---
license: mit
tags:
- biology
- protein
- structure
---

# Fusion AFDB Quality Dataset

Synthetic demonstration dataset for FusionUncertaintyNet.

- 500 synthetic proteins (30-250 aa)
- Fields: sequence, target (0-100 per residue, synthetic lDDT-like), plddt, phi, psi
- Real training would use AFdb 501k + PDB alignment with UniRef clustering split
- Uploaded from Kaggle P100 pipeline demo

This dataset is used to train Adaptive Fusion + EDR heads.