Upload ov_metadata.json with huggingface_hub

ov_metadata.json

@@ -0,0 +1,34 @@
+{
+  "dataset_name": "ov",
+  "download_urls": {
+    "OV_mRNA": "https://huggingface.co/datasets/AIBIC/MLOmics/resolve/main/Main_Dataset/Classification_datasets/GS-OV/Original/OV_mRNA.csv",
+    "OV_label_num": "https://huggingface.co/datasets/AIBIC/MLOmics/resolve/main/Main_Dataset/Classification_datasets/GS-OV/Original/OV_label_num.csv"
+  },
+  "download_timestamp": "2026-06-01T13:48:09.093293",
+  "statistics": {
+    "num_samples": 284,
+    "num_features": 14229,
+    "target_stats": {
+      "class_mapping": {
+        "Immunoreactive": 0,
+        "Mesenchymal": 1,
+        "Proliferative": 2,
+        "Differentiated": 3
+      },
+      "num_classes": 4,
+      "class_names": [
+        "Immunoreactive",
+        "Mesenchymal",
+        "Proliferative",
+        "Differentiated"
+      ],
+      "samples_per_class": {
+        "Immunoreactive": 64,
+        "Mesenchymal": 79,
+        "Proliferative": 67,
+        "Differentiated": 74
+      }
+    }
+  },
+  "preprocessing_notes": "MLOmics log-transformed FPKM values (no additional log transform). Dropped genes with >50% zero values in a random 70% training split. Duplicate gene symbols are dropped keeping first occurrence (1-Mar, 2-Mar). Parquet feature names replace dots with underscores; gene map string_id retains original HGNC symbols."
+}