file}')
+except FileNotFoundError as e:
+    raise SystemExit(f"Error loading feature names: {e}. Please ensure 'feature_names.json' is available.")
+
+# Load SHAP background data
+try:
+    background_data = torch.load(shap_background_data_file, map_location=device)
+except FileNotFoundError as e:
+    try:
+        background_data = torch.load(f'src/{shap_background_data_file}', map_location=device)
+    except FileNotFoundError:
+        raise SystemExit(f"Error loading SHAP background data: {e}. Please ensure 'shap_background_data.pt' is available.")
+
+# --- Load Trained Model ---
+# Determine input_dim from feature_names for model initialization
+input_dim = len(feature_names)
+loaded_model = MultiTaskMLP(input_dim, num_classes, dropout_rate=best_hyperparams['dropout']).to(device)
+
+# Check model path: root, then src/
+alt_model_save_path = f'src/{model_save_path}'
+if os.path.exists(model_save_path):
+    final_model_path_to_load = model_save_path
+elif os.path.exists(alt_model_save_path):
+    final_model_path_to_load = alt_model_save_path
+else:
+    raise SystemExit(f"Error: Model file not found at {model_save_path} or {alt_model_save_path}")
+
+loaded_model.load_state_dict(torch.load(final_model_path_to_load, map_location=device))
+loaded_model.eval() # Ensure eval mode is set immediately after loading
+
+# --- SHAP Explainers Setup ---
+model_reg_wrapper = RegressionHeadWrapper(loaded_model).to(device)
+model_cls_wrapper = ClassificationHeadWrapper(loaded_model).to(device)
+
+explainer_reg = shap.GradientExplainer(model_reg_wrapper, background_data)
+explainer_cls = shap.GradientExplainer(model_cls_wrapper, background_data)
+
+# --- Main Prediction Function for Gradio ---
+def predict_stability_change(
+    weight, blosum62, pos, year, aro, ca_depth, mut_count, neg, sul,
+    relative_bfactor, ph, neu, phi, psi, rsa, res_depth, temperature,
+    acc, don, pam250, length, dtm,
+    sst, measure, protein_name, source, original_aa, mutant_aa, mutated_chain, mutation_type, method_val
+):
+    # Explicitly cast numerical inputs to float and handle None values
+    # Gradio's gr.Number will return float or int, no need for manual casting if types are consistent.
+    # However, if using gr.Slider with None as default, it might return None. Ensure defaults or cast.
+    # For robustness, we still ensure float type here.
+
+    input_data_df_raw = pd.DataFrame({
+        'weight': [float(weight) if weight is not None else 0.0],
+        'blosum62': [float(blosum62) if blosum62 is not None else 0.0],
+        'pos': [float(pos) if pos is not None else 0.0],
+        'year': [float(year) if year is not None else 0.0],
+        'aro': [float(aro) if aro is not None else 0.0],
+        'ca_depth': [float(ca_depth) if ca_depth is not None else 0.0],
+        'mut_count': [float(mut_count) if mut_count is not None else 0.0],
+        'neg': [float(neg) if neg is not None else 0.0],
+        'sul': [float(sul) if sul is not None else 0.0],
+        'relative_bfactor': [float(relative_bfactor) if relative_bfactor is not None else 0.0],
+        'ph': [float(ph) if ph is not None else 0.0],
+        'neu': [float(neu) if neu is not None else 0.0],
+        'phi': [float(phi) if phi is not None else 0.0],
+        'psi': [float(psi) if psi is not None else 0.0],
+        'rsa': [float(rsa) if rsa is not None else 0.0],
+        'res_depth': [float(res_depth) if res_depth is not None else 0.0],
+        'temperature': [float(temperature) if temperature is not None else 0.0],
+        'acc': [float(acc) if acc is not None else 0.0],
+        'don': [float(don) if don is not None else 0.0],
+        'pam250': [float(pam250) if pam250 is not None else 0.0],
+        'length': [float(length) if length is not None else 0.0],
+        'dtm': [float(dtm) if dtm is not None else 0.0],
+        'sst': [sst],
+        'measure': [measure],
+        'protein': [protein_name],
+        'source': [source],
+        'original_aa': [original_aa],
+        'mutant_aa': [mutant_aa],
+        'mutated_chain': [mutated_chain],
+        'mutation_type': [mutation_type],
+        'method': [method_val]
+    })
+
+    # Preprocess the input data
+    processed_input = preprocessor.transform(input_data_df_raw)
+    if hasattr(processed_input, 'toarray'):
+        processed_input = processed_input.toarray()
+
+    input_tensor = torch.tensor(processed_input, dtype=torch.float32).to(device)
+
+    # Make predictions
+    loaded_model.eval()
+    with torch.no_grad():
+        ddg_pred, effect_logits = loaded_model(input_tensor)
+
+    predicted_ddg = ddg_pred.item()
+    predicted_effect_class_idx = torch.argmax(effect_logits, dim=1).item()
+    predicted_effect_label = label_encoder.inverse_transform([predicted_effect_class_idx])[0]
+
+    # 4. Generate AA Property Comparison
+    aa_comparison_results = get_aa_property_comparison(original_aa, mutant_aa)
+    aa_props_html = "<h3>Amino Acid Property Comparison:</h3>"
