Upload model_info.txt with huggingface_hub

model_info.txt

@@ -0,0 +1,56 @@
+
+DNA Mutation Pathogenicity Predictor - Model Information
+=========================================================
+
+Model Type: Convolutional Neural Network (CNN)
+Parameters: 191,522
+Input Dimension: 1101
+Output: Binary (Pathogenic/Benign) + Importance Score
+
+DATASET:
+--------
+Source: ClinVar (GRCh38)
+Total Variants: 16,000
+- Pathogenic: 8,000
+- Benign: 8,000
+
+Chromosomes: 1, 2, 3, 5, 6, 7, 11, 13, 17, 19, 22, X
+
+PERFORMANCE:
+-----------
+Test Accuracy: 57.54%
+Test AUC-ROC: 0.6166
+Precision: 0.5705
+Recall: 0.6100
+F1 Score: 0.5896
+
+ARCHITECTURE:
+------------
+Input: 1101-dim encoding
+- Reference sequence: 99bp × 5 nucleotides = 495
+- Mutated sequence: 99bp × 5 nucleotides = 495  
+- Difference mask: 99
+- Mutation type: 12
+
+Layers:
+1. Conv1D (11→64) + BatchNorm + ReLU
+2. Conv1D (64→128) + BatchNorm + ReLU
+3. Conv1D (128→256) + BatchNorm + ReLU
+4. Global Average Pooling
+5. FC layers (256+32 → 128 → 64 → 1)
+6. Importance head (256 → 1)
+
+Outputs:
+- Pathogenic logit (apply sigmoid for probability)
+- Importance score (mutation position explainability)
+
+TRAINING:
+--------
+Optimizer: Adam (lr=0.0005)
+Loss: BCE + 0.1 × Contrastive Importance Loss
+Epochs: 60
+Batch Size: 32
+Split: 70% train / 15% val / 15% test
+
+Created: February 18, 2026
+Framework: PyTorch