Update app.py

app.py

@@ -1133,17 +1133,75 @@ with gr.Blocks(css=css, title="K R&D Lab · S1 Biomedical") as demo:
                 # === Learning ===
                 with gr.TabItem("📚 Learning"):
                     gr.Markdown("""
-## 🧪 Guided Investigations
+## 🧪 Guided Investigations — S1 Biomedical
 > 🟢 Beginner → 🟡 Intermediate → 🔴 Advanced
 
-**S1-A·R1a** OpenVariant – try BRCA1:p.R1699Q vs R1699W  
-**S1-D·R1a** Corona – vary PEG% and observe dominant protein  
-**S1-D·R2a** Flow Corona – compare flow 0 vs 40 cm/s  
-**S1-B·R2a** siRNA – count "Novel" targets across cancer types  
-**S1-E·R1a** Liquid Biopsy – find minimal signal for CANCER  
-**S1-G·3D Lab** – explore nanoparticle, DNA, and protein corona models
-                    """)
+---
+### 🟢 Case 1 · S1-A·R1a
+**Why does the same position give two different outcomes?**
+1. Go to **S1-A · R1a · OpenVariant**.
+2. Enter `BRCA1:p.R1699Q` → you get **Benign**.
+3. Enter `BRCA1:p.R1699W` → you get **Pathogenic**.
+4. Same codon, different amino acid — Q (polar, neutral) vs W (bulky, aromatic).  
+   *This illustrates how a single nucleotide change can radically alter pathogenicity.*
 
+---
+### 🟢 Case 2 · S1-D·R1a + S1-D·R3a
+**How does PEG density control which protein forms the corona?**
+1. Go to **S1-D · R1a · LNP Corona**.
+2. Set: Size=100 nm, Zeta=-5 mV, Lipid=Ionizable, PEG=**0.5%** → note the dominant protein.
+3. Change PEG to **2.5%** → run again → dominant protein changes.
+4. Now go to **S1-D · R3a · LNP Brain**, use pKa≈6.5, Zeta≈-3 mV → observe ApoE%.  
+   *Higher PEG shields the surface, reducing ApoE adsorption and brain targeting.*
+
+---
+### 🟡 Case 3 · S1-D·R2a
+**Does blood flow change the corona composition over time?**
+1. Go to **S1-D · R2a · Flow Corona**.
+2. Set Flow = 0 (static) → run → note when ApoE becomes dominant (≈ ? min).
+3. Set Flow = 40 cm/s (arterial) → run again → compare curves.  
+   *Flow accelerates the Vroman effect: ApoE dominates earlier under flow.*
+
+---
+### 🟡 Case 4 · S1-B·R2a
+**Which cancer type has the most novel (undrugged) siRNA targets?**
+1. Go to **S1-B · R2a · siRNA**.
+2. Select LUAD → count how many targets are marked "Novel".
+3. Repeat for BRCA and COAD.  
+   *Novel targets have no approved drug – they represent high‑opportunity research areas.*
+
+---
+### 🔴 Case 5 · S1-E·R1a
+**What is the minimum protein signal that flips the classifier to CANCER?**
+1. Go to **S1-E · R1a · Liquid Biopsy**.
+2. Set all sliders to 0 → result = HEALTHY.
+3. Increase only CTHRC1 step by step (e.g., 0.5, 1.0, 1.5…) until the label becomes CANCER.
+4. Reset and try the same with FHL2 or LDHA.  
+   *CTHRC1 has the highest weight; you need ≈2.5 to cross the threshold.*
+
+---
+### 🔴 Case 6 · Cross‑tool convergence
+**Do different RNA tools point to the same cancer drivers?**
+1. Go to **S1-B · R1a · miRNA** → select TP53 → note top targets (BCL2, CDK6).
+2. Go to **S1-C · R1a · FGFR3** → look for CDK6 in the pathway column.
+3. Go to **S1-B · R2a · siRNA** → select BRCA → check if CDK6 appears.  
+   *CDK6 is a common node – targeted by miRNAs, siRNAs, and existing drugs.*
+
+---
+### 📖 Tool Index
+| Code | Module | Tool | Metric |
+|------|--------|------|--------|
+| S1-A·R1a | PHYLO-GENOMICS | OpenVariant | AUC=0.939 |
+| S1-B·R1a | PHYLO-RNA | miRNA silencing | top: hsa-miR-148a |
+| S1-B·R2a | PHYLO-RNA | siRNA SL | SPC24 top LUAD |
+| S1-B·R3a | PHYLO-RNA | lncRNA-TREM2 | CYTOR→AKT1 |
+| S1-C·R1a | PHYLO-DRUG | FGFR3 drug | score=0.793 |
+| S1-D·R1a | PHYLO-LNP | Corona | AUC=0.791 |
+| S1-D·R2a | PHYLO-LNP | Flow Vroman | 3–4× faster |
+| S1-D·R3a | PHYLO-LNP | LNP Brain | pKa 6.2–6.8 |
+| S1-E·R1a | PHYLO-BIOMARKERS | Liquid Biopsy | AUC=0.992* |
+| S1-D·R4a | PHYLO-LNP | AutoCorona NLP | F1=0.71 |
+""")
                 # === Journal (окрема вкладка) ===
                 with gr.TabItem("📓 Journal"):
                     gr.Markdown("## Lab Journal — Full History")