+    if isinstance(aa_comparison_results, str):
+        aa_props_html += f"<p>{aa_comparison_results}</p>"
+    else:
+        aa_props_html += "<table><tr><th>Original</th><th>Mutant</th><th>Change</th><th>Property</th></tr>"
+        for prop, vals in aa_comparison_results.items():
+            aa_props_html += f"<tr><td>{vals['original']}</td><td>{vals['mutant']}</td><td>{vals['change']}</td><td><b>{prop}</b></td></tr>"
+        aa_props_html += "</table>"
+
+    # 5. Generate SHAP Explanations (Waterfall Plots as static images)
+    # Regression SHAP
+    fig_reg, ax_reg = plt.subplots(figsize=(10, 6))
+    shap.waterfall_plot(shap.Explanation(values=explainer_reg.shap_values(input_tensor)[0].flatten(),
+                                         base_values=expected_value_reg,
+                                         data=input_tensor.cpu().numpy().flatten(),
+                                         feature_names=feature_names),
+                         max_display=15, show=False)
+    ax_reg.set_title("SHAP Waterfall Plot for ΔΔG Prediction")
+    shap_html_reg_img = plot_to_base64(fig_reg)
+
+    # Classification SHAP (for predicted class)
+    fig_cls, ax_cls = plt.subplots(figsize=(10, 6))
+    shap.waterfall_plot(shap.Explanation(values=explainer_cls.shap_values(input_tensor)[0, :, predicted_effect_class_idx].flatten(),
+                                         base_values=expected_value_cls[predicted_effect_class_idx],
+                                         data=input_tensor.cpu().numpy().flatten(),
+                                         feature_names=feature_names),
+                         max_display=15, show=False)
+    ax_cls.set_title(f"SHAP Waterfall Plot for Effect Prediction (Class: {predicted_effect_label})")
+    shap_html_cls_img = plot_to_base64(fig_cls)
+
+    # 6. Generate Local Feature Importance Chart (uses base64 plotting function)
+    local_fi_reg_img = plot_local_feature_importance(explainer_reg.shap_values(input_tensor)[0], feature_names,
+                                                      title="Local Feature Importance for ΔΔG Prediction")
+    local_fi_cls_img = plot_local_feature_importance(explainer_cls.shap_values(input_tensor)[0, :, predicted_effect_class_idx], feature_names,
+                                                      title=f"Local Feature Importance for Effect Prediction ({predicted_effect_label})")
+
+    # 7. Generate Top Global Feature Importance (using the background data)
+    top_reg_features = get_top_features_by_shap(explainer_reg, background_data, feature_names, top_n=5, task_type='regression')
+    top_cls_features = get_top_features_by_shap(explainer_cls, background_data, feature_names, top_n=5, task_type='classification')
+
+    top_features_html = "<h3>Top 5 Globally Important Features:</h3><p><b>Regression (ΔΔG):</b></p><ul>"
+    for item in top_reg_features:
+        top_features_html += f"<li>{item['Feature']}: {item['Mean_Abs_SHAP']:.4f}</li>"
+    top_features_html += "</ul><p><b>Classification (Effect):</b></p><ul>"
+    for item in top_cls_features:
+        top_features_html += f"<li>{item['Feature']}: {item['Mean_Abs_SHAP']:.4f}</li>"
+    top_features_html += "</ul>"
+
+    # 8. Get Mutation Context Summary
+    mutation_context_html = get_mutation_context_summary(input_data_df_raw)
+
+    # Return all outputs for Gradio
+    return (
+        f"Predicted ΔΔG: **{predicted_ddg:.4f}** kcal/mol",
+        f"Predicted Effect: **{predicted_effect_label.upper()}**",
+        aa_props_html,
+        mutation_context_html,
+        shap_html_reg_img,
+        local_fi_reg_img,
+        shap_html_cls_img,
+        local_fi_cls_img,
+        top_features_html
+    )
+
+# --- Gradio Interface Layout ---
+# Define Input Components grouped by sections
+input_general_props = [
+    gr.Number(minimum=-100.0, maximum=100000.0, value=28726.09, label="Weight (Da)", interactive=True),
+    gr.Number(minimum=0.0, maximum=1000.0, value=268.0, label="Protein Length (AA)", step=1, interactive=True),
+    gr.Textbox(value="Tryptophan synthase alpha chain", label="Protein Name", interactive=True),
+    gr.Dropdown(choices=['Escherichia coli (strain K12)', 'Enterobacteria phage T4', 'Homo sapiens', 'Rattus norvegicus', 'Pseudomonas putida', 'Saccharomyces cerevisiae', 'Bacillus subtilis', 'Thermococcus kodakarensis', 'Unknown'], value='Escherichia coli (strain K12)', label="Source Organism", interactive=True),
+    gr.Number(minimum=1970, maximum=2024, value=1979, step=1, label="Year of Study", interactive=True)
+]
+
+input_mutation_details = [
+    gr.Dropdown(choices=list(AA_PROPERTIES.keys()), value='E', label="Original Amino Acid", interactive=True),
+    gr.Dropdown(choices=list(AA_PROPERTIES.keys()), value='M', label="Mutant Amino Acid", interactive=True),
+    gr.Number(minimum=-10.0, maximum=1000.0, value=0.0, label="Position (Sequence Index)", step=1, interactive=True),
