Integrate real AlphaGenome API with fallback

app.py

@@ -3,6 +3,15 @@ import pandas as pd
 import numpy as np
 import matplotlib.pyplot as plt
 
+# Try importing AlphaGenome, handle failure gracefully
+try:
+    from alphagenome.models import dna_client
+    from alphagenome.data import genome
+    from alphagenome.visualization import plot_components
+    HAS_ALPHAGENOME = True
+except ImportError:
+    HAS_ALPHAGENOME = False
+
 # Page Config
 st.set_page_config(
     page_title="Virtual Gene Scope",
@@ -10,25 +19,13 @@ st.set_page_config(
     layout="wide"
 )
 
-# Custom CSS for Nursing/Medical Aesthetic
+# Custom CSS
 st.markdown("""
     <style>
-    .main {
-        background-color: #f8f9fa;
-    }
-    h1 {
-        color: #2c3e50;
-    }
-    .stButton>button {
-        background-color: #007bff;
-        color: white;
-    }
-    .metric-card {
-        background-color: white;
-        padding: 20px;
-        border-radius: 10px;
-        box-shadow: 0 4px 6px rgba(0,0,0,0.1);
-    }
+    .main { background-color: #f8f9fa; }
+    h1 { color: #2c3e50; }
+    .stButton>button { background-color: #007bff; color: white; }
+    .metric-card { background-color: white; padding: 20px; border-radius: 10px; box-shadow: 0 4px 6px rgba(0,0,0,0.1); }
     </style>
 """, unsafe_allow_html=True)
 
@@ -36,7 +33,7 @@ st.markdown("""
 st.title("🧬 Virtual Gene Scope: The Regulatory Genome")
 st.markdown("""
 **Nursing Context**: 98% of the genome does not code for proteins. Instead, it acts as a "control panel" turning genes on and off.
-**AlphaGenome** allows us to see this hidden layer. Use this tool to visualize how non-coding mutations affects patient health.
+**AlphaGenome** allows us to see this hidden layer.
 """)
 
 # Sidebar controls
@@ -44,53 +41,103 @@ st.sidebar.header("Patient & Gene Settings")
 gene_choice = st.sidebar.selectbox("Select Gene of Interest", ["INS (Insulin - Diabetes)", "SCN9A (Pain Sensitivity)", "MMP9 (Wound Healing)"])
 mutation_intensity = st.sidebar.slider("Mutation Impact (Simulation)", 0.0, 1.0, 0.5, help="Simulate the severity of the regulatory disruption.")
 
-# Mock Data Generation (Simulating AlphaGenome Output)
-def get_mock_tracks(gene, mutation_factor):
-    # Base signal (Promoter activity)
-    x = np.linspace(0, 100, 500)
-    
-    # "Normal" gene expression peak
-    base_signal = np.exp(-((x - 50)**2) / 20) 
-    
-    # "Mutated" signal - reduced by mutation factor
-    mutated_signal = base_signal * (1 - mutation_factor * 0.8) # up to 80% reduction
+# Setup API
+api_key = st.secrets.get("ALPHAGENOME_API_KEY")
+
+if not HAS_ALPHAGENOME:
+    st.error("AlphaGenome library not installed. Please check requirements.txt")
+elif not api_key:
+    st.warning("⚠️ No API Key found! Please add `ALPHAGENOME_API_KEY` to your secrets to see real data. Showing MOCK data for now.")
     
-    return x, base_signal, mutated_signal
+    # MOCK FALLBACK
+    def get_mock_tracks(gene, mutation_factor):
+        x = np.linspace(0, 100, 500)
+        base_signal = np.exp(-((x - 50)**2) / 20) 
+        mutated_signal = base_signal * (1 - mutation_factor * 0.8) 
+        return x, base_signal, mutated_signal
 
-# Main Content Area
-col1, col2 = st.columns([1, 2])
+    col1, col2 = st.columns([1, 2])
+    with col1:
+        st.subheader("Variant details (MOCK)")
+        st.info(f"Analyzing Regulatory Region for **{gene_choice.split(' ')[0]}**")
+        st.write("Variant Type: **Non-Coding / Regulatory**")
+        if mutation_intensity > 0.7:
+            st.error("⚠️ HIGH RISK: Significant reduction in gene expression predicted.")
+        else:
+            st.success("✅ LOW RISK: Benign variant predicted.")
 
