# SurGBSA: Molecular Dynamics Trajectories and MM/GBSA Scores for PDBbind CASF-2016

## Facts

- Provider: Lawrence Livermore National Laboratory (Jones, Yang, Lightstone, Moshiri, Allen, Rosing)
- License: CC-BY-4.0 (https://huggingface.co/datasets/llnl/SurGBSA/raw/main/README.md)
- Size: 36 files
- Format: CSV (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/llnl/SurGBSA
- Access: public, readable without an account
- DOI: https://doi.org/10.57967/hf/9379
- Category: Drug Discovery
- Source: https://arxiv.org/abs/2509.03084
- Page: https://scigantic.com/archive/f19a8e0b-1480-5d5e-b4b1-36b79210bba2
- Keywords: Drug Discovery, molecular dynamics, benchmark, CSV, Hugging Face

## About

Atomistic molecular dynamics (MD) trajectories for the PDBbind CASF-2016 benchmark, with snapshot-resolved Molecular Mechanics Generalized Born Surface Area (MM/GBSA) binding energy estimates, released by Lawrence Livermore National Laboratory with SurGBSA: Learning Representations From Molecular Dynamics Simulations (arXiv:2509.03084; LLNL-DATA-2016815). 10 ns Amber production simulations cover 242 of the 285 CASF-2016 protein-ligand complexes, each from up to 6 starting points: the crystal structure plus up to 5 AutoDock Vina docking poses. The release includes raw Amber MD files, ML-ready preprocessed arrays in NumPy format, per-snapshot MM/GBSA scores, k-fold cross-validation splits and trained model weights. Listing the repository (measured 2026-11-13) shows md.tar.gz (95.2 GB of raw trajectories), processed_data.tar.gz (4.45 GB), splits/ with 17 coreMD fold CSV files and weights/kfold/ with 15 EGMN model checkpoints; about 101 GB in all. Extract the archives and read trajectories with MDAnalysis, MDTraj or cpptraj, and the processed arrays with NumPy. Uses include surrogate models for binding free energy, protein-ligand affinity prediction and representation learning from MD. Released under CC-BY-4.0; cite the SurGBSA paper and PDBbind.

## Load the data

```python
import pandas as pd
df = pd.read_csv("hf://datasets/llnl/SurGBSA/splits/coreMD-fold-0-test.csv")
df.head()
```

## Data structure

File types in sampled files: CSV

Example file: `splits/coreMD-fold-0-test.csv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/f19a8e0b-1480-5d5e-b4b1-36b79210bba2
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/f19a8e0b-1480-5d5e-b4b1-36b79210bba2
- Over MCP: `get_schema_card` with id `f19a8e0b-1480-5d5e-b4b1-36b79210bba2` at https://api.scigantic.com/mcp

## Related topics

- [CSV datasets](https://scigantic.com/topics/csv)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Drug Discovery datasets](https://scigantic.com/topics/drug-discovery)
- [Molecular dynamics datasets](https://scigantic.com/topics/molecular-dynamics)
