Influence of Model Structures on Predictors of Protein Stability Changes from Single-Point Mutations

Abstract

Missense variation in genomes can affect protein structure stability and, in turn, the cell physiology behavior. Predicting the impact of those variations is relevant, and the best-performing computational tools exploit the protein structure information. However, most of the current protein sequence variants are unresolved, and comparative or ab initio tools can provide a structure. Here, we evaluate the impact of model structures, compared to experimental structures, on the predictors of protein stability changes upon single-point mutations, where no significant changes are expected between the original and the mutated structures. We show that there are substantial differences among the computational tools. Methods that rely on coarse-grained representation are less sensitive to the underlying protein structures. In contrast, tools that exploit more detailed molecular representations are sensible to structures generated from comparative modeling, even on single-residue substitutions.

Keywords: machine learning; performance evaluation; protein stability; single-point mutation; stability change.

Figure 1

Distribution of $RMSD$ values between residues’ distance matrices of experimental and Modeller/Rosetta predicted protein 3D structures.

Figure 2

Comparative analysis of $\Delta \Delta G$ predictors’ performance (on the total set of mutations) in terms of Pearson correlation, $MAE$, $RMSE$, antisymmetry, and bias. Differential scores between experimental and Modeller/Rosetta-derived structures are displayed for each metrics. The names of the models and their respective absolute metric values on the experimental structures are annotated on the x axis.

Figure 3

$\Delta \Delta G$ values, expressed in kcal/mol, obtained from experimental and from Modeller/Rosetta-predicted structures, for all the models. Pearson correlation coefficients r are annotated in each legend.

Figure 4

Pearson coefficients comparison between $\Delta \Delta G$ predictions from experimental and Modeller/Rosetta structures, which are categorized according to the solvent accessibility of the mutated residue.