DCFold: Efficient Protein Structure Generation with Single Forward Pass

2026-05-18

AI Summary

This work addresses the computational inefficiency of AlphaFold3, whose iterative diffusion architecture impedes its application in time-sensitive tasks such as virtual screening and protein design. To overcome this limitation, the authors propose DCFold, a single-step forward generative model that achieves structural prediction accuracy comparable to AlphaFold3 through a single inference pass. DCFold leverages an innovative Dual Consistency training framework and a Temporal Geodesic Matching scheduler, combined with all-atom coordinate modeling, to enable rapid yet precise structure generation. Experimental results demonstrate that DCFold attains a 15-fold acceleration in inference speed while preserving high prediction accuracy, with strong performance validated on standard benchmarks for both protein structure prediction and complex design.

Abstract

AlphaFold3 introduces a diffusion-based architecture that elevates protein structure prediction to all-atom resolution with improved accuracy. This state-of-the-art performance has established AlphaFold3 as a foundation model for diverse generation and design tasks. However, its iterative design substantially increases inference time, limiting practical deployment in downstream settings such as virtual screening and protein design. We propose DCFold, a single-step generative model that attains AlphaFold3-level accuracy. Our Dual Consistency training framework, which incorporates a novel Temporal Geodesic Matching (TGM) scheduler, enables DCFold to achieve a 15x acceleration in inference while maintaining predictive fidelity. We validate its effectiveness across both structure prediction and binder design benchmarks.

Problem

• protein structure prediction
• diffusion model
• inference acceleration
• virtual screening
• protein design

Innovation

• DCFold
• Single Forward Pass
• Dual Consistency
• Temporal Geodesic Matching
• Protein Structure Generation