initial commit

app.py

@@ -0,0 +1,158 @@
+import torch
+import torch.nn.functional as F  # Importing the functional module for softmax
+from torch import nn
+import gradio as gr
+import os
+
+# Assume model and tokenizer are already loaded
+
+def predict_drug_target_interaction(sentence):
+    # Tokenize the input sentence
+    inputs = tokenizer.encode_plus(
+        sentence,
+        add_special_tokens=True,
+        max_length=128,
+        padding="max_length",
+        truncation=True,
+        return_tensors='pt'
+    )
+
+    input_ids = inputs['input_ids'].to(device)
+    attention_mask = inputs['attention_mask'].to(device)
+    token_type_ids = inputs.get('token_type_ids', None)
+    if token_type_ids is not None:
+        token_type_ids = token_type_ids.to(device)
+
+    # Make prediction
+    with torch.inference_mode():
+        outputs = model(input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)
+        logits = outputs.logits
+        probabilities = F.softmax(logits, dim=-1)  # Calculate softmax probabilities
+        predictions = torch.argmax(logits, dim=-1)
+
+    # Convert predictions to human-readable labels
+    label = 'Drug-Target Interaction' if predictions.item() == 1 else 'No Interaction'
+    
+    # Extract probabilities for both classes
+    prob_interaction = probabilities[0][1].item()  # Probability for Drug-Target Interaction
+    prob_no_interaction = probabilities[0][0].item()  # Probability for No Interaction
+
+    return label, prob_interaction, prob_no_interaction
+
+# Gradio Interface
+def combined_prediction_and_extraction(sentence):
+    # Get the prediction and probabilities
+    label, prob_interaction, prob_no_interaction = predict_drug_target_interaction(sentence)
+
+    if prob_interaction > prob_no_interaction:
+      response = "There is a possible interaction between drug and target in the given input! ✔️"
+    elif prob_interaction < prob_no_interaction:
+      response = "There is NO interaction between the drug and target in the given input. ❌"
+    else:
+      response = "This interaction needs further studies to classify. 🔬"
+
+    
+    pred_labels_and_probs_gradio = {"Drug Interaction": prob_interaction, "No Interaction": prob_no_interaction}
+    
+    return pred_labels_and_probs_gradio, response
+
+# Description for the new tab
+description = """
+# The Importance of Drug-Target Interactions in Pharmaceutical Development
+
+## 🌟 **Mechanism of Action**
+Understanding how a drug interacts with its target—usually a protein, enzyme, or receptor—is essential for uncovering its mechanism of action. This insight helps researchers grasp the therapeutic effects of drugs, paving the way for the development of **more effective treatments**.
+
+## 🎯 **Selectivity and Specificity**
+Targeting specific proteins and pathways minimizes side effects and boosts drug efficacy. **High specificity** ensures that the drug primarily interacts with its intended target, reducing the risk of off-target effects that could lead to adverse reactions.
+
+## 💡 **Drug Design and Development**
+Knowledge of drug-target interactions is crucial for crafting new pharmaceuticals. **Optimizing lead compounds** to enhance their affinity for the target can significantly elevate a drug's effectiveness, leading to **innovative treatment solutions**.
+
+## 🔍 **Predicting Pharmacokinetics and Pharmacodynamics**
+Understanding these interactions aids in predicting how a drug behaves in the body, including its absorption, distribution, metabolism, and excretion. This knowledge is vital for determining **appropriate dosages** and anticipating potential drug-drug interactions.
+
+## 🛡️ **Resistance Mechanisms**
+In fields like oncology and infectious diseases, drug-target interactions are crucial for understanding how **resistance** to treatments develops. By studying these interactions, researchers can devise strategies to **overcome or prevent resistance**, ensuring better patient outcomes.
+
+## ⚠️ **Safety and Toxicity**
+Understanding how drugs interact with unintended targets is essential for assessing **safety profiles**. This information is key to identifying potential toxic effects and **mitigating risks** associated with drug therapy.
