Drug Development Platform Backend 🧬¶

Overview 🎯¶

Backend for the Drug Development Platform, leveraging Azure AI Foundry SDKs for molecular analysis and clinical trial monitoring.

Azure AI Integration 🤖¶

SDKs Used¶

• 🎯 azure-ai-projects: Project and agent management

  # main.py
  project_client = AIProjectClient.from_connection_string(
      credential=credential,
      conn_str=PROJECT_CONNECTION_STRING
  )
  

• 🔬 azure-ai-inference: Molecular analysis and predictions

  # molecular_design.py
  inference_client = project_client.inference.get_chat_completions_client()
  

• 📊 azure-ai-evaluation: Result analysis and confidence scoring

  # automated_testing.py
  evaluation = Evaluation(
      display_name="Drug Analysis Evaluation",
      description="Evaluation of drug analysis outputs",
      data=Dataset(id=data_id),
      evaluators={
          "f1_score": EvaluatorConfiguration(id=F1ScoreEvaluator.id),
          "relevance": EvaluatorConfiguration(id=RelevanceEvaluator.id)
      }
  )
  

• 📈 opentelemetry-sdk: Performance monitoring and tracing

  # main.py
  trace.set_tracer_provider(TracerProvider())
  tracer = trace.get_tracer(__name__)
  

• 🔐 azure-identity: Secure Azure authentication

  # main.py
  credential = DefaultAzureCredential()
  

Getting Started 🚀¶

1. Create Virtual Environment:

   python -m venv venv
   source venv/bin/activate  # Windows: venv\Scripts\activate
   

2. Install Dependencies:

   pip install -r requirements.txt
   

3. Configure Environment:

4. Copy .env.example to .env
5. Update with your Azure credentials:

   # Azure Authentication (DO NOT commit actual values!)
   AZURE_CLIENT_ID=your_client_id_here
   AZURE_CLIENT_SECRET=your_client_secret_here
   AZURE_TENANT_ID=your_tenant_id_here
   
   # AI Foundry Configuration (use your project-specific values)
   PROJECT_CONNECTION_STRING=your_connection_string_here
   MODEL_DEPLOYMENT_NAME=your_model_name_here
   
   # OpenTelemetry Settings (adjust based on your environment)
   OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
   OTEL_SERVICE_NAME=drug-discovery-platform
   OTEL_RESOURCE_ATTRIBUTES=deployment.environment=development
   

⚠️ Security Note: Never commit real credentials or sensitive information to version control. Always use environment variables or secure secret management solutions in production.

Required Environment Variables: - AZURE_CLIENT_ID: Azure AD application ID for authentication - AZURE_CLIENT_SECRET: Azure AD application secret for authentication - AZURE_TENANT_ID: Azure AD tenant ID for authentication - PROJECT_CONNECTION_STRING: AI Foundry project connection string for agent and model access - MODEL_DEPLOYMENT_NAME: Name of your deployed model for agent creation - OTEL_EXPORTER_OTLP_ENDPOINT: OpenTelemetry collector endpoint for tracing - OTEL_SERVICE_NAME: Service name for tracing (default: drug-discovery-platform) - OTEL_RESOURCE_ATTRIBUTES: Additional tracing attributes (default: deployment.environment=development)

1. Run the Server:

   uvicorn main:app --reload
   

2. Access API Documentation:

3. OpenAPI: http://localhost:8000/docs
4. ReDoc: http://localhost:8000/redoc

API Architecture 🏗️¶

flowchart TB
    subgraph Client
        FE[Frontend]
    end

    subgraph Backend
        API[FastAPI]
        AI[Azure AI Clients]
        TEL[OpenTelemetry]
    end

    subgraph Azure["Azure AI Foundry"]
        PRJ[AI Projects]
        AGT[AI Agents]
        INF[AI Inference]
        EVAL[AI Evaluation]
        subgraph Agents["AI Agent Tools"]
            BING[Bing Grounding]
            FUNC[Function Calling]
            CODE[Code Interpreter]
        end
    end

    FE <--> API
    API <--> AI
    AI <--> PRJ
    PRJ <--> AGT
    AGT <--> BING
    AGT <--> FUNC
    AGT <--> CODE
    PRJ <--> INF
    PRJ <--> EVAL
    API --> TEL

Endpoints 🛠️¶

Molecular Design Endpoints¶

1. Analyze Molecule¶

POST /molecular-design/analyze

Request:

{
    "id": "DRUG-001",
    "molecule_type": "Small Molecule",
    "therapeutic_area": "Oncology",
    "target_proteins": ["EGFR", "HER2"],
    "development_stage": "Phase 1"
}

Response:

{
    "message": "Molecular analysis complete",
    "analysis": {
        "efficacy_score": 0.85,
        "safety_score": 0.92,
        "confidence": 0.89
    }
}

Azure AI Agent Endpoints¶

1. Literature Search¶

POST /agents/literature-search

Request:

{
    "query": "EGFR inhibitors in lung cancer"
}

Response:

{
    "query": "EGFR inhibitors in lung cancer",
    "summary": "Analysis of recent publications...",
    "agent_id": "agent-123"
}

2. Molecule Analysis¶

POST /agents/molecule-analysis

Request:

