We show that this approach is promising towards addressing the above issues, namely, folding trajectory summarization, folding events detection and ordering, and consensus partial folding pathway identification across trajectories.
For instance, the α-motif corresponds to three types of bit-patterns 4, 7, and 9. Figure 9 shows a segment in each summarized BBA5 folding trajectory before and after being generalized with 3D motifs.
The Q53D and Q53T mutants also refolded faster than FKBP∗ but lacked the folding intermediate, indicating that these mutants experienced a different folding trajectory and transition state than FKBP∗ and Q53N.
This is a learning optimal control algorithm whereby the Gaussian network is slowly learned as contacts are made at each folding step and the entries of the contact map are filled dynamically along the optimal folding trajectory.
The eight different segments of the two-dimensional version of the protein folding trajectory are presented in the figure.
These 'hidden intermediates' cannot be observed using common spectroscopic techniques in bulk kinetic experiments, as they are usually populated after the rate-limiting transition state on the folding trajectory.
