ConforNets: Latents-Based Conformational Control in OpenFold3

2026-04-20

AI Summary

AlphaFold3 struggles to reliably generate multiple biologically relevant conformations of proteins. To address this limitation, this work introduces ConforNets into the OpenFold3 architecture, enabling global and transferable control over protein conformations through channel-wise affine transformations applied to the pair latent representation prior to the Pairformer module. This approach uniquely supports conformational control that can be reused across different proteins, offering both unsupervised multi-conformation generation and supervised conformation transfer capabilities. Experimental results demonstrate that the method achieves state-of-the-art performance in unsupervised generation success rates across all existing multi-conformation benchmarks and successfully enables conformation transfer across diverse protein families.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: 3D Computer Vision
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Models from the AlphaFold (AF) family reliably predict one dominant conformation for most well-ordered proteins but struggle to capture biologically relevant alternate states. Several efforts have focused on eliciting greater conformational variability through ad hoc inference-time perturbations of AF models or their inputs. Despite their progress, these approaches remain inefficient and fail to consistently recover major conformational modes. Here, we investigate both the optimal location and manner-of-operation for perturbing latent representations in the AF3 architecture. We distill our findings in ConforNets: channel-wise affine transforms of the pre-Pairformer pair latents. Unlike previous methods, ConforNets globally modulate AF3 representations, making them reusable across proteins. On unsupervised generation of alternate states, ConforNets achieve state-of-the-art success rates on all existing multi-state benchmarks. On the novel supervised task of conformational transfer, ConforNets trained on one source protein can induce a conserved conformational change across a protein family. Collectively, these results introduce a mechanism for conformational control in AF3-based models.

Problem

• conformational control
• protein conformation
• AlphaFold
• alternate states
• latent representations

Innovation

• ConforNets
• conformational control
• AlphaFold3
• latent representation
• protein dynamics