Added proper dataset loader, data folder, and README for Hugging Face dataset compatibility

5ht_ki_prediction.py

@@ -0,0 +1,36 @@
+from datasets import DatasetInfo, GeneratorBasedBuilder, SplitGenerator, Split, Value, Features
+import datasets
+import csv
+
+class FiveHTKiPrediction(datasets.GeneratorBasedBuilder):
+    VERSION = datasets.Version("1.0.0")
+
+    def _info(self):
+        return DatasetInfo(
+            description="Curated dataset of serotonin receptor Ki values from PDSP Ki Database.",
+            features=Features({
+                "smiles": Value("string"),
+                "ki": Value("float"),
+                "receptor": Value("string"),
+                "source": Value("string"),
+            }),
+            supervised_keys=None,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_path = dl_manager.download_and_extract("data/KiDatabase.csv")
+        return [SplitGenerator(name=Split.TRAIN, gen_kwargs={"filepath": data_path})]
+
+    def _generate_examples(self, filepath):
+        with open(filepath, encoding="utf-8") as f:
+            reader = csv.DictReader(f)
+            for idx, row in enumerate(reader):
+                try:
+                    yield idx, {
+                        "smiles": row["SMILES"].strip(),
+                        "ki": float(row["ki Val"]) if row["ki Val"] else None,
+                        "receptor": row["Name"].strip(),
+                        "source": row["source"].strip(),
+                    }
+                except Exception:
+                    continue

README.md

@@ -1,16 +1,22 @@
 ---
+datasets:
+  - sarahantgan/5HT_Ki_Prediction
 language:
-- en
+  - en
+license: cc-by-4.0
 tags:
-- bioactivity
-- cheminformatics
-- regression
-- serotonin
-- binding-affinity
-pretty_name: 5-HT Ki Prediction Dataset
+  - bioactivity
+  - cheminformatics
+  - regression
+  - serotonin
+  - binding-affinity
 size_categories:
-- n<1K
+  - 100-1K
+task_categories:
+  - regression
+pretty_name: 5-HT Ki Prediction Dataset
 ---
+
 # Serotonin Receptor (5-HT) Binding Affinity Prediction Dataset
 
 This dataset was curated from the PDSP Ki Database to support training machine learning models that predict binding affinity (Ki in nM) of ligands to serotonin (5-HT) receptors.
@@ -32,15 +38,15 @@ This dataset was curated from the PDSP Ki Database to support training machine l
 
 ### 📈 Model Performance
 
-- **R² Score**: 0.257
-- **RMSE**: 4193.72 nM
+- **R² Score**: `your_R2_here`
+- **RMSE**: `your_RMSE_here` nM
 
 ## Source
 
 - PDSP Ki Database: https://pdsp.unc.edu/databases/kidb.php
 
-## Author
+## Project by 
 
 Sara Hantgan  
 University of Michigan | BIOINF 595 Final Project  
-Winter 2025
+Winter 2025

data/KiDatabase.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93a15227e00534bed1d6e6bc777af679b66eb8180c797fbb18248df4602aeefe
+size 23869637