+    gr.Dropdown(choices=['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'unsigned', 'Unknown'], value='A', label="Mutated Chain", interactive=True),
+    gr.Dropdown(choices=['Single', 'Multiple', 'Unknown'], value='Single', label="Mutation Type", interactive=True),
+    gr.Number(minimum=0, maximum=10, value=0, step=1, label="Mutation Count", interactive=True)
+]
+
+input_physicochemical = [
+    gr.Number(minimum=-10.0, maximum=10.0, value=-1.0, label="Blosum62 Score", step=0.1, interactive=True),
+    gr.Number(minimum=-10.0, maximum=10.0, value=0.0, label="PAM250 Score", step=0.1, interactive=True),
+    gr.Number(minimum=-10.0, maximum=10.0, value=0.0, label="Aromatic AA Count", step=0.1, interactive=True),
+    gr.Number(minimum=-10.0, maximum=10.0, value=-2.0, label="Negative AA Count", step=0.1, interactive=True),
+    gr.Number(minimum=-10.0, maximum=10.0, value=-1.0, label="Neutral AA Count", step=0.1, interactive=True),
+    gr.Number(minimum=-10.0, maximum=10.0, value=1.0, label="Sulfur AA Count", step=0.1, interactive=True),
+    gr.Number(minimum=-10.0, maximum=10.0, value=-2.0, label="Acceptor Count", step=0.1, interactive=True),
+    gr.Number(minimum=-10.0, maximum=10.0, value=0.0, label="Donor Count", step=0.1, interactive=True)
+]
+
+input_structural = [
+    gr.Dropdown(choices=['Strand', 'AlphaHelix', 'None', 'Turn', 'Bend', 'Isolatedbeta-bridge', '3-10Helix', 'PiHelix', 'polyproline', 'Unknown'], value='Strand', label="Secondary Structure Type", interactive=True),
+    gr.Number(minimum=0.0, maximum=1.0, value=0.0, label="RSA (Relative Solvent Accessibility)", step=0.01, interactive=True),
+    gr.Number(minimum=0.0, maximum=10.0, value=4.14, label="CA Depth (Å)", step=0.01, interactive=True),
+    gr.Number(minimum=0.0, maximum=10.0, value=3.53, label="Residue Depth (Å)", step=0.01, interactive=True),
+    gr.Number(minimum=0.0, maximum=10.0, value=3.47, label="Relative B-Factor", step=0.01, interactive=True),
+    gr.Number(minimum=-180.0, maximum=180.0, value=-118.5, label="Phi Angle (°)", step=0.1, interactive=True),
+    gr.Number(minimum=-180.0, maximum=180.0, value=113.0, label="Psi Angle (°)", step=0.1, interactive=True)
+]
+
+input_experimental = [
+    gr.Dropdown(choices=['CD', 'Unavailable', 'Fluorescence', 'DSC', 'NMR', 'Absorbance', 'Activity', 'ITC', 'Other', 'Unknown'], value='CD', label="Measurement Method", interactive=True),
+    gr.Dropdown(choices=['GdnHCl', 'Thermal', 'Unavailable', 'Urea', 'GdnSCN', 'pH-stability', 'TFE', 'Proteolysis', 'DSC, CD', 'Absorbance, Fluorescence', 'Fluorescence, GdnHCl', 'Unknown'], value='GdnHCl', label="Denaturation Method", interactive=True),
+    gr.Number(minimum=0.0, maximum=14.0, value=7.0, label="pH", step=0.1, interactive=True),
+    gr.Number(minimum=273.0, maximum=373.0, value=298.95, label="Temperature (K)", step=0.01, interactive=True),
+    gr.Number(minimum=-100.0, maximum=100.0, value=0.0, label="dTM (Change in Melting Temp)", step=0.1, interactive=True)
+]
+
+# Combine all inputs into a single list for the fn call, ordered as expected by predict_stability_change
+all_input_components = [
+    input_general_props[0], input_physicochemical[0], input_mutation_details[2], input_general_props[4],
+    input_physicochemical[2], input_structural[2], input_mutation_details[5], input_physicochemical[3],
+    input_physicochemical[5], input_structural[4], input_experimental[2], input_physicochemical[4],
+    input_structural[5], input_structural[6], input_structural[1], input_structural[3],
+    input_experimental[3], input_physicochemical[6], input_physicochemical[7], input_physicochemical[1],
+    input_general_props[1], input_experimental[4],
+    input_structural[0], input_experimental[0], input_general_props[2], input_general_props[3],
+    input_mutation_details[0], input_mutation_details[1], input_mutation_details[3], input_mutation_details[4],
+    input_experimental[1]
+]
+
+output_components = [
+    gr.Markdown(label="Predicted ΔΔG"),
+    gr.Markdown(label="Predicted Effect"),
+    gr.HTML(label="Amino Acid Property Comparison"),
+    gr.HTML(label="Mutation Structural Context"),
+    gr.HTML(label="SHAP Waterfall Plot (Regression)"),
+    gr.HTML(label="Local Feature Importance (Regression)"),
+    gr.HTML(label="SHAP Waterfall Plot (Classification)"),
+    gr.HTML(label="Local Feature Importance (Classification)"),
+    gr.HTML(label="Top Global Feature Importance")
+]
+
+# Create the Gradio Interface
+with gr.Blocks(theme=gr.themes.Soft()) as iface:
+    gr.Markdown("# Protein Stability Change (ΔΔG) Prediction with Explainability")
+    gr.Markdown(
+        "Predict ΔΔG and mutation effect for single amino acid substitutions in proteins. "
+        "Explore physicochemical changes, mutation structural context, and feature importance with SHAP explanations. "
+        "Designed for researchers and doctors for practical applications." 
+        "All backend features are included and frontend template is professional and exceptional."