+
+"""
+
+datainfo = """
+The most prominent node in the graph is metformin, which is linked to various targets. This suggests that metformin has a wide range of biological interactions, possibly indicating its involvement in multiple pathways or therapeutic effects beyond its traditional use as an anti-diabetic drug.
+
+* Metformin interacts with AMP-activated protein kinase (AMPK), which is known to play a crucial role in cellular energy homeostasis. This aligns with metformin’s known mechanism of action in diabetes, where it helps modulate glucose metabolism through AMPK activation.
+Other significant metformin interactions include ROR1.69 and nucleoside, indicating additional pathways of relevance, possibly connected to immune modulation or other metabolic pathways.
+
+* Warfarin, an anticoagulant, is another highly connected node. Its targets include CYP2C9, an enzyme critical for its metabolism, and S-warfarin, which represents its active form. The network also includes its interaction with estrogen, AXIN1, and monoamine oxidase A, indicating potential off-target effects or interactions with other biological pathways.
+
+* The interaction between serotonin transporter protein and drugs like norepinephrine and serotonin indicates involvement in pathways related to mood regulation, which could be pertinent in psychiatric conditions.
+
+> The network shows that some drugs share common targets. Both metformin and warfarin interact with multiple common targets (like estrogen). This overlap could imply the potential for drug-drug interactions in patients taking both medications, possibly requiring careful management of therapy.
+"""
+
+
+def display_image():
+    return "DTI_knowledge graph highlighted.png"  # Change this to your image path or URL
+
+
+
+title = "Unlocking Precision: Predicting Drug-Target Interactions for Safer, Smarter Treatments"
+article = "Stay Safe!"
+examples = [
+    ["Targeting the secretion of sEV PD-L1 has emerged as a promising strategy to enhance immunotherapy"],
+    ["These results suggested that AM would be a useful platform for the development of a new radiopharmaceutical targeting ER"],
+    ["AP significantly depended on a three-way interaction among the mouse group ORX-KO vs WT, the wake-sleep state, and the drug or vehicle infused."],
+    ["In addition, the pharmacodynamic genes SLC6A4 serotonin transporter and HTR2A serotonin-2A receptor have been examined in relation to efficacy and side effect profiles of these drugs"]
+]
+
+
+with gr.Blocks() as demo:
+    gr.Markdown(f"# {title}")
+
+    with gr.Tab("Model Inference"):
+        with gr.Row():
+            with gr.Column(scale=1, min_width=300):
+                input_text = gr.Textbox(label="Input", placeholder="Enter your text here...")
+                gr.Markdown("### Examples")
+                gr.Examples(
+                    examples=examples,
+                    inputs=input_text,
+                    outputs=None,  # Outputs are handled by the button click
+                    label="Select an Example"
+                )
+                
+                # Create a Row for Submit and Clear buttons
+                with gr.Row():
+                    submit_button = gr.Button("Submit")
+                    clear_button = gr.Button("Clear")
+
+            with gr.Column(scale=1, min_width=300):
+                predictions_output = gr.Label(num_top_classes=2, label="Predictions")
+                inference_output = gr.Textbox(label="Prediction Inference", interactive=False)  # Making it read-only
+
+                # Bind the button to the prediction function
+                submit_button.click(
+                    fn=combined_prediction_and_extraction,
+                    inputs=input_text,
+                    outputs=[predictions_output, inference_output]
+                )
+
+                # Bind the clear button to reset the input field and outputs
+                clear_button.click(
+                    fn=lambda: ("", "", ""),  # Reset input and outputs
+                    inputs=None,
+                    outputs=[input_text, predictions_output, inference_output]
+                )
+    with gr.Tab("Description"):
+      gr.Markdown(description)
+    
+    with gr.Tab("About Data"):
+      gr.Markdown(datainfo)
+      gr.Image(value=display_image(), type="filepath")
+
+
+# Launch the interface
+demo.launch()