{
    "smiles": "CC1=CC=C(C=C1)NC(=O)C2=CC=C(Cl)C=C2",
    "target_proteins": ["EGFR", "HER2"],
    "therapeutic_area": "Oncology"
}

Response:

{
    "molecule": "CC1=CC=C(C=C1)NC(=O)C2=CC=C(Cl)C=C2",
    "analysis": "Detailed molecular property analysis...",
    "agent_id": "agent-456"
}

3. Clinical Trial Data Analysis¶

POST /agents/data-analysis

Request:

multipart/form-data
- file: trial_data.csv

Response:

{
    "filename": "trial_data.csv",
    "analysis": "Statistical analysis and visualizations...",
    "agent_id": "agent-789"
}

Clinical Trials Endpoints¶

1. Monitor Trials¶

GET /clinical-trials/monitor?trial_id=TRIAL-001

Response:

{
    "trial_id": "TRIAL-001",
    "phase": "Phase 2",
    "status": "Active",
    "real_time_metrics": {
        "enrollment_rate": 0.75,
        "retention_rate": 0.92,
        "safety_signals": []
    }
}

2. Predict Patient Response¶

POST /clinical-trials/predict-response

Request:

{
    "trial_id": "TRIAL-001",
    "patient_id": "PAT-001"
}

Response:

{
    "predicted_response": 0.85,
    "confidence": 0.92,
    "recommendations": [
        "Continue monitoring key biomarkers",
        "Schedule follow-up in 2 weeks"
    ]
}

Automated Testing Endpoints¶

1. Run Evaluation Demo¶

sequenceDiagram
    participant Client
    participant API
    participant Evaluator
    participant Dataset

    Client->>API: POST /evaluation/run-demo
    API->>Dataset: Upload JSONL Data
    API->>Evaluator: Configure Evaluation
    Evaluator->>Evaluator: Run F1 Score
    Evaluator->>Evaluator: Run Relevance
    Evaluator-->>API: Results
    API-->>Client: Metrics

OpenTelemetry Integration 📊¶

This project uses OpenTelemetry for distributed tracing to monitor and debug the drug development pipeline. Traces help us understand:

• 🔍 Performance bottlenecks
• 🔗 Request flows through the system
• ❌ Error patterns and their context
• 📈 AI model inference timing

Viewing Traces¶

1. Traces are collected by the OpenTelemetry collector at:

   http://localhost:4318/v1/traces
   

2. Key spans to monitor:

3. molecular_design.analyze: Molecule analysis and AI inference
4. clinical_trials.monitor: Trial monitoring and metrics

5. clinical_trials.predict_response: Patient response predictions

6. Important attributes in traces:

7. molecule.id: Unique identifier for drug candidates
8. molecule.type: Type of molecule being analyzed
9. therapeutic.area: Target therapeutic area
10. analysis.efficacy: Predicted efficacy score
11. analysis.safety: Safety assessment score
12. trial.id: Clinical trial identifier
13. patient.id: Patient identifier for specific analyses

Example Trace Analysis¶

sequenceDiagram
    participant C as Client
    participant A as API
    participant AI as Azure AI
    participant DB as Database

    C->>A: POST /molecular-design/analyze
    activate A
    A->>AI: Analyze molecule
    AI-->>A: Inference results
    A->>DB: Store results
    A-->>C: Analysis response
    deactivate A

Security 🔒¶

1. Environment Variables:
2. All secrets stored in .env

3. Never commit sensitive data

4. Authentication:

5. Azure AD integration

6. Role-based access control

7. Data Protection:

8. Encryption at rest
9. Secure communication

Contributing 🤝¶

1. Fork the repository
2. Create a feature branch
3. Submit a Pull Request

Setup and Running¶

1. Install dependencies:

   pip install -r requirements.txt
   

2. Start the server (runs on port 8002):

Windows (PowerShell):

.\start.ps1

Unix/Linux/MacOS:

./start.sh

Environment Variables¶

The backend server uses the following environment variables:

• PORT: Server port (defaults to 8002)
• PROJECT_CONNECTION_STRING: Azure project connection string
• MODEL_DEPLOYMENT_NAME: Azure model deployment name
• Other environment variables as specified in main.py

API Documentation¶

Once the server is running, you can access the API documentation at: - http://localhost:8002/ (ReDoc interface)

Development¶

The server includes: - FastAPI framework - OpenTelemetry instrumentation - CORS middleware configured for frontend at http://localhost:3000 - Health check endpoint at /health

Deployment with Azure Developer CLI (azd) 🚀¶

1. Install Azure Developer CLI:

   curl -fsSL https://aka.ms/install-azd.sh | bash
   

2. Login to Azure:

   azd auth login
   

3. Deploy from Root Directory: From the root of "3-e2e-drug-discovery-sample/", run:

   azd init
   azd up
   

This will: - Create Azure Container App for the backend - Configure environment variables from your .env - Deploy the backend service - Output the public endpoint URL

After deployment: - Access API documentation at https://<your-app>.azurecontainerapps.io/docs - Use the endpoint URL in your frontend configuration

Learn More 📚¶

• Azure AI Projects SDK
• Azure AI Inference SDK
• Azure AI Evaluation SDK
• OpenTelemetry Documentation
• FastAPI Documentation