-with col1:
-    st.subheader("Variant details")
-    st.info(f"Analyzing Regulatory Region for **{gene_choice.split(' ')[0]}**")
-    
-    st.write("Variant Type: **Non-Coding / Regulatory**")
-    st.write("Location: **Promoter / Enhancer Region**")
-    
-    if mutation_intensity > 0.7:
-        st.error("⚠️ HIGH RISK: Significant reduction in gene expression predicted.")
-    elif mutation_intensity > 0.3:
-        st.warning("⚠️ MODERATE RISK: Partial loss of regulation.")
-    else:
-        st.success("✅ LOW RISK: Benign variant predicted.")
+    with col2:
+        st.subheader("AlphaGenome Predicted Activity (MOCK)")
+        x, normal, mutant = get_mock_tracks(gene_choice, mutation_intensity)
+        fig, ax = plt.subplots(figsize=(10, 5))
+        ax.plot(x, normal, label="Healthy Control", color='green')
+        ax.plot(x, mutant, label="Patient Variant", color='red', linestyle='--')
+        ax.legend()
+        st.pyplot(fig)
 
-with col2:
-    st.subheader("AlphaGenome Predicted Activity Tracks")
+else:
+    # REAL API LOGIC
+    st.sidebar.success("✅ AlphaGenome API Connected")
     
-    x, normal, mutant = get_mock_tracks(gene_choice, mutation_intensity)
+    # Initialize Client
+    model = dna_client.create(api_key)
     
-    fig, ax = plt.subplots(figsize=(10, 5))
-    ax.plot(x, normal, label="Healthy Control (Reference)", color='green', linewidth=2, alpha=0.7)
-    ax.plot(x, mutant, label="Patient Variant (Alternative)", color='red', linewidth=2, linestyle='--')
+    # Define dummy coordinates for the demo genes (simplified)
+    # in a real app, you'd look these up properly
+    gene_coords = {
+        "INS": {"chr": "chr11", "start": 2159779, "end": 2161221, "pos": 2160000},
+        "SCN9A": {"chr": "chr2", "start": 166195190, "end": 166205190, "pos": 166200000}, # rounded
+        "MMP9": {"chr": "chr20", "start": 46010000, "end": 46020000, "pos": 46015000}
+    }
     
-    ax.set_title(f"Promoter Activity: {gene_choice.split(' ')[0]}", fontsize=12)
-    ax.set_xlabel("Genomic Position (bp)")
-    ax.set_ylabel("Predicted RNA Expression")
-    ax.legend()
-    ax.grid(True, alpha=0.3)
-    ax.fill_between(x, normal, mutant, color='orange', alpha=0.2, label='Lost Expression')
+    gene_sym = gene_choice.split(' ')[0]
+    coords = gene_coords.get(gene_sym, gene_coords["INS"])
     
-    st.pyplot(fig)
+    try:
+        with st.spinner("Querying Google DeepMind TPUs..."):
+            interval = genome.Interval(chromosome=coords["chr"], start=coords["start"], end=coords["end"])
+            # Create a dummy variant at the center
+            variant = genome.Variant(
+                chromosome=coords["chr"], 
+                position=coords["pos"], 
+                reference_bases='A', 
+                alternate_bases='C' # simple mutation
+            )
+            
+            # Predict
+            outputs = model.predict_variant(
+                interval=interval,
+                variant=variant,
+                requested_outputs=[dna_client.OutputType.RNA_SEQ],
+            )
+            
+            col1, col2 = st.columns([1, 2])
+            with col1:
+                st.subheader("Real Analysis")
+                st.markdown(f"**Gene**: {gene_sym}")
+                st.markdown(f"**Locus**: {coords['chr']}:{coords['start']}-{coords['end']}")
+                st.success("Prediction successful!")
+                
+            with col2:
+                st.subheader("AlphaGenome Tracks")
+                # Use plot_components from the library
+                fig = plot_components.plot(
+                    [
+                        plot_components.OverlaidTracks(
+                            tdata={
+                                'REF': outputs.reference.rna_seq,
+                                'ALT': outputs.alternate.rna_seq,
+                            },
+                            colors={'REF': 'dimgrey', 'ALT': 'red'},
+                        ),
+                    ],
+                    interval=outputs.reference.rna_seq.interval.resize(2**15),
+                    annotations=[plot_components.VariantAnnotation([variant], alpha=0.8)],
+                )
+                st.pyplot(fig)
+                
+    except Exception as e:
+        st.error(f"API Error: {str(e)}")
 
 st.markdown("---")
-st.caption("Powered by AlphaGenome Architecture (Mock Demo) | Nursing Genomics Education Toolkit")
+st.caption("Powered by AlphaGenome Architecture | Nursing Genomics Education Toolkit")

requirements.txt

@@ -3,4 +3,4 @@ pandas
 matplotlib
 numpy
 huggingface_hub
-# alphagenome
+git+https://github.com/google-deepmind/alphagenome.git