+    )
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            gr.Markdown("## Input Features")
+            with gr.Accordion("General Properties", open=True):
+                for component in input_general_props:
+                    component.render()
+            with gr.Accordion("Mutation Details", open=False):
+                for component in input_mutation_details:
+                    component.render()
+            with gr.Accordion("Physicochemical & Substitution Scores", open=False):
+                for component in input_physicochemical:
+                    component.render()
+            with gr.Accordion("Structural Features", open=False):
+                for component in input_structural:
+                    component.render()
+            with gr.Accordion("Experimental Conditions", open=False):
+                for component in input_experimental:
+                    component.render()
+
+            submit_btn = gr.Button("Predict & Explain", variant="primary")
+
+        with gr.Column(scale=2):
+            gr.Markdown("## Prediction Results & Explanations")
+            with gr.Tab("Summary & Predictions"):
+                output_components[0].render()
+                output_components[1].render()
+                output_components[2].render()
+                output_components[3].render()
+            with gr.Tab("SHAP Explanations (ΔΔG)"):
+                gr.Markdown("### How features influence ΔΔG Prediction (SHAP Waterfall Plot)")
+                output_components[4].render()
+                gr.Markdown("### Local Feature Contributions (ΔΔG)")
+                output_components[5].render()
+            with gr.Tab("SHAP Explanations (Effect)"):
+                gr.Markdown("### How features influence Effect Classification (SHAP Waterfall Plot)")
+                output_components[6].render()
+                gr.Markdown("### Local Feature Contributions (Effect)")
+                output_components[7].render()
+            with gr.Tab("Global Feature Importance"):
+                output_components[8].render()
+
+    submit_btn.click(
+        fn=predict_stability_change,
+        inputs=all_input_components,
+        outputs=output_components
+    )
+
+iface.launch(debug=True, share=True)

feature_names.json

@@ -0,0 +1 @@
+["num__weight", "num__blosum62", "num__pos", "num__year", "num__aro", "num__ca_depth", "num__mut_count", "num__neg", "num__sul", "num__relative_bfactor", "num__ph", "num__neu", "num__phi", "num__psi", "num__rsa", "num__res_depth", "num__temperature", "num__acc", "num__don", "num__pam250", "num__length", "num__dtm", "cat__sst_3-10Helix", "cat__sst_AlphaHelix", "cat__sst_Bend", "cat__sst_Isolatedbeta-bridge", "cat__sst_None", "cat__sst_PiHelix", "cat__sst_Strand", "cat__sst_Turn", "cat__measure_-", "cat__measure_Abs", "cat__measure_Abs, CD, Fluorescence", "cat__measure_Absorbance", "cat__measure_Absorsance", "cat__measure_Activity", "cat__measure_CD", "cat__measure_CD (Far-UV)", "cat__measure_CD (far-UV)", "cat__measure_CD, DSC", "cat__measure_CD, DSF", "cat__measure_CD, Fluorescence", "cat__measure_CD, Fluorescence, DSC", "cat__measure_CD, Fluorescence, Thiol reactivity", "cat__measure_CD, ITC", "cat__measure_CD, Optical Probes", "cat__measure_CD,DSC, Fluorescence", "cat__measure_CD,Fluorescence", "cat__measure_Chromatography", "cat__measure_DSC", "cat__measure_DSC, CD, Fluorescence", "cat__measure_DSC, Fluorescence", "cat__measure_DSF", "cat__measure_ESR", "cat__measure_Enzyme Activity", "cat__measure_Enzyme activity", "cat__measure_Fluorescence", "cat__measure_Fluorescence, CD", "cat__measure_Gel electrophoresis", "cat__measure_HPLC", "cat__measure_Hydrogen exchange", "cat__measure_ITC", "cat__measure_Isothermal denaturation", "cat__measure_Isothermal titration calorimetry", "cat__measure_MCD", "cat__measure_NMR", "cat__measure_NMR, Fluorescence", "cat__measure_PTS, Fluorescence", "cat__measure_RP-HPLC", "cat__measure_Stopped-flow fluorescence", "cat__measure_UV", "cat__measure_UV/vis spectrophotometer", "cat__measure_Unavailable", "cat__measure_fluorescence", "cat__measure_fluorescence / CD", "cat__source_Aeromonas hydrophila", "cat__source_Amphitrite ornata", "cat__source_Anopheles_dirus", "cat__source_Aquifex aeolicus (strain VF5)", "cat__source_Arthrobacter sp.", "cat__source_Aspergillus kawachii", "cat__source_Aspergillus oryzae (strain ATCC 42149 / RIB 40)", "cat__source_Avena sativa (Oat)", "cat__source_Bacillus amyloliquefaciens", "cat__source_Bacillus caldolyticus", "cat__source_Bacillus licheniformis", "cat__source_Bacillus subtilis (strain 168)", "cat__source_Bacillus thuringiensis subsp. entomocidus", "cat__source_Barley", "cat__source_Borrelia burgdorferi", "cat__source_Bos taurus", "cat__source_Bovine", "cat__source_Brevundimonas diminuta", "cat__source_Caenorhabditis elegans", "cat__source_Canis lupus familiaris", "cat__source_Capra hircus", "cat__source_Clostridium pasteurianum", "cat__source_Cucurbita maxima", "cat__source_Dengue virus type 2 (strain Puerto Rico/PR159-S1/1969)", "cat__source_Designed", "cat__source_Desulfovibrio desulfuricans", "cat__source_Dioscoreophyllum cumminsii", "cat__source_Drosophila melanogaster", "cat__source_Enterobacteria phage M13", "cat__source_Enterobacteria phage T4", "cat__source_Enterobacteria phage fd", "cat__source_Equus caballus", "cat__source_Escherichia