Beyond rotamers: a generative, probabilistic model of side chains in proteins

Abstract

BackgroundAccurately covering the conformational space of amino acid side chains is essential for important applications such as protein design, docking and high resolution structure prediction. Today, the most common way to capture this conformational space is through rotamer libraries - discrete collections of side chain conformations derived from experimentally determined protein structures. The discretization can be exploited to efficiently search the conformational space. However, discretizing this naturally continuous space comes at the cost of losing detailed information that is crucial for certain applications. For example, rigorously combining rotamers with physical force fields is associated with numerous problems.ResultsIn this work we present BASILISK: a generative, probabilistic model of the conformational space of side chains that makes it possible to sample in continuous space. In addition, sampling can be conditional upon the protein's detailed backbone conformation, again in continuous space - without involving discretization.ConclusionsA careful analysis of the model and a comparison with various rotamer libraries indicates that the model forms an excellent, fully continuous model of side chain conformational space. We also illustrate how the model can be used for rigorous, unbiased sampling with a physical force field, and how it improves side chain prediction when used as a pseudo-energy term. In conclusion, BASILISK is an important step forward on the way to a rigorous probabilistic description of protein structure in continuous space and in atomic detail.

New Dynamic Rotamer Libraries: Data-Driven Analysis of Side-Chain Conformational Propensities

“…Another discrepancy raised by Shapovalov and Dunbrack (2011) as a critique of BASILISK (Harder et al, 2010) was the ranking of the Ser rotamers; the authors stated that the Ser g− rotamer was not the most probable. However, in agreement with Harder et al, we find that the g − rotamer of Ser is dominant (75%, BBIND) (Table S1; Figures 7, 8, and 9).…”

Generative Models of Conformational Dynamics

“…Subsequent uses of graphical models for proteins have considered a wide range of problems, including density fitting [7], structure prediction [4, 14], protein design [12, 11, 3, 33, 49], free energy calculations [18], and predicting resistance mutations [20]. Their growing popularity in structural biology is due to their ability to represent complex distributions and solve challenging inference problems.…”

Mixtures of concentrated multivariate sine distributions with applications to bioinformatics

“…basic building blocks of protein 3D structures. On the other hand, Harder et al [6] proposed a dynamic Bayesian network model to simulate the conformation space of the protein dihedral angles. These mixture models have potential to improve their current model in the way that each amino acid is modelled by a mixture model rather than by four independent von Mises distributions.…”

“…Our ILE data contain 2000 observations with 4 angles; these are a random sample from a large data set used in [6]. It is expected from the geometry of the ILE that there are nearly 18 groups, as the means of the components 'sit' on a nearly regular lattice.…”

“…An obvious example is where X-ray determination of protein structures is validated through an examination of the corresponding dihedral angles using PROCHECK [8]. Other examples, which explore protein conformational space through the adoption of bivariate or univariate von Mises distributions, can be found in [2,6].…”