Section: New Results

SBROD protein quality assessment method

Participants : Mikhail Karasikov, Sergei Grudinin.

Protein quality assessment (QA) is a crucial element of protein structure prediction, a fundamental but yet open problem in structural bioinformatics. QA aims at ranking predicted protein models from a set of proposed candidates. Although consensus-model QA methods often outperform single-model QA methods, their performance substantially depends on the pool of available candidates. This makes single-model QA methods a particularly important research target since these usually assist in the sampling of candidates.

We developed a novel single-model QA method called SBROD. The SBROD (Smooth Backbone-Reliant Orientation-Dependent) method uses only the conformation of the protein backbone, and hence it can be applied to scoring the coarse-grained protein models. The proposed method deduces the scoring function from a training set of protein models. This function is composed of four terms related to different structural features, residue-residue orientations, contacts between the backbone atoms, hydrogen bonding, and solvent-solvate interactions. The SBROD scoring function is smooth with respect to atomic coordinates and thus is applicable to continuous gradient-based optimization of protein conformations. Furthermore, it can also be used for coarse-grained protein modeling and computational protein design. Computational experiments conducted on diverse datasets (Stage1 and Stage2 from CASP11, and MOULDER) proved SBROD to achieve the state-of-the-art performance among single-model QA methods including meta algorithms.

The standalone application implemented in C++ and Python is freely available at https://team.inria.fr/nano-d/software/SBROD and supported on Linux, MacOS, and Windows.