coli", "cat__source_Escherichia coli (strain K12)", "cat__source_Escherichia coli O157:H7", "cat__source_Escherichia coli O6:H1 (strain CFT073 / ATCC 700928 / UPEC)", "cat__source_Escherichia phage lambda", "cat__source_Escherichia_coli", "cat__source_Finegoldia_magna", "cat__source_Gallus gallus", "cat__source_Geobacillus stearothermophilus", "cat__source_Halobacterium salinarum (strain ATCC 700922 / JCM 11081 / NRC-1)", "cat__source_Hirudo medicinalis", "cat__source_Homo sapiens", "cat__source_Hordeum vulgare", "cat__source_Human", "cat__source_Human immunodeficiency virus type 1 group M subtype B (isolate NY5)", "cat__source_Katsuwonus pelamis", "cat__source_Kitasatospora aureofaciens", "cat__source_Lama glama", "cat__source_Lithobates pipiens", "cat__source_Locusta migratoria", "cat__source_Mesocricetus auratus", "cat__source_Methanothermus fervidus", "cat__source_Mnemiopsis_leidyi", "cat__source_Mus musculus", "cat__source_Mycobacterium tuberculosis", "cat__source_Mycobacterium tuberculosis (strain ATCC 25618 / H37Rv)", "cat__source_None", "cat__source_Nostoc sp. (strain ATCC 29151 / PCC 7119)", "cat__source_Nostoc sp. (strain PCC 7120 / SAG 25.82 / UTEX 2576)", "cat__source_Oryctolagus cuniculus", "cat__source_Oryza sativa subsp. japonica", "cat__source_Ovis aries", "cat__source_Paenibacillus polymyxa", "cat__source_Pan paniscus", "cat__source_Pentadiplandra brazzeana", "cat__source_Photinus pyralis", "cat__source_Physeter macrocephalus", "cat__source_Podospora anserina", "cat__source_Proteus hauseri", "cat__source_Pseudomonas aeruginosa (strain ATCC 15692 / DSM 22644 / CIP 104116 / JCM 14847 / LMG 12228 / 1C / PRS 101 / PAO1)", "cat__source_Pseudomonas putida", "cat__source_Pseudomonas syringae pv. tomato", "cat__source_Pyrococcus furiosus (strain ATCC 43587 / DSM 3638 / JCM 8422 / Vc1)", "cat__source_Pyrococcus_furiosus", "cat__source_Rattus norvegicus", "cat__source_Rhodobacter capsulatus (strain ATCC BAA-309 / NBRC 16581 / SB1003)", "cat__source_Rhodopseudomonas palustris (strain ATCC BAA-98 / CGA009)", "cat__source_Rous sarcoma virus (strain Prague C)", "cat__source_Saccharolobus shibatae", "cat__source_Saccharolobus solfataricus (strain ATCC 35092 / DSM 1617 / JCM 11322 / P2)", "cat__source_Saccharomyces cerevisiae (strain ATCC 204508 / S288c)", "cat__source_Salmonella phage P22", "cat__source_Salmonella typhimurium (strain LT2 / SGSC1412 / ATCC 700720)", "cat__source_Schizosaccharomyces pombe (strain 972 / ATCC 24843)", "cat__source_Severe acute respiratory syndrome coronavirus", "cat__source_Shigella flexneri", "cat__source_Simian foamy virus type 1", "cat__source_Spodoptera frugiperda (Fall armyworm)", "cat__source_Staphylococcus aureus", "cat__source_Stichodactyla helianthus", "cat__source_Streptococcus sp. group G", "cat__source_Streptomyces albogriseolus", "cat__source_Streptomyces lividans", "cat__source_Streptomyces sp. (strain N174)", "cat__source_Sulfolobus acidocaldarius (strain ATCC 33909 / DSM 639 / JCM 8929 / NBRC 15157 / NCIMB 11770)", "cat__source_Sus scrofa", "cat__source_Thermococcus celer", "cat__source_Thermococcus kodakarensis (strain ATCC BAA-918 / JCM 12380 / KOD1)", "cat__source_Thermotoga maritima (strain ATCC 43589 / MSB8 / DSM 3109 / JCM 10099)", "cat__source_Thermus thermophilus", "cat__source_Thermus thermophilus (strain HB8 / ATCC 27634 / DSM 579)", "cat__source_Tobacco etch virus", "cat__source_Vibrio harveyi", "cat__source_Xenopus laevis", "cat__source_Zoarces americanus", "cat__source_designed", "cat__source_human", "cat__source_none", "cat__protein_10 kDa chaperonin", "cat__protein_3-isopropylmalate dehydrogenase", "cat__protein_30S ribosomal protein S16", "cat__protein_30S ribosomal protein S6", "cat__protein_5'-AMP-activated protein kinase subunit beta-1", "cat__protein_5'-AMP-activated protein kinase subunit beta-2", "cat__protein_50S ribosomal protein L23", "cat__protein_50S ribosomal protein L30e", "cat__protein_50S ribosomal protein L9", "cat__protein_60 kDa chaperonin", "cat__protein_6aJL2", "cat__protein_A26.8 VHH antibody", "cat__protein_AL-103", "cat__protein_AL-12", "cat__protein_ALPHA-T-ALPHA", "cat__protein_Actin-binding protein", "cat__protein_Acyl-CoA-binding protein", "cat__protein_Acylphosphatase", "cat__protein_Acylphosphatase-1", "cat__protein_Adenosine deaminase", "cat__protein_Adenylate kinase", "cat__protein_Adrenodoxin, mitochondrial", "cat__protein_Aerolysin", "cat__protein_Alkanal monooxygenase alpha chain", "cat__protein_Alpha-1-antitrypsin", "cat__protein_Alpha-lactalbumin", "cat__protein_Apolipophorin-3b", "cat__protein_Apolipoprotein A-I", "cat__protein_Apolipoprotein(a)", "cat__protein_Asparaginase", "cat__protein_Aspartate aminotransferase", "cat__protein_Attachment protein G3P", "cat__protein_Azurin", "cat__protein_BRCA1-associated RING domain protein 1", "cat__protein_Bacteriorhodopsin", "cat__protein_Barstar", "cat__protein_Beta lactamase", "cat__protein_Beta-2-microglobulin", "cat__protein_Beta-galactosidase", "cat__protein_Beta-glucosidase", "cat__protein_Beta-glucosidase B", "cat__protein_Beta-lactamase", "cat__protein_Beta-lactamase TEM", "cat__protein_Beta-lactoglobulin", "cat__protein_Bifunctional ligase/repressor BirA", "cat__protein_Blue-light photoreceptor", "cat__protein_Bromodomain-containing protein 2", "cat__protein_Bromodomain-containing protein 3", "cat__protein_Bromodomain-containing