MARTINI-Based Protein-DNA Coarse-Grained HADDOCKing

Abstract

Modeling biomolecular assemblies is an important field in computational structural biology. The inherent complexity of their energy landscape and the computational cost associated with modeling large and complex assemblies are major drawbacks for integrative modeling approaches. The so-called coarse-graining approaches, which reduce the degrees of freedom of the system by grouping several atoms into larger "pseudo-atoms," have been shown to alleviate some of those limitations, facilitating the identification of the global energy minima assumed to correspond to the native state of the complex, while making the calculations more efficient. Here, we describe and assess the implementation of the MARTINI force field for DNA into HADDOCK, our integrative modeling platform. We combine it with our previous implementation for protein-protein coarse-grained docking, enabling coarse-grained modeling of protein-nucleic acid complexes. The system is modeled using MARTINI topologies and interaction parameters during the rigid body docking and semi-flexible refinement stages of HADDOCK, and the resulting models are then converted back to atomistic resolution by an atom-to-bead distance restraints-guided protocol. We first demonstrate the performance of this protocol using 44 complexes from the protein-DNA docking benchmark, which shows an overall ~6-fold speed increase and maintains similar accuracy as compared to standard atomistic calculations. As a proof of concept, we then model the interaction between the PRC1 and the nucleosome (a former CAPRI target in round 31), using the same information available at the time the target was offered, and compare all-atom and coarse-grained models.

Keywords: biomolecular complexes; coarse-graining; docking; force field; nucleic acids.

Figure 1

Performance of the all-atom and coarse-grained protocols in HADDOCK on the 44 unbound protein-DNA complexes of the benchmark. (A) Overall success rates (%) of the all-atom protocol on ranking single models as a function of the number of models considered. (B) Same as (A) but for the coarse-grained protocol. (C,D) The quality of the docking models for all 44 cases as a function of the number of models considered. The complexes are ordered by increasing degree of difficulty (from top to bottom) for both all-atom and CG docking runs. The color coding indicates the quality of the docked models according to CAPRI criteria.

Figure 2

Single structure comparison of top-ranking models predicted by HADDOCK. Superimposition of the best models (top-ranked) predicted by HADDOCK using atomistic (blue) or coarse-grained (orange) docking onto the experimental crystal structure (PDB-ID 4r8p, green; McGinty et al., 2014). The two residues PRC1-Cys85 and H2A-Lys119 which are expected to form a covalent bond (Kerscher et al., ; an information used to guide the docking) are shown as spheres. The interface RMSD of the all-atom and coarse-grained top rankings models against the reference crystal structure are 3.23 and 3.0 Å, respectively.