Predicting protein structures at any folding stage from sequence
Abstract
Protein structure predictors reliably identify native conformations, but do not describe the continuum connecting unfolded ensembles to folded states. Here we present FoldScape, a sequence-conditioned deep learning framework that generates Cα conformational ensembles at a specified folding fraction Q. FoldScape reconstructs three-dimensional chains through hierarchical, sliding-window regressions of one-dimensional inter-residue distances, followed by chain assembly and sampling with a rectified-flow decoder. It is trained on multi-scale coarse-grained folding trajectories and all-atom mechanical-unfolding trajectories. Across 20 human proteins (36-857 residues), at matched sample size, FoldScape covers about an order of magnitude more of the folding coordinate than seed-varied AlphaFold 3 ensembles, robustly across proteins, order parameters and binning choices. AlphaFold 3 better anchors the immediate native vicinity, which FoldScape does not reach, making the approaches complementary. FoldScape reconstruction accuracy remains approximately constant along Q, from collapsed coil to near-native globule, rather than deteriorating away from the native state. A stride-one crop-and-superpose scheme extends inference to chains several times longer than a single chain-level forward pass. We also show that training trajectories resolve mechanical unfolding at residue resolution, that elongation-conditioned ensembles exhibit reproducible continuous observables across pulling velocities, and that partially unfolded chains form geometrically coherent segments. FoldScape thus combines broad coverage of folding stages with stable reconstruction accuracy, without per-protein simulation.
Related
- An inverse-protein-folding pre-trained model for predicting the mutational effects in protein-protein interactions
- Structures of protein folding intermediates on the ribosome
- AmyloDeep: protein language model-based ensemble model for predicting amyloid propensity from the amino acid sequence
- Sequence design for three-dimensional genome folding using Akita Semifreddo
- From sequences and structures to fitness, folding and function: challenges in the age of AI
BibTeX
@misc{zhang2026predicting,
title = {{Predicting protein structures at any folding stage from sequence}},
author = {Lisheng Zhang and Qingsong Li and Zinan Xie and Yanting Yang and Qikui Yang and Peng Lu and Zhican Fu and Wentao Cui and Runtong Hou and Zhengwei Xie},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.04287},
eprint = {2611.04287},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.04287},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Predicting protein structures at any folding stage from sequence
AU - Lisheng Zhang
AU - Qingsong Li
AU - Zinan Xie
AU - Yanting Yang
AU - Qikui Yang
AU - Peng Lu
AU - Zhican Fu
AU - Wentao Cui
AU - Runtong Hou
AU - Zhengwei Xie
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.04287
UR - https://arxiv.org/abs/2611.04287
AB - Protein structure predictors reliably identify native conformations, but do not describe the continuum connecting unfolded ensembles to folded states. Here we present FoldScape, a sequence-conditioned deep learning framework that generates Cα conformational ensembles at a specified folding fraction Q. FoldScape reconstructs three-dimensional chains through hierarchical, sliding-window regressions of one-dimensional inter-residue distances, followed by chain assembly and sampling with a rectified-flow decoder. It is trained on multi-scale coarse-grained folding trajectories and all-atom mechanical-unfolding trajectories. Across 20 human proteins (36-857 residues), at matched sample size, FoldScape covers about an order of magnitude more of the folding coordinate than seed-varied AlphaFold 3 ensembles, robustly across proteins, order parameters and binning choices. AlphaFold 3 better anchors the immediate native vicinity, which FoldScape does not reach, making the approaches complementary. FoldScape reconstruction accuracy remains approximately constant along Q, from collapsed coil to near-native globule, rather than deteriorating away from the native state. A stride-one crop-and-superpose scheme extends inference to chains several times longer than a single chain-level forward pass. We also show that training trajectories resolve mechanical unfolding at residue resolution, that elongation-conditioned ensembles exhibit reproducible continuous observables across pulling velocities, and that partially unfolded chains form geometrically coherent segments. FoldScape thus combines broad coverage of folding stages with stable reconstruction accuracy, without per-protein simulation.
ER -