ConforNets: Latents-Based Conformational Control in OpenFold3
For protein structure prediction, ConforNets provide an efficient and reusable method to capture biologically relevant alternate conformations, addressing a key limitation of AlphaFold models.
ConforNets introduce channel-wise affine transforms of pre-Pairformer pair latents in AlphaFold3 to generate alternate protein conformations, achieving state-of-the-art success rates on all multi-state benchmarks and enabling supervised conformational transfer across protein families.
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.