“…Specifically, we tested different combinations of predicted models by Alpha-Fold (Jumper et al, 2021) from CASP14 (Kryshtafovych, Schwede et al, 2021) in three cases: full-length model, multiple single-domain models and individual single-domain model. In this strategy, we performed MR using Phaser (McCoy et al, 2007) in CCP4 (Winn et al, 2011), followed by phases extension using OASIS (Zhang et al, 2010), density modification using DM (Cowtan, 1994) or Parrot (Cowtan, 2010), and alternative model building using Phenix.AutoBuild (Terwilliger et al, 2008) and Buccaneer (Cowtan, 2006) within the framework of IPCAS 2.0 (Ding et al, 2020). Our results demonstrate that our approach effectively corrects model bias introduced by the predicted models and improves the final structures.…”