Guided by the cryo‐EM density map 1 (Appendix Fig S3) and XL‐MS distance restraints (Fig 2C), a partial molecular model of E27 was generated de novo using the program Coot 0.9.5 (Emsley & Cowtan, 2004 (link); Emsley et al, 2010 (link)). In addition, in silico AlphaFold2 structure prediction of E27 was performed (Jumper et al, 2021 (link)) using the ColabFold web server (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb; Mirdita et al, 2022 (link)). The AlphaFold2 output coordinates were in very good agreement with the experimentally determined E27 structure (RMSD of 1.489 Å, 2764 aligned atoms). Accordingly, the model was completed by merging the initial model with the AlphaFold2 template and adjusting backbone and side chain positions where necessary using Coot 0.9.5. DDB1 coordinates were taken from a previously solved DDB1/DCAF1‐CtD complex structure (Banchenko et al, 2021 (link)), rigid body‐fitted and adjusted manually using Coot 0.9.5. The coordinates of the rnSTAT2 CCD were extracted from the AlphaFold protein structure database (https://alphafold.ebi.ac.uk/entry/Q5XI26; Varadi et al, 2022 (link)), rigid body‐fitted and adjusted manually using Coot 0.9.5. As final step, automated real space refinement was performed using the program Phenix 1.19.2–4158 (Liebschner et al, 2019 (link)). The refined DDB1/E27/STAT2 CCD structure was then fitted as rigid body in cryo‐EM density map 2 (Appendix Fig S3). Full‐length rnSTAT2 coordinates (https://alphafold.ebi.ac.uk/entry/Q5XI26) were placed by structural superposition of the CCDs. The STAT2 NTD and TAD were removed, and flexible loops in the SH2 domain were trimmed, because they could not be assigned to the experimental density, indicating that they are flexibly attached and mobile relative to the other STAT2 domains. Lastly, a chain break was introduced between STAT2 residues 315 and 317, and the STAT2 portion encompassing residues 317–701 (DBD, LD and SH2) was once more separately fitted as rigid body, to account for a small movement relative to the CCD. Subsequently, the model was subjected to rigid body and grouped b‐factor real space refinement using Phenix 1.19.2–4158, followed by molecular dynamics flexible fitting using the Namdinator server (Kidmose et al, 2019 ), applying default parameters.