Uncovering Hidden Functional States in Cryo-EM Datasets with EMPROVE

Single-particle cryo-EM has revolutionised structural biology, but heterogeneous particles, variable local resolution, and orientation bias often obscure secondary structure details and key functional states. Increasing the particle count alone is often insufficient to resolve dynamic regions. We present EMPROVE, an integrative pipeline for particle selection and 3D class reassignment that systematically overcomes these challenges. EMPROVE introduces the Structural Cross-correlation Index to quantify structural consistency at the single-particle level, enabling high-quality subset identification guided by local structural quality, thereby improving relevant conformational states. We applied EMPROVE to unpublished and published (EMPIAR) cryo-EM datasets at 2.6-4.5Å resolution. EMPROVE resolved novel functional motifs and detected previously-uncharacterized conformations in membrane proteins and RNAs of 112-388 kDa. Moreover, it revealed novel poses of ligands bound to the active pockets of medically-relevant biomolecules. EMPROVE advances our understanding of fundamental biological mechanisms, such as splicing and cell signalling, and facilitates structure-based design of pharmaceutical compounds.