Create README.md
Browse filesESM2-GBT: Gradient Boosted Trees on ESM2 Embeddings
📌 Model Overview
The ESM2-GBT model is a Gradient Boosted Trees (GBT) regressor trained on ESM2 embeddings from Meta’s ESM2 protein language model. It is designed for protein-related predictive tasks.
🧪 Available Models:
Model Name Description
ACE2_RBD_ESM2-GBT.json Predicts binding affinity between ACE2 and RBD proteins.
General_ESM2-GBT.json General-purpose GBT model trained on ESM2 embeddings.
🏗 Model Details
• Base Model: ESM2
• Architecture: Gradient Boosted Trees (CatBoostRegressor)
• Framework: CatBoost
• Task: Regression
🧑💻 How to Use
1️⃣ Download Model from Hugging Face
from huggingface_hub import hf_hub_download
# Download ACE2 RBD model
model_path = hf_hub_download(repo_id="hbp5181/ESM2-GBT", filename="ACE2_RBD_ESM2-GBT.json")
2️⃣ Load Model in CatBoost
from catboost import CatBoostRegressor
model = CatBoostRegressor()
model.load_model(model_path, format="json")
# Now ready for predictions!
🔬 Training Details
• Feature Extraction: ESM2 embeddings (33-layer transformer, 650M params)
• Training Algorithm: CatBoost Gradient Boosting
• Dataset: homology wildtype ACE2 RBD complex sequence features
• Evaluation Metrics: RMSE, R^2
📌 Applications
• Binding affinity predictions
💡 Limitations & Considerations
• The model is trained on ESM2 embeddings and is limited by the quality of those embeddings.
• Performance depends on the training dataset used.
• Not a deep-learning model; instead, it leverages GBTs for fast, interpretable predictions.
📄 Citation
👤 Maintainer: hbp5181@psu.edu
📅 Last Updated: February 2025
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
metrics:
|
| 4 |
+
- mae
|
| 5 |
+
- r_squared
|
| 6 |
+
base_model:
|
| 7 |
+
- facebook/esm2_t33_650M_UR50D
|
| 8 |
+
pipeline_tag: tabular-regression
|
| 9 |
+
tags:
|
| 10 |
+
- PLM
|
| 11 |
+
- ESM2
|
| 12 |
+
- Regression
|
| 13 |
+
- GBT
|
| 14 |
+
---
|