henhaobio
Sign in

bioRxivBioinformaticsPreprint

EquiFold: Protein Structure Prediction with a Novel Coarse-Grained Structure Representation

Jae Hyeon LeeiD, Payman YadollahpouriD, Andrew M. WatkinsiD, Nathan C. FreyiD, Andrew Leaver‐FayiD, Stephen RaiD, Kyunghyun ChoiD, Vladimir GligorijevićiD, Aviv RegeviD, Richard BonneauiD

PDFDOI
Structured summary available · Jump to summary ↓

Abstract

Designing proteins to achieve specific functions often requires in silico modeling of their properties at high throughput scale and can significantly benefit from fast and accurate protein structure prediction. We introduce EquiFold, a new end-to-end differentiable, SE(3)-equivariant, all-atom protein structure prediction model. EquiFold uses a novel coarse-grained representation of protein structures that does not require multiple sequence alignments or protein language model embeddings, inputs that are commonly used in other state-of-the-art structure prediction models. Our method relies on geometrical structure representation and is substantially smaller than prior state-of-the-art models. In preliminary studies, EquiFold achieved comparable accuracy to AlphaFold but was orders of magnitude faster. The combination of high speed and accuracy make EquiFold suitable for a number of downstream tasks, including protein property prediction and design.

Abstract summary

Claude Opus 5.5 · xhigh effort · 2 Nov 2026 · from the abstract · CC BY-NC 4.0

EquiFold predicts all-atom protein structures using an SE(3)-equivariant coarse-grained geometric representation without sequence alignments or language model embeddings, matching AlphaFold accuracy orders of magnitude faster in preliminary tests.

High-throughput in silico protein modeling and design require fast, accurate structure prediction, but existing state-of-the-art predictors are large and depend on multiple sequence alignments (MSAs) or protein language model embeddings. EquiFold replaces these inputs with a compact, end-to-end differentiable SE(3)-equivariant model built on a coarse-grained geometric representation.

Key results

  1. SuggestedIn preliminary studies, EquiFold achieves all-atom structure prediction accuracy comparable to AlphaFold while running orders of magnitude faster and using a substantially smaller model without MSAs or protein language model embeddings.

    Preliminary evaluations comparing EquiFold against AlphaFold showed comparable accuracy and orders-of-magnitude faster runtime. · Abstract

Methods, data and reproducibility

Developed EquiFold, an end-to-end differentiable, SE(3)-equivariant all-atom protein structure prediction model that uses a coarse-grained geometric structure representation instead of multiple sequence alignments or protein language model embeddings.

Methods & tools
EquiFold, AlphaFold
Limitations (2)
  • The evaluations are described as preliminary studies, and the abstract does not report benchmark datasets, quantitative structural accuracy scores, or exact runtime measurements. · noted in this summary
  • No public code or model weights URL is provided in the abstract. · noted in this summary

Read this if you work on high-throughput protein design or property prediction and need an MSA-free, SE(3)-equivariant all-atom structure predictor that runs much faster than AlphaFold.

Citations
47

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISAlQuraishi '19Kroll '26Seo '26Chowdhury '22Roney '22Li '24
76% semantic matchBioinformatics
Mohammed AlQuraishi·Cell Systems·Apr 1, 2019· 371 citations