Prediction of structures of multidomain proteins from structures of the individual domains

Abstract

We describe the development of a method for assembling structures of multidomain proteins from structures of isolated domains. The method consists of an initial low-resolution search in which the conformational space of the domain linker is explored using the Rosetta de novo structure prediction method, followed by a high-resolution search in which all atoms are treated explicitly and backbone and side chain degrees of freedom are simultaneously optimized. The method recapitulates, often with very high accuracy, the structures of existing multidomain proteins.Keywords: domain assembly; protein-protein docking; protein structure prediction Proteins are frequently composed of multiple domains (Ponting and Russell 2002;Vogel et al. 2004) that are likely to fold independently (Shen et al. 2005). Determining the structure of multidomain complexes at atomic resolution is critical to understanding the underpinnings of much of biology (Lupas et al. 2001;Aloy and Russell 2006). While structures of single domains can be readily determined through X-ray or NMR techniques, the structures of large multipart proteins are often more difficult to elucidate (Aloy et al. 2003).There are two general approaches to predicting structures of multidomain proteins from structures of individual domains. First, the domain assembly problem may be treated as a docking problem. For example, Inbar et al. (2005) used rigid body docking methods to predict the structure of the resulting complex. A second approach to domain assembly, which we describe here, is to explicitly sample the degrees of freedom of the linker rather than the rigid body degrees of freedom of the two domains. Approached in this manner, the domain assembly problem may be viewed as an ab initio prediction problem for a relatively short amino acid sequence with preformed N-and C-terminal structures.The Rosetta protein modeling method has had success in folding small protein chains ab initio , and in protein-protein docking with flexible side chains ). Here we combine these methods to assemble structures of isolated domains into a multidomain complex. The conformation of the linker is explored, keeping the backbone of the individual domains fixed but allowing the side chains in the linker and at the domain interface to sample a full range of rotamer conformations. The lowest energy models found are often very close to the correct structure. Results and DiscussionSeventy-six two-domain proteins were culled from a nonredundant database of proteins (Berman et al. 2000), as described in Materials and Methods. These proteins contained no cofactors or ligands near the interface of the domains, as the focus was on modeling the interface Abbreviations: RMSD, root mean square deviation. Article published online ahead of print. Article and publication date are at

Structure prediction for CASP8 with all‐atom refinement using Rosetta

“…For T0416 domain 2(server‐only GDT‐TS Z‐score is 3.0) and T0464(server‐only GDT‐TS Z‐score is 2.41), increasing the sampling for modeling long loops may have been a factor in their successful predictions. T0416 domain 2 is a domain insertion and was modeled as a long loop in the context of a BLAST template, and T0464 contained a 33 residue insertion.T0462(server‐only Z‐score is 3.61) consisted of two domains modeled with different templates that were successfully assembled by Robetta's domain assembly method 33. Among all predictors Robetta's best model for T0449 was the top performer with a GDT‐TS Z‐score of 1.70 and 3.10 among human/server and server only predictors, respectively.…”

Structure prediction for CASP7 targets using extensive all-atom refinement with Rosetta@home

“…Inspection of the secondary structure prediction and domain boundary predictions suggested an approximately symmetric conformation with two ferredoxin‐like folds (cf. T0272 in CASP61); assembling models of the two domains26 led to a prediction with an accuracy of 5.1 Å over 180 residues [Fig. 5(A)].…”

RosettaRemodel: A Generalized Framework for Flexible Backbone Protein Design

“…For a simple linear two-domain assembly [43], one can simply define the linear region with its appropriate secondary structures and run RosettaRemodel with the -no_jumps flag to fold through the linker to sample the degrees of freedom. We offer here a recipe for a more complicated scenario that combines the use of several features in RosettaRemodel.…”