protein 4", "cat__protein_CREB-binding protein", "cat__protein_Calmodulin-2 A", "cat__protein_Carbonic anhydrase 2", "cat__protein_Carboxypeptidase A2", "cat__protein_Cellular retinoic acid-binding protein 1", "cat__protein_Cellular tumor antigen p53", "cat__protein_Chemotaxis protein CheY", "cat__protein_Chitinase", "cat__protein_Chitosanase", "cat__protein_Chymotrypsin inhibitor 2", "cat__protein_Chymotrypsin inhibitor-2", "cat__protein_Cold shock protein CspA", "cat__protein_Cold shock protein CspB", "cat__protein_Cold shock-like protein", "cat__protein_Colicin-E7 immunity protein", "cat__protein_Colicin-E9 immunity protein", "cat__protein_Collagen alpha-1(XVIII) chain", "cat__protein_Copper resistance protein C", "cat__protein_Copper-transporting ATPase 2", "cat__protein_Creatine kinase M-type", "cat__protein_Crystaline entomocidal protoxin", "cat__protein_Cyclin-dependent kinase inhibitor 2A", "cat__protein_Cyclin-dependent kinases regulatory subunit", "cat__protein_Cyclin-dependent kinases regulatory subunit 1", "cat__protein_Cystatin-B", "cat__protein_Cytochrome b", "cat__protein_Cytochrome b5", "cat__protein_Cytochrome c", "cat__protein_Cytochrome c isoform 1", "cat__protein_Cytochrome c isoform 2", "cat__protein_Cytochrome c mitochondrial import factor CYC2", "cat__protein_Cytochrome c'", "cat__protein_Cytochrome c, somatic", "cat__protein_Cytochrome c-551", "cat__protein_Cytochrome c2", "cat__protein_DELTA-stichotoxin-She4a", "cat__protein_DELTA-stichotoxin-She4b", "cat__protein_DF1", "cat__protein_DNA-Binding protein G5P", "cat__protein_DNA-binding protein 7d", "cat__protein_DNA-binding protein Fis", "cat__protein_DNA-binding protein HMf-2", "cat__protein_DNA-binding protein HU", "cat__protein_DNA/RNA-binding protein Alba 1", "cat__protein_Defensin-like protein", "cat__protein_Dehaloperoxidase A", "cat__protein_Dihydrofolate reductase", "cat__protein_Dihydrolipoyllysine-residue acetyltransferase component of pyruvate dehydrogenase complex", "cat__protein_Dihydrolipoyllysine-residue succinyltransferase component of 2-oxoglutarate dehydrogenase complex", "cat__protein_Disks large homolog 4", "cat__protein_Disulfide bond formation protein D", "cat__protein_Dystrophin", "cat__protein_E3 ubiquitin-protein ligase RSP5", "cat__protein_E3 ubiquitin-protein ligase parkin", "cat__protein_Eglin C", "cat__protein_Endo-1,4-beta-xylanase A", "cat__protein_Endolysin", "cat__protein_Estrogen receptor", "cat__protein_Fatty acid-binding protein, brain", "cat__protein_Fatty acid-binding protein, heart", "cat__protein_Fatty acid-binding protein, intestinal", "cat__protein_Ferrienterobactin receptor", "cat__protein_Fibroblast growth factor 1", "cat__protein_Fibronectin", "cat__protein_FimH", "cat__protein_Flavodoxin", "cat__protein_Forkhead box protein P1", "cat__protein_Formamidopyrimidine-DNA glycosylase", "cat__protein_Frataxin, mitochondrial", "cat__protein_Gag polyprotein", "cat__protein_Gag-Pol polyprotein", "cat__protein_Gamma-crystallin D", "cat__protein_Gastrotropin", "cat__protein_Gelsolin", "cat__protein_General control protein GCN4", "cat__protein_Genome polyprotein", "cat__protein_Glucoamylase I", "cat__protein_Glucocorticoid receptor", "cat__protein_Glutamate dehydrogenase", "cat__protein_Glutathione S-transferase GstA", "cat__protein_Glutathione S-transferase Mu 1", "cat__protein_Glycerol-3-phosphate dehydrogenase [NAD(+)], cytoplasmic", "cat__protein_Granulocyte-macrophage colony-stimulating factor", "cat__protein_Growth factor receptor-bound protein 2", "cat__protein_Growth hormone receptor", "cat__protein_Guanyl-specific ribonuclease Sa", "cat__protein_Guanyl-specific ribonuclease Sa3", "cat__protein_Guanyl-specific ribonuclease T1", "cat__protein_HIV-1 capsid protein", "cat__protein_Heterokaryon incompatibility protein s", "cat__protein_High mobility group protein B1", "cat__protein_Ig gamma-1 chain C region secreted form", "cat__protein_Ig heavy chain V region 6.96", "cat__protein_Immunoglobulin G-binding protein A", "cat__protein_Immunoglobulin G-binding protein G", "cat__protein_Immunoglobulin heavy constant gamma 1", "cat__protein_Immunoglobulin kappa variable 1D-33", "cat__protein_Inhibitor of trypsin and hageman factor", "cat__protein_Insulin", "cat__protein_Insulin receptor", "cat__protein_Interleukin-1 beta", "cat__protein_Interleukin-4", "cat__protein_Isocitrate lyase", "cat__protein_LamA", "cat__protein_Lipase EstA", "cat__protein_Lipid A palmitoyltransferase PagP", "cat__protein_Lipocalin-1", "cat__protein_Luciferin 4-monooxygenase", "cat__protein_Lymphotactin", "cat__protein_Lysozyme C", "cat__protein_Lysozyme C, milk isozyme", "cat__protein_MAK33", "cat__protein_Major capsid protein", "cat__protein_Major prion protein", "cat__protein_Maltose/maltodextrin-binding periplasmic protein", "cat__protein_Membrane-associated guanylate kinase, WW and PDZ domain-containing protein 1", "cat__protein_Methyl-CpG-binding protein 2", "cat__protein_Microtubule-associated proteins 1A/1B light chain 3B", "cat__protein_Monellin chain A", "cat__protein_Monellin chain B", "cat__protein_Myelin P2 protein", "cat__protein_Myoglobin", "cat__protein_NF-kappa-B inhibitor alpha", "cat__protein_NPH1-1", "cat__protein_Neuronal calcium sensor 1", "cat__protein_Non-specific lipid-transfer protein", "cat__protein_Non-specific lipid-transfer protein 2", "cat__protein_Nuclear