ConforNets introduces a novel approach to controlling conformational variability in protein structures, leveraging latent-based methods to improve upon the limitations of AlphaFold models. By integrating latent-based conformational control into OpenFold3, researchers can more effectively capture biologically relevant alternate states, addressing a key challenge in protein structure prediction. Previous efforts to elicit greater conformational variability through inference-time perturbations have been inefficient and limited in their impact1. The ConforNets method offers a more efficient and effective solution, with potential applications in fields such as structural biology and drug discovery. This breakthrough has significant implications for the field of protein structure prediction, enabling researchers to better understand the complex relationships between protein structure and function. So what matters to practitioners is that ConforNets' ability to capture alternate states can reveal new insights into protein behavior, ultimately informing the development of more effective therapeutic strategies.