Estimating quality of template‐based protein models by alignment stability

Abstract

The error in protein tertiary structure prediction is unavoidable, but it is not explicitly shown in most of the current prediction algorithms. Estimated error of a predicted structure is crucial information for experimental biologists to use the prediction model for design and interpretation of experiments. Here, we propose a method to estimate errors in predicted structures based on the stability of the optimal target-template alignment when compared with a set of suboptimal alignments. The stability of the optimal alignment is quantified by an index named the SuboPtimal Alignment Diversity (SPAD). We implemented SPAD in a profile-based threading algorithm and investigated how well SPAD can indicate errors in threading models using a large benchmark dataset of 5232 alignments. SPAD shows a very good correlation not only to alignment shift errors but also structure-level errors, the root mean square deviation (RMSD) of predicted structure models to the native structures (i.e. global errors), and local errors at each residue position. We have further compared SPAD with seven other quality measures, six from sequence alignment-based measures and one atomic statistical potential, discrete optimized protein energy (DOPE), in terms of the correlation coefficient to the global and local structure-level errors. In terms of the correlation to the RMSD of structure models, when a target and a template are in the same SCOP family, the sequence identity showed a best correlation to the RMSD; in the superfamily level, SPAD was the best; and in the fold level, DOPE was best. However, in a head-to-head comparison, SPAD wins over the other measures. Next, SPAD is compared with three other measures of local errors. In this comparison, SPAD was best in all of the family, the superfamily and the fold levels. Using the discovered correlation, we have also predicted the global and local error of our predicted structures of CASP7 targets by the SPAD. Finally, we proposed a sausage representation of predicted tertiary structures which intuitively indicate the predicted structure and the estimated error range of the structure simultaneously.

STAP Refinement of the NMR database: a database of 2405 refined solution NMR structures

“…9v7) (19), normalized DOPE (nDOPE, ver. 9v7) (20) and dipolar Distance-scaled, Finite-Ideal gas REference (dDFIRE, ver. 1.1) (21).…”

Prediction of Local Quality of Protein Structure Models Considering Spatial Neighbors in Graphical Models

“…We used the same set of features as our previous work1921. They can be classified into structural, target-template alignment-based, and composite scores.…”

“…Alignment-features used were the amino acid similarity score, which is the average BLOSUM45 score45 of columns in the multiple sequence alignment (MSA); the gap ratio, which is the fraction of gaps in columns in the MSA; conservation, the fraction of the most abundant residue type at columns; PRSS46, which is a Z-score of the target-template alignment score computed against a distribution of the alignment score of shuffled sequences; and four scores derived from SPAD19, which quantifies the consistency of a template-target alignment with a set of suboptimal alignments: local SPAD, which is the SPAD score computed for each residue; global SPAD, which is the average local SPAD (lSPAD) along the entire model; log(SPAD), and log(lSPAD/(SPAD + 1)).…”

“…Global quality of models can be predicted from structural and sequence (target-template alignment) features of models or combinations thereof, often formulated in a machine learning framework. Useful structural features include residue/atom-contact potentials1415, main-chain torsion angles16, burial/exposure propensity of residues, and residue environment propensity17, while alignment features include alignment scores computed for a target and a template sequence/profile the sequence identity, and their statistical significance1819. To combine features, methods such as linear combination20 or regression21, support vector machine (SVM)2223, and neural network242526 have been used.…”

Using Structure to Explore the Sequence Alignment Space of Remote Homologs

“…We calculate the distance between any two alignments using a measure similar to the gALD measure developed by Chen and Kihara [48]. If we plot two alignments of the same two sequences on the dynamic programming matrix (Figure 1, blue and green lines) there is a region between them for which we can calculate the area.…”