RNA export factor 1", "cat__protein_Nuclease-Concanavalin A hybrid protein", "cat__protein_OPG2_FAB_VL", "cat__protein_Orotidine 5'-phosphate decarboxylase", "cat__protein_Outer membrane protein A", "cat__protein_PTS system mannose-specific EIIAB component", "cat__protein_Pancreatic trypsin inhibitor", "cat__protein_Parathion hydrolase", "cat__protein_Parvalbumin alpha", "cat__protein_Peptide YY", "cat__protein_Peptidyl-prolyl cis-trans isomerase FKBP1A", "cat__protein_Peptidyl-prolyl cis-trans isomerase NIMA-interacting 1", "cat__protein_Periplasmic chaperone Spy", "cat__protein_Peroxisome proliferator-activated receptor gamma", "cat__protein_Phosphocarrier protein HPr", "cat__protein_Phosphoglycerate kinase", "cat__protein_Phosphoglycerate kinase 1", "cat__protein_Phospholipase A2", "cat__protein_Phospholipase A2, major isoenzyme", "cat__protein_Plasminogen activator inhibitor 1", "cat__protein_Polyglutamine-binding protein 1", "cat__protein_Polyubiquitin-B", "cat__protein_Pre-mRNA-processing factor 40 homolog A", "cat__protein_Pre-mRNA-splicing factor URN1", "cat__protein_Pro-Pol polyprotein", "cat__protein_Pro-epidermal growth factor", "cat__protein_Profilin-1", "cat__protein_Prolactin", "cat__protein_Protein P-30", "cat__protein_Protein RecA", "cat__protein_Protein S100-B", "cat__protein_Protein S100-G", "cat__protein_Protein THO1", "cat__protein_ProteinL", "cat__protein_Proteinase inhibitor", "cat__protein_Proto-oncogene tyrosine-protein kinase Src", "cat__protein_Pyrin", "cat__protein_RAF proto-oncogene serine/threonine-protein kinase", "cat__protein_Receptor-interacting serine/threonine-protein kinase 2", "cat__protein_Regulatory protein cro", "cat__protein_Regulatory protein rop", "cat__protein_Replicase polyprotein 1ab", "cat__protein_Repressor protein cI", "cat__protein_Riboflavin-binding protein", "cat__protein_Ribonuclease", "cat__protein_Ribonuclease H", "cat__protein_Ribonuclease HI", "cat__protein_Ribonuclease HII", "cat__protein_Ribonuclease P protein component", "cat__protein_Ribonuclease pancreatic", "cat__protein_Ribose import binding protein RbsB", "cat__protein_Rubredoxin", "cat__protein_S-adenosylmethionine synthase isoform type-1", "cat__protein_SCO1 protein homolog", "cat__protein_SH2 domain-containing protein 1A", "cat__protein_Serine hydroxymethyltransferase", "cat__protein_Serine/threonine-protein kinase pim-1", "cat__protein_Sialidase", "cat__protein_Soluble cytochrome b562", "cat__protein_Spectrin alpha chain, non-erythrocytic 1", "cat__protein_Steroid Delta-isomerase", "cat__protein_Subtilisin BPN'", "cat__protein_Subtilisin inhibitor", "cat__protein_Subtilisin-chymotrypsin inhibitor-2A", "cat__protein_Superoxide dismutase", "cat__protein_Superoxide dismutase [Cu-Zn]", "cat__protein_Superoxide dismutase [Mn], mitochondrial", "cat__protein_T-cell surface antigen CD2", "cat__protein_TC10b", "cat__protein_Tail spike protein", "cat__protein_Tenascin", "cat__protein_Tetracycline repressor protein class D", "cat__protein_Thermonuclease", "cat__protein_Thiol:disulfide interchange protein DsbA", "cat__protein_Thioredoxin 1", "cat__protein_Thioredoxin-like protein 4A", "cat__protein_Tissue factor", "cat__protein_Tissue-type plasminogen activator", "cat__protein_Titin", "cat__protein_Transcription antitermination protein NusB", "cat__protein_Transcription elongation regulator 1", "cat__protein_Transcriptional activator Myb", "cat__protein_Transcriptional coactivator YAP1", "cat__protein_Transcriptional enhancer factor TEF-1", "cat__protein_Transcriptional enhancer factor TEF-3", "cat__protein_Transcriptional enhancer factor TEF-4", "cat__protein_Transcriptional repressor arc", "cat__protein_Transforming growth factor-beta-induced protein ig-h3", "cat__protein_Transketolase 1", "cat__protein_Triosephosphate isomerase", "cat__protein_Tryptophan synthase alpha chain", "cat__protein_Twitchin", "cat__protein_Type-2 restriction enzyme PvuII", "cat__protein_Type-3 ice-structuring protein HPLC 12", "cat__protein_Tyrosine--tRNA ligase", "cat__protein_Tyrosine-protein kinase Fyn", "cat__protein_Tyrosine-protein phosphatase non-receptor type 13", "cat__protein_U1 small nuclear ribonucleoprotein A", "cat__protein_Ubiquitin", "cat__protein_Variable large protein", "cat__protein_Villin-1", "cat__protein_WD repeat-containing protein 5", "cat__protein_cAMP-activated global transcriptional regulator CRP", "cat__protein_de novo designed three helix bundle GRa3D", "cat__protein_glutathione_transferases", "cat__protein_mnemiopsin", "cat__original_aa_A", "cat__original_aa_C", "cat__original_aa_D", "cat__original_aa_E", "cat__original_aa_F", "cat__original_aa_G", "cat__original_aa_H", "cat__original_aa_I", "cat__original_aa_K", "cat__original_aa_L", "cat__original_aa_M", "cat__original_aa_N", "cat__original_aa_P", "cat__original_aa_Q", "cat__original_aa_R", "cat__original_aa_S", "cat__original_aa_T", "cat__original_aa_V", "cat__original_aa_W", "cat__original_aa_Y", "cat__mutant_aa_A", "cat__mutant_aa_C", "cat__mutant_aa_D", "cat__mutant_aa_E", "cat__mutant_aa_F", "cat__mutant_aa_G", "cat__mutant_aa_H", "cat__mutant_aa_I", "cat__mutant_aa_K", "cat__mutant_aa_L", "cat__mutant_aa_M", "cat__mutant_aa_N", "cat__mutant_aa_P", "cat__mutant_aa_Q", "cat__mutant_aa_R", "cat__mutant_aa_S", "cat__mutant_aa_T", "cat__mutant_aa_V", "cat__mutant_aa_W", "cat__mutant_aa_Y", "cat