Researchers at IST Austria and international collaborators reported an approach to steer AlphaFold using experimental data to address a key limitation: the model’s tendency to collapse heterogeneous structures into a single dominant conformation. The work, published in Nature Biotechnology, uses experimental signals to guide predicted conformations more faithfully to real biological variability. AlphaFold’s high accuracy has accelerated structural biology, but many protein systems switch between local states depending on conditions. This new “experiment-guided” strategy aims to retain multiple relevant shapes and incorporate the influence of measured experimental context, improving utility for downstream drug discovery and protein engineering. For biotech teams, the advance strengthens the pipeline from sequence to structure by reducing the need to treat predictions as condition-agnostic. It also creates a clearer framework for integrating wet-lab measurements directly into next-generation predictive modeling.