Ablating Design Choices for Protein Ensemble Generators Across Structural Regimes
Abstract
Generative models of protein conformational ensembles inherit much of their design from structure prediction for folded proteins. Ordered and disordered proteins, however, occupy distinct structural regimes, and we investigate how design choices hold up across them. Using a minimal Cα flow-matching model, we ablate choices of architecture, sequence conditioning, rigid-body alignment and training source, and evaluate each variant against experimental SAXS and NMR chemical-shift data partitioned into ordered, disordered, multi-domain and transitional proteins. Mixture-of-experts routing, a static pair representation and a disorder probe have no detectable effect, a smaller protein language model performs worse on ordered proteins but not on disordered ones, and alignment in the loss degrades agreement with SAXS. Replacing coarse-grained simulations of disordered proteins with fragment-assembled ensembles improves agreement with chemical shifts at a cost to SAXS, while an equal mixture of the two keeps most of the gain at a smaller cost. This points to the training data, rather than the architecture, as the main constraint.
Related
- Bayesian Sampling of Structural Ensembles: The Role of Ensemble-Counting Measures
- AmyloDeep: protein language model-based ensemble model for predicting amyloid propensity from the amino acid sequence
- Multi-Ensemble Reweighting Enables Structural Interpretation of Differential HDX-MS
- Novel hybrid protein scaffold gap filling using weighted machine learning ensemble, beam search, and mass-constrained reranking
- COPAL: An ensemble of protein-ligand co-folding models enriches preferred ligand predictions for the LuxR-family of quorum sensing receptors
BibTeX
@misc{quinn2026ablating,
title = {{Ablating Design Choices for Protein Ensemble Generators Across Structural Regimes}},
author = {Eoin Quinn and Oliver Bent},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.05055},
eprint = {2611.05055},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.05055},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Ablating Design Choices for Protein Ensemble Generators Across Structural Regimes
AU - Eoin Quinn
AU - Oliver Bent
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.05055
UR - https://arxiv.org/abs/2611.05055
AB - Generative models of protein conformational ensembles inherit much of their design from structure prediction for folded proteins. Ordered and disordered proteins, however, occupy distinct structural regimes, and we investigate how design choices hold up across them. Using a minimal Cα flow-matching model, we ablate choices of architecture, sequence conditioning, rigid-body alignment and training source, and evaluate each variant against experimental SAXS and NMR chemical-shift data partitioned into ordered, disordered, multi-domain and transitional proteins. Mixture-of-experts routing, a static pair representation and a disorder probe have no detectable effect, a smaller protein language model performs worse on ordered proteins but not on disordered ones, and alignment in the loss degrades agreement with SAXS. Replacing coarse-grained simulations of disordered proteins with fragment-assembled ensembles improves agreement with chemical shifts at a cost to SAXS, while an equal mixture of the two keeps most of the gain at a smaller cost. This points to the training data, rather than the architecture, as the main constraint.
ER -