__mutated_chain_0", "cat__mutated_chain_1", "cat__mutated_chain_A", "cat__mutated_chain_B", "cat__mutated_chain_E", "cat__mutated_chain_H", "cat__mutated_chain_I", "cat__mutated_chain_L", "cat__mutated_chain_X", "cat__mutated_chain_unsigned", "cat__mutation_type_Single", "cat__method_Acid", "cat__method_Activity", "cat__method_GSSG, GSH", "cat__method_GdmCl", "cat__method_Gdn-HCl", "cat__method_GdnCl", "cat__method_GdnCl.", "cat__method_GdnHCL", "cat__method_GdnHCl", "cat__method_GdnHCl, Urea", "cat__method_GdnSCN", "cat__method_GnHCl", "cat__method_GndHCl", "cat__method_GuHCI", "cat__method_GuHCl", "cat__method_NMR", "cat__method_Pressure", "cat__method_SDS", "cat__method_Thermal", "cat__method_Thermal, Urea", "cat__method_Unavailable", "cat__method_Urea", "cat__method_Urea,Thermal", "cat__method_pressure", "cat__method_thermal"]

final_multi_task_protein_stability_model_non_plm.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2512e198b0fd1a08e0f757c4f6c49e45a296ab4c60ee3e58afd4c5ddc92550a5
+size 777049

global_vars.json

@@ -0,0 +1,148 @@
+{
+    "label_encoder_classes": [
+        "destabilizing",
+        "neutral",
+        "stabilizing"
+    ],
+    "num_classes": 3,
+    "best_hyperparams": {
+        "learning_rate": 0.0001,
+        "dropout": 0.3,
+        "lambda_reg": 0.5,
+        "lambda_cls": 0.5
+    },
+    "expected_value_reg": -1.0198863744735718,
+    "expected_value_cls": [
+        -0.5417895317077637,
+        0.3860914409160614,
+        0.6462976932525635
+    ],
+    "AA_PROPERTIES": {
+        "A": {
+            "hydrophobicity": 1.8,
+            "volume": 88.6,
+            "charge": 0,
+            "polarity": 0
+        },
+        "R": {
+            "hydrophobicity": -4.5,
+            "volume": 173.4,
+            "charge": 1,
+            "polarity": 1
+        },
+        "N": {
+            "hydrophobicity": -3.5,
+            "volume": 114.1,
+            "charge": 0,
+            "polarity": 1
+        },
+        "D": {
+            "hydrophobicity": -3.5,
+            "volume": 111.1,
+            "charge": -1,
+            "polarity": 1
+        },
+        "C": {
+            "hydrophobicity": 2.5,
+            "volume": 108.5,
+            "charge": 0,
+            "polarity": 0
+        },
+        "Q": {
+            "hydrophobicity": -3.5,
+            "volume": 143.8,
+            "charge": 0,
+            "polarity": 1
+        },
+        "E": {
+            "hydrophobicity": -3.5,
+            "volume": 138.4,
+            "charge": -1,
+            "polarity": 1
+        },
+        "G": {
+            "hydrophobicity": -0.4,
+            "volume": 60.1,
+            "charge": 0,
+            "polarity": 0
+        },
+        "H": {
+            "hydrophobicity": -3.2,
+            "volume": 153.2,
+            "charge": 0.1,
+            "polarity": 1
+        },
+        "I": {
+            "hydrophobicity": 4.5,
+            "volume": 166.7,
+            "charge": 0,
+            "polarity": 0
+        },
+        "L": {
+            "hydrophobicity": 3.8,
+            "volume": 166.7,
+            "charge": 0,
+            "polarity": 0
+        },
+        "K": {
+            "hydrophobicity": -3.9,
+            "volume": 168.6,
+            "charge": 1,
+            "polarity": 1
+        },
+        "M": {
+            "hydrophobicity": 1.9,
+            "volume": 162.9,
+            "charge": 0,
+            "polarity": 0
+        },
+        "F": {
+            "hydrophobicity": 2.8,
+            "volume": 203.4,
+            "charge": 0,
+            "polarity": 0
+        },
+        "P": {
+            "hydrophobicity": -1.6,
+            "volume": 112.7,
+            "charge": 0,
+            "polarity": 0
+        },
+        "S": {
+            "hydrophobicity": -0.8,
+            "volume": 89.0,
+            "charge": 0,
+            "polarity": 1
+        },
+        "T": {
+            "hydrophobicity": -0.7,
+            "volume": 116.1,
+            "charge": 0,
+            "polarity": 1
+        },
+        "W": {
+            "hydrophobicity": -0.9,
+            "volume": 227.8,
+            "charge": 0,
+            "polarity": 1
+        },
+        "Y": {
+            "hydrophobicity": -1.3,
+            "volume": 203.0,
+            "charge": 0,
+            "polarity": 1
+        },
+        "V": {
+            "hydrophobicity": 4.2,
+            "volume": 117.5,
+            "charge": 0,
+            "polarity": 0
+        },
+        "X": {
+            "hydrophobicity": 0,
+            "volume": 0,
+            "charge": 0,
+            "polarity": 0
+        }
+    }
+}

label_encoder.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f56301b90a626279392c84a763c2e4793164d9361258aaf3ba65569120e4597
+size 283

preprocessor.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7b482f0df5b1269ea593a189dc4dbc321c7f24b7d45a2bfb037fae1e18c392a
+size 16417

requirements.txt

@@ -0,0 +1,10 @@
+pandas
+numpy
+scikit-learn
+torch
+gradio==4.36.1
+matplotlib
+seaborn
+shap
+requests
+huggingface_hub

shap_background_data.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bec90c6f95e7fc71e3bdd4b221dc1d923cff3dac97284859b9e8754419c1ab65
+size 218820

thermomutdb.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54d53ad4aae1b402a3fd7b5a470edcf082dc499509023c70d8a998aecffca68b
+size 17047379