「view this page in B3 βῆτα server」

Revisions №57127

branch: master 「№57127」
Commited by: Roland Pache
GitHub commit link: 「ea3780887273283c」
Difference from previous tested commit:  code diff
Commit date: 2014-07-29 01:10:42

This commit adds a new loop modeling protocol to the Rosetta code-base, KIC with fragments, which combines the power of kinematic closure with sampling coupled degrees of freedom from protein fragment data. In the KIC with fragments protocol, coupled phi/psi/omega torsion angles from consecutive residues in protein fragments are used to sample all but 6 torsional degrees of freedom. Those remaining 6 phi/psi torsions are solved analytically by KIC. This combination allows for rapid sampling of large conformational spaces. The protocol is based on Kale Kundert's refactored KIC code, which provides a more flexible and more easily extendable framework than its legacy counterpart. Our current benchmark results, using the 45 12-residue loops KIC benchmark set, indicate that KIC with fragments retains the sub- Angstrom loop reconstruction accuracy of the next generation KIC protocol, while considerably outperforming previous methods in the sampling of sub-Angstrom models (median fraction of sub-Angstrom models = 33% (without using homologs during fragment generation), compared to only 13% for next generation KIC, 6% for legacy KIC and 1% for CCD). KIC with fragments can be activated for the centroid (perturb) and fullatom (refine) stages using the following flags: -loops:remodel perturb_kic_with_fragments -loops:refine refine_kic_with_fragments Fragment data can be loaded using the standard flags -loops:frag_sizes and -loops:frag_files A comprehensive documentation for KIC with fragments, incl. algorithmic details, protocol usage (such as necessary input files and available options) and a quick-start example can be found at https://www.rosettacommons.org/docs/wiki/application_documentation/KIC_with_fragments This commit also contains a subtle bugfix by Kale in KicMover that was causing chain breaks during initial loop building. That bug only affected the refactored loop modeling framework and in most cases led to reduced performance (in one benchmark case it caused the initial loop build to fail). The newly-added KIC_with_fragments integration test is expected to change. The following integration tests have already failed before and are not affected by this commit: torsion_restricted_sampling centroid_disulfide_scores No other test changes expected. NEWS: Adding a new loop modeling protocol, KIC with fragments, which combines the power of kinematic closure with sampling coupled phi/psi/omega degrees of freedom from protein fragment data.

...