A composite score for predicting errors in protein structure models

Abstract

Reliable prediction of model accuracy is an important unsolved problem in protein structure modeling. To address this problem, we studied 24 individual assessment scores, including physics-based energy functions, statistical potentials, and machine learning-based scoring functions. Individual scores were also used to construct ;85,000 composite scoring functions using support vector machine (SVM) regression. The scores were tested for their abilities to identify the most native-like models from a set of 6000 comparative models of 20 representative protein structures. Each of the 20 targets was modeled using a template of <30% sequence identity, corresponding to challenging comparative modeling cases. The best SVM score outperformed all individual scores by decreasing the average RMSD difference between the model identified as the best of the set and the model with the lowest RMSD (DRMSD) from 0.63 Å to 0.45 Å , while having a higher Pearson correlation coefficient to RMSD (r ¼ 0.87) than any other tested score. The most accurate score is based on a combination of the DOPE non-hydrogen atom statistical potential; surface, contact, and combined statistical potentials from MODPIPE; and two PSIPRED/DSSP scores. It was implemented in the SVMod program, which can now be applied to select the final model in various modeling problems, including fold assignment, target-template alignment, and loop modeling.Keywords: model assessment; comparative modeling; fold assignment; statistical potentials; support vector machine; protein structure prediction Supplemental material: see www.proteinscience.org Genomics efforts are providing researchers with the genomes of many species, including Homo sapiens. More difficult tasks lie ahead in annotating, understanding, and modifying the functions of the proteins encoded by these genomes. The structures of proteins aid in these efforts, as the biochemical function of a protein is determined by its structure and dynamics. Atomic structures can be determined for a small subset of proteins by X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy. However, for many proteins of interest, such methods are often costly, time-consuming, and challenging. In the absence of an experimentally determined structure, structure models are often valuable for rationalizing existing evidence and guiding new experiments (Baker and Sali 2001 1653The accuracy of a model determines its utility, making a means of reliably determining the accuracy of a model an important problem in protein structure prediction (Baker and Sali 2001;Ginalski et al. 2005). Model assessment has been previously applied to (1) determine whether or not a model has the correct fold (Miyazawa and Jernigan 1996;Domingues et al. 1999;Melo et al. 2002;McGuffin and Jones 2003), (2) discriminate between the native and near-native states (Lazaridis and Karplus 1999a;Gatchell et al. 2000;Vorobjev and Hermans 2001;Seok et al. 2003;Tsai et al. 2003;Zhu et al. 2003), and (3) select the most native-like model in a set o...

Structure of intact Thermus thermophilus V-ATPase by cryo-EM reveals organization of the membrane-bound V O motor

“…segmentation was performed manually with Qsegment in EMAN (52), and a comparative model of the L 12 − ring was generated with MODELLER (53,54). Automated and semiautomated rigid-body fitting was carried out with Situs Colores (55) and with UCSF Chimera, respectively (56), and all image rendering was performed with UCSF Chimera (56).…”

Mimicking the folding pathway to improve homology-free protein structure prediction

“…As demonstrated below, our Monte Carlo simulated annealing (MCSA) algorithm, using a statistical potential (StatPot) (16)(17)(18)(19) and an increasingly restrictive PDB-based move set ( Fig. 1), recaptures the requisite side-chain information and performs remarkably well in the prediction of 2°and 3°structure without invoking homology information (Figs.…”

“…Because each factor strongly influences the other two, a major challenge lies in simultaneously including all three factors with appropriate weights into a folding algorithm. Because the model retains the backbone heavy atoms and the side-chain C ␤ atoms (16)(17)(18)(19), the 2N backbone dihedral , angles are the major degrees of freedom for a chain of N residues in our treatment (cisconformers are occasionally allowed, see Methods).…”

“…The 3°interaction energies are obtained from the StatPot discrete optimized protein energy (DOPE)-C ␤ (18,26) derived from an all-atom pairwise additive StatPot (16,17) that uses a novel reference state and distinguishes the backbone atoms according to amino acid type. Our version removes all contributions involving hydrogen and side-chain atoms beyond the C ␤ atom.…”

On-line hierarchy of general linear models for selecting and ranking the best predicted protein structures

“…Comparison to The Related Work: we compare our work to the other machine learning based model quality assessment programs, namely: ProQ [3], SVMod [6], and SVR [7]. The following seven aspects highlight the differences and the similarities between our work and the related work: (i) the framework learns a unique function per target protein; however, the other methods learn only one function for all target proteins (ii) ZicoSTP uses a subset of proteins which are related to the input target protein to build online a custom-trained hierarchy of GLM's.…”