(A) Number of satisfied intra-module cross-links plotted against the score for all 165,523 models. The vertical brown line represents the score threshold used to select the best scoring 500 models. (B) RMSD distance matrix for the 500 best-scoring models. The distance matrix is represented by a heat map, where the bin color is proportional to the RMSD distance of the corresponding solution pair. (C) Difference between the four clusters represented by a hierarchical graph. Clusters 1, 2, and 4 have the middle module oriented in the ‘normal’, expected orientation (i.e., Med1 and Med19 are at the bottom and at the top of the module, respectively). The three clusters differ in the arrangement of the tail subunits. Cluster 3 has the middle module oriented in a ‘flipped’ orientation (i.e., Med1 and Med19 are at the top and at the bottom of the module, respectively). (D) Detailed statistics of cross-link satisfaction for the whole dataset (top), intra-module cross links (middle), and inter-module cross-links (bottom). (E) Student t-test analysis of the intra-module cross-link violation distance. (F) Precision of the four clusters. (G) Statistical properties of the four clusters. (H) Agreement between subunit localization in Cluster 1 (left) and the interpretation of localization data mapped on the EM density map of the Tail module (right).