Fig. 2: Detailed illustration of iterative flexible fitting.

a Schematic of the domain refinement procedure applied in EMBuild. The main-chain probability map is in transparent magenta, the reference PDB structure is in blue, and the fitted query model is colored from green to red according to the Cα displacements between the query model and the PDB structure. Owing to the conformational deviation between the query model and the PDB structure, the query chain matches poorly to the reference structure as a rigid entirety (bottom left). For better fitting, the query chain is then split into multiple domains, as indicated by numbers 1–6 (top left). At last, all domains are iteratively optimized to achieve their best fitting into the main-chain probability map (top right). As such, the final fitted model shows good consistency with the PDB structure (bottom right). b The iterative building strategy adopted by EMBuild. After each cycle of building, the map regions with fitted structures are removed from further fitting. The remaining chains are iteratively assembled to the complex under the guidance of the updated main-chain probability map.