To generate a model of the human PM5 C* complex, we first rigid-body–fitted previously published protein and RNA structures from the human spliceosomal C* and P complexes using UCSF Chimera (51 (link)). To obtain a better fit into the EM density map, individual RNAs and proteins and domains thereof were subsequently refitted using UCSF Chimera and manually readjusted in Coot (52 (link)). After an initial round of real-space refinement in Phenix (53 (link)) and a manual optimization in Coot to improve the fit, the map was searched for unassigned elements. Guided by the composition of the PM5 C* complex (fig. S1 and table S1) and CXMS data (data S8), candidate proteins that were cross-linked to the already modeled parts of the complex were selected. Published experimental models or AlphaFold-predicted models (29 (link), 30 (link)) of these candidates were examined and docked using UCSF Chimera. The individual models were manually checked and rebuilt in Coot if the resolution of the map allowed it. Using this approach, we were able to extend the models of several proteins including NKAP, PRP22, SDE2, and SYF2. Furthermore, it was possible to build partial models of ESS2 and NOSIP, as well as to rigid-body dock parts of CXORF56, PPIL3, and TLS1 not localized/modeled in previously published spliceosomal complexes. All AlphaFold models of various structural domains and helices of C* proteins (CXORF56, FAM50A, NKAP, NOSIP, SDE2, TLS1, and ESS2) that were fitted into our PM5 hC* density belong to the confident and very confident classes (as defined by the AlphaFold program). Although some single-stranded stretches (e.g., FAM50A amino acids 118 to 128 and 154 to 168 and NOSIP amino acids 121 to 158) belong to lower confidence classes, they fit well into our hC* EM density and, in some cases, even better into unassigned EM density of the previously published hP complex (see fig. S11). In addition, their location in PM5 C* is supported by protein cross-linking. For PRP22, AlphaFold-predicted models of some of the peripheral parts of its helicase domain (e.g., amino acids 107 to 154, 386 to 501, and 530 to 556) had a lower confidence score, but they fit well into the PM5 hC*, as well as hP densities, and were further supported by protein cross-linking data (see fig. S6). Several nucleotides of the intronic RNA (the PPT loop) were built de novo. The RNA fragment that mimics the 3′ exon was docked using a corresponding RNA element from the S. cerevisiae P complex structure and rebuilt in Coot. The model, excluding its parts located in the less well-resolved peripheral parts of the map, was iteratively refined in Phenix and inspected/adjusted in Coot. The model was validated in Phenix using a cryo-EM validation package (table S4). A summary of the appropriate existing atomic coordinate models used as templates and the procedures used to generate the model is provided in table S5. PyMOL (https://pymol.org/2/) and UCSF Chimera were used to generate the figures.