Fig. 3

Circularly-permuted models constructed by several modeling methods for proteins requiring a long termini linker. The CPM structure of a β-glucanase has been determined [24]. In this experiment, the 17 N-terminal residues of the native β-glucanase were removed to test the linker design capability of CirPred. The model constructed by CirPred was complete, and the designed linker was similar to the removed 17-residue fragment both in structure and sequence. Conventional modeling methods either twisted the structure to connect the native termini or failed to fill the gap in between. a Crystal structure of the native 1,3–1,4-β-glucanase (PDB 2ayhA). The yellow fragment corresponds to the 17 N-terminal residues for truncation. b The N-terminus truncated β-glucanase (denoted β-glucanase_∆1–17). c Crystal structure of the CPM59 of the β-glucanase (PDB 1cpmA). The yellow fragment (and sequence) is equivalent to the truncated part of the native protein and was the target linker to be redesigned. d Model constructed by CirPred using b as the template. The yellow fragment is the redesigned linker, whose sequence similarity to the native linker is 70.6% (aligned by Stretcher [32]). e Model constructed by SWISS-MODEL using b as the template. The C-terminal proportion of the structure was missing. f Model constructed by SWISS-MODEL using c as the template. The longest main-frame β-strand was twisted to connect the native termini. g Model constructed by RaptorX using c as the template. As the dotted line indicates, an unreasonable gap remained between the native termini. h Model constructed by Robetta using c as the template. The direct connection between the native termini twisted the longest main-frame β-strand