Fig. 7: Predicting E. coli proteins that may switch folds.

a We ran the sequences of 2126 E. coli and phage proteins through CF-random. In total 3237 sequences were attempted, but in 1111 cases, MSAs deep enough for coevolutionary inference could not be generated. Seashell-like image represents these 3237 proteins by length; the inner circle represents 1111 rejected candidates, and outer, the 2126 proteins that were then run through CF-random. If two or more distinct conformations were identified, such as in the case of the successfully identified fold-switching E. coli protein, RfaH, we tested for dual-fold coevolution using ACE. If coevolutionary evidence for both folds was found, the protein was considered a putative fold switcher. Light gray/black contacts on upper/lower diagonals are unique to CF-random predicted dominant/alternative conformations. Teal contacts are from coevolutionary inference. Medium gray contacts are unique to both folds. b Putative fold-switching proteins are involved in diverse functions. C Examples of putative hits. The phage tail tube protein is part of a large assembly that penetrates its host cell envelope. The dominant conformation predicts an extended β-sheet, while in the alternative form, the sheets are incorporated into the larger β-sheet structure. Both conformations were predicted with plDDT scores > 70. NinH is transcription factor protein that may undergo an α-helix-to-β-sheet transition; both conformations were predicted with plDDT scores > 70. Finally, YmcE is a bacterial antitoxin predicted to assume two different folds. Its dominant form was predicted with plDDT > 80, while its alternative was predicted with plDDT > 66 excluding disordered ends. Throughout this figure, dominant folds are blue and alternatives are pink. Source data are provided as a Source Data file.