Upload 3 files

app.py

@@ -0,0 +1,165 @@
+import streamlit as st
+import torch
+import numpy as np
+import pandas as pd
+
+from torch_geometric.nn import SAGEConv, global_mean_pool
+from torch_geometric.data import Batch
+from torch_geometric.utils import from_smiles
+
+# ===============================================================
+# Model Definition (MUST MATCH TRAINING)
+# ===============================================================
+class MolEncoderSAGE(torch.nn.Module):
+    def __init__(self, in_dim=9, hidden=128, layers=3):
+        super().__init__()
+        self.convs = torch.nn.ModuleList()
+        self.convs.append(SAGEConv(in_dim, hidden))
+        for _ in range(layers - 1):
+            self.convs.append(SAGEConv(hidden, hidden))
+
+    def forward(self, data):
+        x = data.x.float()
+        edge_index = data.edge_index
+        batch = data.batch
+
+        for conv in self.convs:
+            x = torch.relu(conv(x, edge_index))
+
+        return global_mean_pool(x, batch)
+
+
+class SolvSAGENet(torch.nn.Module):
+    def __init__(self, hidden=128, layers=3, dropout=0.1):
+        super().__init__()
+        self.solute = MolEncoderSAGE(9, hidden, layers)
+        self.solvent = MolEncoderSAGE(9, hidden, layers)
+
+        self.mlp = torch.nn.Sequential(
+            torch.nn.Linear(2 * hidden, 256),
+            torch.nn.ReLU(),
+            torch.nn.Dropout(dropout),
+            torch.nn.Linear(256, 128),
+            torch.nn.ReLU(),
+            torch.nn.Linear(128, 1)
+        )
+
+    def forward(self, s, v):
+        z = torch.cat([self.solute(s), self.solvent(v)], dim=1)
+        return self.mlp(z)
+
+# ===============================================================
+# Load Model (cached)
+# ===============================================================
+@st.cache_resource
+def load_model():
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    model = SolvSAGENet(hidden=128, layers=3, dropout=0.1).to(device)
+    model.load_state_dict(torch.load("best_solv_sage.pth", map_location=device))
+    model.eval()
+    return model, device
+
+model, device = load_model()
+
+# ===============================================================
+# Streamlit UI
+# ===============================================================
+st.set_page_config(page_title="ΔG_solv Prediction", layout="centered")
+
+st.title("🔬 ΔGₛₒₗᵥ Prediction (GraphSAGE)")
+st.markdown("""
+Enter **solute and solvent SMILES** to predict  
+**Solvation Free Energy (ΔGₛₒₗᵥ)** in kcal/mol.
+""")
+
+# ===============================================================
+# Single Prediction
+# ===============================================================
+st.header("🧪 Single Prediction")
+
+solute_smiles = st.text_input(
+    "Solute SMILES",
+    value="CCO",
+    help="Example: CCO (ethanol)"
+)
+
+solvent_smiles = st.text_input(
+    "Solvent SMILES",
+    value="O",
+    help="Example: O (water)"
+)
+
+if st.button("Predict ΔGₛₒₗᵥ"):
+    try:
+        # Convert SMILES → graphs
+        solute_graph = from_smiles(solute_smiles)
+        solvent_graph = from_smiles(solvent_smiles)
+
+        # Create batch
+        solute_batch = Batch.from_data_list([solute_graph]).to(device)
+        solvent_batch = Batch.from_data_list([solvent_graph]).to(device)
+
+        # Predict
+        with torch.no_grad():
+            prediction = model(solute_batch, solvent_batch).item()
+
+        st.success(f"✅ Predicted ΔGₛₒₗᵥ: **{prediction:.3f} kcal/mol**")
+
+    except Exception as e:
+        st.error("❌ Invalid SMILES or model error")
+        st.write(e)
+
+# ===============================================================
+# Batch Prediction
+# ===============================================================
+st.header("📂 Batch Prediction (CSV Upload)")
+
+st.markdown("""
+Upload a CSV file with **columns**:
+- `mol_solute`
+- `mol_solvent`
+""")
+
+uploaded_file = st.file_uploader("Upload CSV", type=["csv"])
+
+if uploaded_file:
+    df = pd.read_csv(uploaded_file)
+
+    if {"mol_solute", "mol_solvent"}.issubset(df.columns):
+        predictions = []
+
+        with torch.no_grad():
+            for _, row in df.iterrows():
+                try:
+                    s = from_smiles(row["mol_solute"])
+                    v = from_smiles(row["mol_solvent"])
+
+                    sb = Batch.from_data_list([s]).to(device)
+                    vb = Batch.from_data_list([v]).to(device)
+
+                    pred = model(sb, vb).item()
+                    predictions.append(pred)
+                except:
+                    predictions.append(np.nan)
+
+        df["predicted_Gsolv"] = predictions
+
+        st.dataframe(df)
+
+        st.download_button(
+            label="⬇️ Download Predictions",
+            data=df.to_csv(index=False),
+            file_name="predicted_gsolv.csv",
+            mime="text/csv"
+        )
+    else:
+        st.error("CSV must contain columns: mol_solute, mol_solvent")
+
+# ===============================================================
+# Footer
+# ===============================================================
+st.markdown("---")
+st.markdown(
+    "🧠 **Graph Neural Network (GraphSAGE)**  \n"
+    "🔗 PyTorch Geometric | Molecular ML"
+)

best_solv_sage.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:114d73ec142074f3c79013af6169de54ed1730f7e6876fa155a3ef9a9ae5a21c
+size 950317

requirements.txt

@@ -0,0 +1,6 @@
+streamlit
+torch
+torch-geometric
+pandas
+numpy
+scikit-learn