Broken chain sampling (Jumping) is an important tool to improve sampling of proteins with high-contact order beta-strands and to allow resampling of specific beta-strand topologies during RASREC. To form a specific strand-strand contact between two residues i and j, the chain is cut in any location (preferably loop regions) between residues i and j. The two independent chains are then put into contact at residue i and j using one of the relative rigid body orientation that is found in the strand-strand database (rosetta_database/scoring/score_functions/jump_templates_SSpairs_v2.dat). The generalized fragment sampling implemented in ROSETTA3 allows further to sample other relative rigid body orientations from the strand-strand database during the fragment assembly stages of the structure calculation protocols. Fragments taken from the strand-strand database are characterized by their direction (parallels, anti-parallel) and their pleating (CA atoms point inward/outward). Since exchange of conformations between these classes would be too disruptive during folding, we select one of the 4 classes for each jumping position at the beginning of a folding simulation and then exchange fragments for each jumping position according to its class identity.
The position and class of strand-strand contacts can be defined by input files. It is possible to define multiple pairs for jumps including multiple classes (direction, pleating) for the same pair of residues. In this case specific pairs and contact classes are selected randomly at the beginning of each folding trajectory and subsequently kept constant.
The beta-strand topology is the collection of residue pairs that are bonded by backbone hydrogen bonds and for which both residues are in extended conformation(beta-sheet). These pairs are usually organized in parallel or anti-parallel stretches which we call strand-strand contact. Anti-parallel strand-strand contacts, that start with the pairing of residue i, j have also the pairs i+1, j-1, i+2, j-2 etc. Hence, it makes sense to define an invariant quantity called register for each pair i,j as r=j-i for j>i. In parallel strand-strand contacts the register is defined as r=i+j. Sometimes a residue is skipped which creates a bulge and leads to a change of the register within a single strand-strand contact. ROSETTA3 has a condensed file-format to describe such topologies precisely. Such a topology file can be created using the application r_pdb2top using structures in pdb or silent format as input. If multiple structures are used as input, the program creates a topology description for each input structure and additionally scores the topologies such that topologies with high consensus among the input-models are ranked higher. This scoring is described in detail in Ref. [1].
Using the broker-setup file (-broker:setup) one can add a TemplateJumpClaimer to sample specific beta-sheet topologies. Given a topology file (r_pdb2top) with specific and ranked topolgies, the TemplateJumpClaimer will randomly select a topology from the top-ranking topologies (from a topology file as created with r_pdb2top) for each trajectory. From the selected topology it further selects randomly a subset of the strand-strand contacts to be enforced by jumping. For each chosen strand-strand contact one residue-pair is chosen randomly for the actual residue-residue jump. The pleating and direction for this jump is given by the topology. The flag -templates:topology_rank_cutoff controls which topologies from the input file can be chosen.
If beta-strand topologies are not available one can also just define a plain list of residue-residue pairs including their contact class (direction, pleating) to be enforced as strand-strand jumps. If such a file is given, the TemplateJumpClaimer will select jumps from this list and uses the secondary structure information (as provided by a psi-pred file) to avoid selecting incompatible pairs (such as two pairs within the same strand-strand contact). This mechanism is used during the early stage of the RASREC protocol.
For each jump that fixes a strand-strand contact the protocol has to introduce an artificial chainbreak to avoid circularity in the AtomTree. This chainbreak has to be closed before the structure can be refined during the relax stage. To prepare the closing we ramp up a chainbreak energy penalty. During abinitio stages stageI through stageIII the linear_chainbreak energy term is used. During stage IV the overlap_chainbreak is switched on additionally. The weight for the linear-chainbreak energy is computed as W=S*IC, where IC is given by flag -jumps::increase_chainbreaks
with default 1.0, S depends on the stages as follows 0.25/3, 2.5/3*IP, 0.5/3*IP, (1.5*IP+2.5)/3 for stages stageII, stageIIIa, stageIIIb and stageIV, respectively, and IP the intra-stage progress (0...1, for multi-block stages stageIII and stageIV).
References