I applied the FlexPepDock refinement protocol and the FlexPepDock ab-initio protocol to simulate the same peptide-protein complexe(1K5N) and generated a mass of models(refinement n=1000; ab-initio n=50000). The score function of two protocols is score12. The models was ranked based on the reweighted-score.
However, the score of the top models of refinement is better than ab-initio.Sub-angstrom models were sampled in the cases of refinement. Even though ab-initio protocol generated near-native models,their energy score is high.
In ab-initio i don't observe the trend that near-native models have low energy.In most of cases the top models have unqualified rmsd.
The scatter diagram have been attached.
You are here
the flexpepdock refinement and ab-initio